Compare commits
| Author | SHA1 | Date | |
|---|---|---|---|
 ddidderr
|
4f6a8f0185
|
||
|
ddidderr
|
eecd2e168a
|
@@ -34,7 +34,6 @@ install/
|
||||
|
||||
# Build artefacts
|
||||
/rust/target/
|
||||
/rust/cli/target/
|
||||
contrib/linux-kernel/linux/
|
||||
projects/
|
||||
bin/
|
||||
|
||||
@@ -169,18 +169,6 @@ elseif(_zstd_huf_force_x2)
|
||||
endif()
|
||||
endif()
|
||||
|
||||
# This CMake build compiles every lib/legacy/zstd_v0N.c whenever legacy
|
||||
# support is enabled (ZSTD_LEGACY_LEVEL only selects the C dispatch), so the
|
||||
# Rust archive enables every per-version legacy feature to match. The build
|
||||
# configuration encodes the switch so archives never mix.
|
||||
set(_zstd_rust_legacy 0)
|
||||
if(ZSTD_LEGACY_SUPPORT)
|
||||
set(_zstd_rust_legacy 1)
|
||||
list(APPEND _zstd_rust_features
|
||||
legacy-v01 legacy-v02 legacy-v03 legacy-v04
|
||||
legacy-v05 legacy-v06 legacy-v07)
|
||||
endif()
|
||||
|
||||
set(_zstd_rust_target "${ZSTD_RUST_TARGET}")
|
||||
if(NOT _zstd_rust_target AND CMAKE_SYSTEM_NAME STREQUAL "Linux"
|
||||
AND CMAKE_SIZEOF_VOID_P EQUAL 4)
|
||||
@@ -207,7 +195,7 @@ if(_zstd_rust_features)
|
||||
endif()
|
||||
|
||||
set(_zstd_rust_build_config
|
||||
"c${_zstd_rust_compression}-d${_zstd_rust_decompression}-b${_zstd_rust_dictbuilder}-${_zstd_rust_huf_mode}-legacy${_zstd_rust_legacy}")
|
||||
"c${_zstd_rust_compression}-d${_zstd_rust_decompression}-b${_zstd_rust_dictbuilder}-${_zstd_rust_huf_mode}")
|
||||
set(ZSTD_RUST_MANIFEST "${ZSTD_SOURCE_DIR}/rust/Cargo.toml")
|
||||
set(ZSTD_RUST_TARGET_DIR
|
||||
"${CMAKE_CURRENT_BINARY_DIR}/rust-target/${_zstd_rust_build_config}")
|
||||
|
||||
@@ -88,16 +88,6 @@ elif rust_huf_force_x2
|
||||
rust_huf_c_args += '-DHUF_FORCE_DECOMPRESS_X2'
|
||||
endif
|
||||
|
||||
# Mirror the legacy source selection below: legacy_level N compiles
|
||||
# lib/legacy/zstd_v0N.c .. zstd_v07.c, so the Rust archive enables the
|
||||
# matching per-version features. The build configuration encodes the level
|
||||
# so archives built for different legacy levels never mix.
|
||||
foreach i : [1, 2, 3, 4, 5, 6, 7]
|
||||
if legacy_level != 0 and legacy_level <= i
|
||||
rust_features += 'legacy-v0@0@'.format(i)
|
||||
endif
|
||||
endforeach
|
||||
|
||||
rust_target = get_option('zstd_rust_target')
|
||||
if rust_target == ''
|
||||
if host_machine_os == os_linux and \
|
||||
@@ -108,7 +98,7 @@ if rust_target == ''
|
||||
endif
|
||||
endif
|
||||
|
||||
rust_build_config = 'c1-d1-b1-' + rust_huf_mode + '-legacy@0@'.format(legacy_level)
|
||||
rust_build_config = 'c1-d1-b1-' + rust_huf_mode
|
||||
rust_target_dir = join_paths(meson.current_build_dir(), 'rust-target', rust_build_config)
|
||||
is_msvc = cc_id == compiler_msvc or cc_id == 'clang-cl'
|
||||
rust_staticlib_name = is_msvc ? 'zstd_rs.lib' : 'libzstd_rs.a'
|
||||
|
||||
@@ -104,18 +104,6 @@ ifneq ($(ZSTD_LIB_DECOMPRESSION),0)
|
||||
RUST_CARGO_FEATURES += $(RUST_HUF_FEATURE)
|
||||
endif
|
||||
endif
|
||||
|
||||
# Legacy decoders follow the ZSTD_LEGACY_FILES selection exactly: level N
|
||||
# enables versions v0.N .. v0.7 (0 disables legacy). The build directory
|
||||
# also encodes the level, so archives for different legacy levels never mix.
|
||||
RUST_LEGACY_FEATURES :=
|
||||
ifneq ($(ZSTD_LEGACY_SUPPORT), 0)
|
||||
ifeq ($(shell test $(ZSTD_LEGACY_SUPPORT) -lt 8; echo $$?), 0)
|
||||
RUST_LEGACY_FEATURES := $(addprefix legacy-v0,$(wordlist $(ZSTD_LEGACY_SUPPORT),7,1 2 3 4 5 6 7))
|
||||
endif
|
||||
endif
|
||||
RUST_CARGO_FEATURES += $(RUST_LEGACY_FEATURES)
|
||||
RUST_BUILD_CONFIG := $(RUST_BUILD_CONFIG)-legacy$(ZSTD_LEGACY_SUPPORT)
|
||||
RUST_CARGO_FEATURES := $(subst $(space),$(comma),$(strip $(RUST_CARGO_FEATURES)))
|
||||
|
||||
RUST_TARGET ?=
|
||||
|
||||
+271
-106
@@ -38,62 +38,6 @@ size_t ZSTD_rust_writeFrameHeader(void* dst, size_t dstCapacity,
|
||||
U32 windowLog, U64 pledgedSrcSize,
|
||||
U32 dictID);
|
||||
|
||||
/* Context-free compression-parameter selection and sizing leaves live in
|
||||
* Rust (rust/src/zstd_compress_params.rs). This file retains
|
||||
* configuration-sensitive policy: the excluded-block-compressor strategy
|
||||
* cascade, private ZSTD_CCtx_params handling, and sanitizer workspace
|
||||
* policy, which it feeds to the leaves as explicit scalar inputs. The
|
||||
* ZSTD_CParamMode_e and ZSTD_ParamSwitch_e enums are passed as int; the
|
||||
* Rust side mirrors their values. */
|
||||
int ZSTD_rust_params_maxCLevel(void);
|
||||
int ZSTD_rust_params_minCLevel(void);
|
||||
int ZSTD_rust_params_defaultCLevel(void);
|
||||
ZSTD_bounds ZSTD_rust_params_getBounds(int param);
|
||||
size_t ZSTD_rust_params_checkCParams(ZSTD_compressionParameters cParams);
|
||||
ZSTD_compressionParameters
|
||||
ZSTD_rust_params_clampCParams(ZSTD_compressionParameters cParams);
|
||||
U32 ZSTD_rust_params_cycleLog(U32 hashLog, int strategy);
|
||||
ZSTD_compressionParameters
|
||||
ZSTD_rust_params_selectCParams(int compressionLevel, U64 srcSizeHint,
|
||||
size_t dictSize, int mode);
|
||||
ZSTD_compressionParameters
|
||||
ZSTD_rust_params_adjustCParams(ZSTD_compressionParameters cParams, U64 srcSize,
|
||||
size_t dictSize, int mode,
|
||||
int useRowMatchFinder);
|
||||
ZSTD_parameters ZSTD_rust_params_makeParams(ZSTD_compressionParameters cParams);
|
||||
size_t ZSTD_rust_params_maxNbSeq(size_t blockSize, U32 minMatch,
|
||||
int useSequenceProducer);
|
||||
size_t ZSTD_rust_params_resolveMaxBlockSize(size_t maxBlockSize);
|
||||
|
||||
typedef struct {
|
||||
U32 hashLog3Max;
|
||||
size_t matchTSize;
|
||||
size_t optimalTSize;
|
||||
size_t asanRedzoneSize;
|
||||
} ZSTD_rustMatchStateSizing;
|
||||
size_t ZSTD_rust_params_estimateMatchStateSize(
|
||||
ZSTD_compressionParameters cParams, int useRowMatchFinder,
|
||||
int enableDedicatedDictSearch, U32 forCCtx,
|
||||
const ZSTD_rustMatchStateSizing* sizing);
|
||||
|
||||
typedef struct {
|
||||
size_t cdictSize;
|
||||
size_t hufWorkspaceSize;
|
||||
U32 hashLog3Max;
|
||||
size_t matchTSize;
|
||||
size_t optimalTSize;
|
||||
size_t asanRedzoneSize;
|
||||
} ZSTD_rustCDictSizing;
|
||||
size_t ZSTD_rust_params_estimateCDictSizeFromCParams(
|
||||
size_t dictSize, ZSTD_compressionParameters cParams,
|
||||
int dictLoadMethod, const ZSTD_rustCDictSizing* sizing);
|
||||
|
||||
#if ZSTD_ADDRESS_SANITIZER && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
|
||||
# define ZSTD_RUST_ASAN_REDZONE_SIZE ((size_t)ZSTD_CWKSP_ASAN_REDZONE_SIZE)
|
||||
#else
|
||||
# define ZSTD_RUST_ASAN_REDZONE_SIZE ((size_t)0)
|
||||
#endif
|
||||
|
||||
/* ***************************************************************
|
||||
* Tuning parameters
|
||||
*****************************************************************/
|
||||
@@ -332,7 +276,11 @@ static int ZSTD_resolveExternalSequenceValidation(int mode) {
|
||||
|
||||
/* Resolves maxBlockSize to the default if no value is present. */
|
||||
static size_t ZSTD_resolveMaxBlockSize(size_t maxBlockSize) {
|
||||
return ZSTD_rust_params_resolveMaxBlockSize(maxBlockSize);
|
||||
if (maxBlockSize == 0) {
|
||||
return ZSTD_BLOCKSIZE_MAX;
|
||||
} else {
|
||||
return maxBlockSize;
|
||||
}
|
||||
}
|
||||
|
||||
static ZSTD_ParamSwitch_e ZSTD_resolveExternalRepcodeSearch(ZSTD_ParamSwitch_e value, int cLevel) {
|
||||
@@ -472,17 +420,45 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
|
||||
|
||||
switch(param)
|
||||
{
|
||||
/* The compression level and the seven core compression parameters are
|
||||
* bounded by the Rust leaf. */
|
||||
case ZSTD_c_compressionLevel:
|
||||
bounds.lowerBound = ZSTD_minCLevel();
|
||||
bounds.upperBound = ZSTD_maxCLevel();
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_windowLog:
|
||||
bounds.lowerBound = ZSTD_WINDOWLOG_MIN;
|
||||
bounds.upperBound = ZSTD_WINDOWLOG_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_hashLog:
|
||||
bounds.lowerBound = ZSTD_HASHLOG_MIN;
|
||||
bounds.upperBound = ZSTD_HASHLOG_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_chainLog:
|
||||
bounds.lowerBound = ZSTD_CHAINLOG_MIN;
|
||||
bounds.upperBound = ZSTD_CHAINLOG_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_searchLog:
|
||||
bounds.lowerBound = ZSTD_SEARCHLOG_MIN;
|
||||
bounds.upperBound = ZSTD_SEARCHLOG_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_minMatch:
|
||||
bounds.lowerBound = ZSTD_MINMATCH_MIN;
|
||||
bounds.upperBound = ZSTD_MINMATCH_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_targetLength:
|
||||
bounds.lowerBound = ZSTD_TARGETLENGTH_MIN;
|
||||
bounds.upperBound = ZSTD_TARGETLENGTH_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_strategy:
|
||||
return ZSTD_rust_params_getBounds((int)param);
|
||||
bounds.lowerBound = ZSTD_STRATEGY_MIN;
|
||||
bounds.upperBound = ZSTD_STRATEGY_MAX;
|
||||
return bounds;
|
||||
|
||||
case ZSTD_c_contentSizeFlag:
|
||||
bounds.lowerBound = 0;
|
||||
@@ -1409,7 +1385,14 @@ size_t ZSTD_CCtx_reset(ZSTD_CCtx* cctx, ZSTD_ResetDirective reset)
|
||||
@return : 0, or an error code if one value is beyond authorized range */
|
||||
size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
|
||||
{
|
||||
return ZSTD_rust_params_checkCParams(cParams);
|
||||
BOUNDCHECK(ZSTD_c_windowLog, (int)cParams.windowLog);
|
||||
BOUNDCHECK(ZSTD_c_chainLog, (int)cParams.chainLog);
|
||||
BOUNDCHECK(ZSTD_c_hashLog, (int)cParams.hashLog);
|
||||
BOUNDCHECK(ZSTD_c_searchLog, (int)cParams.searchLog);
|
||||
BOUNDCHECK(ZSTD_c_minMatch, (int)cParams.minMatch);
|
||||
BOUNDCHECK(ZSTD_c_targetLength,(int)cParams.targetLength);
|
||||
BOUNDCHECK(ZSTD_c_strategy, (int)cParams.strategy);
|
||||
return 0;
|
||||
}
|
||||
|
||||
/** ZSTD_clampCParams() :
|
||||
@@ -1418,14 +1401,63 @@ size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
|
||||
static ZSTD_compressionParameters
|
||||
ZSTD_clampCParams(ZSTD_compressionParameters cParams)
|
||||
{
|
||||
return ZSTD_rust_params_clampCParams(cParams);
|
||||
# define CLAMP_TYPE(cParam, val, type) \
|
||||
do { \
|
||||
ZSTD_bounds const bounds = ZSTD_cParam_getBounds(cParam); \
|
||||
if ((int)val<bounds.lowerBound) val=(type)bounds.lowerBound; \
|
||||
else if ((int)val>bounds.upperBound) val=(type)bounds.upperBound; \
|
||||
} while (0)
|
||||
# define CLAMP(cParam, val) CLAMP_TYPE(cParam, val, unsigned)
|
||||
CLAMP(ZSTD_c_windowLog, cParams.windowLog);
|
||||
CLAMP(ZSTD_c_chainLog, cParams.chainLog);
|
||||
CLAMP(ZSTD_c_hashLog, cParams.hashLog);
|
||||
CLAMP(ZSTD_c_searchLog, cParams.searchLog);
|
||||
CLAMP(ZSTD_c_minMatch, cParams.minMatch);
|
||||
CLAMP(ZSTD_c_targetLength,cParams.targetLength);
|
||||
CLAMP_TYPE(ZSTD_c_strategy,cParams.strategy, ZSTD_strategy);
|
||||
return cParams;
|
||||
}
|
||||
|
||||
/** ZSTD_cycleLog() :
|
||||
* condition for correct operation : hashLog > 1 */
|
||||
U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
|
||||
{
|
||||
return ZSTD_rust_params_cycleLog(hashLog, (int)strat);
|
||||
U32 const btScale = ((U32)strat >= (U32)ZSTD_btlazy2);
|
||||
return hashLog - btScale;
|
||||
}
|
||||
|
||||
/** ZSTD_dictAndWindowLog() :
|
||||
* Returns an adjusted window log that is large enough to fit the source and the dictionary.
|
||||
* The zstd format says that the entire dictionary is valid if one byte of the dictionary
|
||||
* is within the window. So the hashLog and chainLog should be large enough to reference both
|
||||
* the dictionary and the window. So we must use this adjusted dictAndWindowLog when downsizing
|
||||
* the hashLog and windowLog.
|
||||
* NOTE: srcSize must not be ZSTD_CONTENTSIZE_UNKNOWN.
|
||||
*/
|
||||
static U32 ZSTD_dictAndWindowLog(U32 windowLog, U64 srcSize, U64 dictSize)
|
||||
{
|
||||
const U64 maxWindowSize = 1ULL << ZSTD_WINDOWLOG_MAX;
|
||||
/* No dictionary ==> No change */
|
||||
if (dictSize == 0) {
|
||||
return windowLog;
|
||||
}
|
||||
assert(windowLog <= ZSTD_WINDOWLOG_MAX);
|
||||
assert(srcSize != ZSTD_CONTENTSIZE_UNKNOWN); /* Handled in ZSTD_adjustCParams_internal() */
|
||||
{
|
||||
U64 const windowSize = 1ULL << windowLog;
|
||||
U64 const dictAndWindowSize = dictSize + windowSize;
|
||||
/* If the window size is already large enough to fit both the source and the dictionary
|
||||
* then just use the window size. Otherwise adjust so that it fits the dictionary and
|
||||
* the window.
|
||||
*/
|
||||
if (windowSize >= dictSize + srcSize) {
|
||||
return windowLog; /* Window size large enough already */
|
||||
} else if (dictAndWindowSize >= maxWindowSize) {
|
||||
return ZSTD_WINDOWLOG_MAX; /* Larger than max window log */
|
||||
} else {
|
||||
return ZSTD_highbit32((U32)dictAndWindowSize - 1) + 1;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/** ZSTD_adjustCParams_internal() :
|
||||
@@ -1442,6 +1474,8 @@ ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,
|
||||
ZSTD_CParamMode_e mode,
|
||||
ZSTD_ParamSwitch_e useRowMatchFinder)
|
||||
{
|
||||
const U64 minSrcSize = 513; /* (1<<9) + 1 */
|
||||
const U64 maxWindowResize = 1ULL << (ZSTD_WINDOWLOG_MAX-1);
|
||||
assert(ZSTD_checkCParams(cPar)==0);
|
||||
|
||||
/* Cascade the selected strategy down to the next-highest one built into
|
||||
@@ -1486,13 +1520,90 @@ ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar,
|
||||
}
|
||||
#endif
|
||||
|
||||
/* The remaining adjustment logic is context-free and lives in Rust. The
|
||||
* short-cache and row-hash tag widths the leaf assumes are fixed
|
||||
* private-header constants; keep them checked here. */
|
||||
ZSTD_STATIC_ASSERT(ZSTD_SHORT_CACHE_TAG_BITS == 8);
|
||||
ZSTD_STATIC_ASSERT(ZSTD_ROW_HASH_TAG_BITS == 8);
|
||||
return ZSTD_rust_params_adjustCParams(cPar, (U64)srcSize, dictSize,
|
||||
(int)mode, (int)useRowMatchFinder);
|
||||
switch (mode) {
|
||||
case ZSTD_cpm_unknown:
|
||||
case ZSTD_cpm_noAttachDict:
|
||||
/* If we don't know the source size, don't make any
|
||||
* assumptions about it. We will already have selected
|
||||
* smaller parameters if a dictionary is in use.
|
||||
*/
|
||||
break;
|
||||
case ZSTD_cpm_createCDict:
|
||||
/* Assume a small source size when creating a dictionary
|
||||
* with an unknown source size.
|
||||
*/
|
||||
if (dictSize && srcSize == ZSTD_CONTENTSIZE_UNKNOWN)
|
||||
srcSize = minSrcSize;
|
||||
break;
|
||||
case ZSTD_cpm_attachDict:
|
||||
/* Dictionary has its own dedicated parameters which have
|
||||
* already been selected. We are selecting parameters
|
||||
* for only the source.
|
||||
*/
|
||||
dictSize = 0;
|
||||
break;
|
||||
default:
|
||||
assert(0);
|
||||
break;
|
||||
}
|
||||
|
||||
/* resize windowLog if input is small enough, to use less memory */
|
||||
if ( (srcSize <= maxWindowResize)
|
||||
&& (dictSize <= maxWindowResize) ) {
|
||||
U32 const tSize = (U32)(srcSize + dictSize);
|
||||
static U32 const hashSizeMin = 1 << ZSTD_HASHLOG_MIN;
|
||||
U32 const srcLog = (tSize < hashSizeMin) ? ZSTD_HASHLOG_MIN :
|
||||
ZSTD_highbit32(tSize-1) + 1;
|
||||
if (cPar.windowLog > srcLog) cPar.windowLog = srcLog;
|
||||
}
|
||||
if (srcSize != ZSTD_CONTENTSIZE_UNKNOWN) {
|
||||
U32 const dictAndWindowLog = ZSTD_dictAndWindowLog(cPar.windowLog, (U64)srcSize, (U64)dictSize);
|
||||
U32 const cycleLog = ZSTD_cycleLog(cPar.chainLog, cPar.strategy);
|
||||
if (cPar.hashLog > dictAndWindowLog+1) cPar.hashLog = dictAndWindowLog+1;
|
||||
if (cycleLog > dictAndWindowLog)
|
||||
cPar.chainLog -= (cycleLog - dictAndWindowLog);
|
||||
}
|
||||
|
||||
if (cPar.windowLog < ZSTD_WINDOWLOG_ABSOLUTEMIN)
|
||||
cPar.windowLog = ZSTD_WINDOWLOG_ABSOLUTEMIN; /* minimum wlog required for valid frame header */
|
||||
|
||||
/* We can't use more than 32 bits of hash in total, so that means that we require:
|
||||
* (hashLog + 8) <= 32 && (chainLog + 8) <= 32
|
||||
*/
|
||||
if (mode == ZSTD_cpm_createCDict && ZSTD_CDictIndicesAreTagged(&cPar)) {
|
||||
U32 const maxShortCacheHashLog = 32 - ZSTD_SHORT_CACHE_TAG_BITS;
|
||||
if (cPar.hashLog > maxShortCacheHashLog) {
|
||||
cPar.hashLog = maxShortCacheHashLog;
|
||||
}
|
||||
if (cPar.chainLog > maxShortCacheHashLog) {
|
||||
cPar.chainLog = maxShortCacheHashLog;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
/* At this point, we aren't 100% sure if we are using the row match finder.
|
||||
* Unless it is explicitly disabled, conservatively assume that it is enabled.
|
||||
* In this case it will only be disabled for small sources, so shrinking the
|
||||
* hash log a little bit shouldn't result in any ratio loss.
|
||||
*/
|
||||
if (useRowMatchFinder == ZSTD_ps_auto)
|
||||
useRowMatchFinder = ZSTD_ps_enable;
|
||||
|
||||
/* We can't hash more than 32-bits in total. So that means that we require:
|
||||
* (hashLog - rowLog + 8) <= 32
|
||||
*/
|
||||
if (ZSTD_rowMatchFinderUsed(cPar.strategy, useRowMatchFinder)) {
|
||||
/* Switch to 32-entry rows if searchLog is 5 (or more) */
|
||||
U32 const rowLog = BOUNDED(4, cPar.searchLog, 6);
|
||||
U32 const maxRowHashLog = 32 - ZSTD_ROW_HASH_TAG_BITS;
|
||||
U32 const maxHashLog = maxRowHashLog + rowLog;
|
||||
assert(cPar.hashLog >= rowLog);
|
||||
if (cPar.hashLog > maxHashLog) {
|
||||
cPar.hashLog = maxHashLog;
|
||||
}
|
||||
}
|
||||
|
||||
return cPar;
|
||||
}
|
||||
|
||||
ZSTD_compressionParameters
|
||||
@@ -1543,31 +1654,47 @@ ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
|
||||
const int enableDedicatedDictSearch,
|
||||
const U32 forCCtx)
|
||||
{
|
||||
ZSTD_rustMatchStateSizing sizing;
|
||||
sizing.hashLog3Max = ZSTD_HASHLOG3_MAX;
|
||||
sizing.matchTSize = sizeof(ZSTD_match_t);
|
||||
sizing.optimalTSize = sizeof(ZSTD_optimal_t);
|
||||
sizing.asanRedzoneSize = ZSTD_RUST_ASAN_REDZONE_SIZE;
|
||||
/* chain table size should be 0 for fast or row-hash strategies */
|
||||
size_t const chainSize = ZSTD_allocateChainTable(cParams->strategy, useRowMatchFinder, enableDedicatedDictSearch && !forCCtx)
|
||||
? ((size_t)1 << cParams->chainLog)
|
||||
: 0;
|
||||
size_t const hSize = ((size_t)1) << cParams->hashLog;
|
||||
U32 const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
|
||||
size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
|
||||
/* We don't use ZSTD_cwksp_alloc_size() here because the tables aren't
|
||||
* surrounded by redzones in ASAN. */
|
||||
size_t const tableSpace = chainSize * sizeof(U32)
|
||||
+ hSize * sizeof(U32)
|
||||
+ h3Size * sizeof(U32);
|
||||
size_t const optPotentialSpace =
|
||||
ZSTD_cwksp_aligned64_alloc_size((MaxML+1) * sizeof(U32))
|
||||
+ ZSTD_cwksp_aligned64_alloc_size((MaxLL+1) * sizeof(U32))
|
||||
+ ZSTD_cwksp_aligned64_alloc_size((MaxOff+1) * sizeof(U32))
|
||||
+ ZSTD_cwksp_aligned64_alloc_size((1<<Litbits) * sizeof(U32))
|
||||
+ ZSTD_cwksp_aligned64_alloc_size(ZSTD_OPT_SIZE * sizeof(ZSTD_match_t))
|
||||
+ ZSTD_cwksp_aligned64_alloc_size(ZSTD_OPT_SIZE * sizeof(ZSTD_optimal_t));
|
||||
size_t const lazyAdditionalSpace = ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)
|
||||
? ZSTD_cwksp_aligned64_alloc_size(hSize)
|
||||
: 0;
|
||||
size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))
|
||||
? optPotentialSpace
|
||||
: 0;
|
||||
size_t const slackSpace = ZSTD_cwksp_slack_space_required();
|
||||
|
||||
/* The Rust leaf hardcodes the frozen format bounds and workspace rules;
|
||||
* keep them checked against the private headers here. */
|
||||
ZSTD_STATIC_ASSERT(MaxML == 52 && MaxLL == 35 && MaxOff == 31);
|
||||
ZSTD_STATIC_ASSERT(Litbits == 8 && ZSTD_OPT_SIZE == 4099);
|
||||
ZSTD_STATIC_ASSERT(ZSTD_CWKSP_ALIGNMENT_BYTES == 64);
|
||||
/* tables are guaranteed to be sized in multiples of 64 bytes (or 16 uint32_t) */
|
||||
ZSTD_STATIC_ASSERT(ZSTD_HASHLOG_MIN >= 4 && ZSTD_WINDOWLOG_MIN >= 4 && ZSTD_CHAINLOG_MIN >= 4);
|
||||
assert(useRowMatchFinder != ZSTD_ps_auto);
|
||||
|
||||
return ZSTD_rust_params_estimateMatchStateSize(*cParams,
|
||||
(int)useRowMatchFinder,
|
||||
enableDedicatedDictSearch,
|
||||
forCCtx, &sizing);
|
||||
DEBUGLOG(4, "chainSize: %u - hSize: %u - h3Size: %u",
|
||||
(U32)chainSize, (U32)hSize, (U32)h3Size);
|
||||
return tableSpace + optSpace + slackSpace + lazyAdditionalSpace;
|
||||
}
|
||||
|
||||
/* Helper function for calculating memory requirements.
|
||||
* Gives a tighter bound than ZSTD_sequenceBound() by taking minMatch into account. */
|
||||
static size_t ZSTD_maxNbSeq(size_t blockSize, unsigned minMatch, int useSequenceProducer) {
|
||||
return ZSTD_rust_params_maxNbSeq(blockSize, minMatch, useSequenceProducer);
|
||||
U32 const divider = (minMatch==3 || useSequenceProducer) ? 3 : 4;
|
||||
return blockSize / divider;
|
||||
}
|
||||
|
||||
static size_t ZSTD_estimateCCtxSize_usingCCtxParams_internal(
|
||||
@@ -5303,15 +5430,15 @@ size_t ZSTD_estimateCDictSize_advanced(
|
||||
size_t dictSize, ZSTD_compressionParameters cParams,
|
||||
ZSTD_dictLoadMethod_e dictLoadMethod)
|
||||
{
|
||||
ZSTD_rustCDictSizing sizing;
|
||||
sizing.cdictSize = sizeof(ZSTD_CDict);
|
||||
sizing.hufWorkspaceSize = HUF_WORKSPACE_SIZE;
|
||||
sizing.hashLog3Max = ZSTD_HASHLOG3_MAX;
|
||||
sizing.matchTSize = sizeof(ZSTD_match_t);
|
||||
sizing.optimalTSize = sizeof(ZSTD_optimal_t);
|
||||
sizing.asanRedzoneSize = ZSTD_RUST_ASAN_REDZONE_SIZE;
|
||||
return ZSTD_rust_params_estimateCDictSizeFromCParams(
|
||||
dictSize, cParams, (int)dictLoadMethod, &sizing);
|
||||
DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict));
|
||||
return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
|
||||
+ ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
|
||||
/* enableDedicatedDictSearch == 1 ensures that CDict estimation will not be too small
|
||||
* in case we are using DDS with row-hash. */
|
||||
+ ZSTD_sizeof_matchState(&cParams, ZSTD_resolveRowMatchFinderMode(ZSTD_ps_auto, &cParams),
|
||||
/* enableDedicatedDictSearch */ 1, /* forCCtx */ 0)
|
||||
+ (dictLoadMethod == ZSTD_dlm_byRef ? 0
|
||||
: ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *))));
|
||||
}
|
||||
|
||||
size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)
|
||||
@@ -7449,12 +7576,11 @@ size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
|
||||
|
||||
|
||||
/*-===== Pre-defined compression levels =====-*/
|
||||
/* The compression-level tables (formerly included from clevels.h) live in
|
||||
* rust/src/zstd_compress_params.rs. */
|
||||
#include "clevels.h"
|
||||
|
||||
int ZSTD_maxCLevel(void) { return ZSTD_rust_params_maxCLevel(); }
|
||||
int ZSTD_minCLevel(void) { return ZSTD_rust_params_minCLevel(); }
|
||||
int ZSTD_defaultCLevel(void) { return ZSTD_rust_params_defaultCLevel(); }
|
||||
int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }
|
||||
int ZSTD_minCLevel(void) { return (int)-ZSTD_TARGETLENGTH_MAX; }
|
||||
int ZSTD_defaultCLevel(void) { return ZSTD_CLEVEL_DEFAULT; }
|
||||
|
||||
static ZSTD_compressionParameters ZSTD_dedicatedDictSearch_getCParams(int const compressionLevel, size_t const dictSize)
|
||||
{
|
||||
@@ -7513,6 +7639,26 @@ static void ZSTD_dedicatedDictSearch_revertCParams(
|
||||
}
|
||||
}
|
||||
|
||||
static U64 ZSTD_getCParamRowSize(U64 srcSizeHint, size_t dictSize, ZSTD_CParamMode_e mode)
|
||||
{
|
||||
switch (mode) {
|
||||
case ZSTD_cpm_unknown:
|
||||
case ZSTD_cpm_noAttachDict:
|
||||
case ZSTD_cpm_createCDict:
|
||||
break;
|
||||
case ZSTD_cpm_attachDict:
|
||||
dictSize = 0;
|
||||
break;
|
||||
default:
|
||||
assert(0);
|
||||
break;
|
||||
}
|
||||
{ int const unknown = srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN;
|
||||
size_t const addedSize = unknown && dictSize > 0 ? 500 : 0;
|
||||
return unknown && dictSize == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : srcSizeHint+dictSize+addedSize;
|
||||
}
|
||||
}
|
||||
|
||||
/*! ZSTD_getCParams_internal() :
|
||||
* @return ZSTD_compressionParameters structure for a selected compression level, srcSize and dictSize.
|
||||
* Note: srcSizeHint 0 means 0, use ZSTD_CONTENTSIZE_UNKNOWN for unknown.
|
||||
@@ -7520,12 +7666,27 @@ static void ZSTD_dedicatedDictSearch_revertCParams(
|
||||
* Note: `mode` controls how we treat the `dictSize`. See docs for `ZSTD_CParamMode_e`. */
|
||||
static ZSTD_compressionParameters ZSTD_getCParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_CParamMode_e mode)
|
||||
{
|
||||
/* Table selection is context-free and lives in Rust; the adjustment step
|
||||
* stays behind ZSTD_adjustCParams_internal() so this build's strategy
|
||||
* cascade applies. */
|
||||
ZSTD_compressionParameters const cp = ZSTD_rust_params_selectCParams(
|
||||
compressionLevel, srcSizeHint, dictSize, (int)mode);
|
||||
return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize, mode, ZSTD_ps_auto);
|
||||
U64 const rSize = ZSTD_getCParamRowSize(srcSizeHint, dictSize, mode);
|
||||
U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);
|
||||
int row;
|
||||
DEBUGLOG(5, "ZSTD_getCParams_internal (cLevel=%i)", compressionLevel);
|
||||
|
||||
/* row */
|
||||
if (compressionLevel == 0) row = ZSTD_CLEVEL_DEFAULT; /* 0 == default */
|
||||
else if (compressionLevel < 0) row = 0; /* entry 0 is baseline for fast mode */
|
||||
else if (compressionLevel > ZSTD_MAX_CLEVEL) row = ZSTD_MAX_CLEVEL;
|
||||
else row = compressionLevel;
|
||||
|
||||
{ ZSTD_compressionParameters cp = ZSTD_defaultCParameters[tableID][row];
|
||||
DEBUGLOG(5, "ZSTD_getCParams_internal selected tableID: %u row: %u strat: %u", tableID, row, (U32)cp.strategy);
|
||||
/* acceleration factor */
|
||||
if (compressionLevel < 0) {
|
||||
int const clampedCompressionLevel = MAX(ZSTD_minCLevel(), compressionLevel);
|
||||
cp.targetLength = (unsigned)(-clampedCompressionLevel);
|
||||
}
|
||||
/* refine parameters based on srcSize & dictSize */
|
||||
return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize, mode, ZSTD_ps_auto);
|
||||
}
|
||||
}
|
||||
|
||||
/*! ZSTD_getCParams() :
|
||||
@@ -7544,9 +7705,13 @@ ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long l
|
||||
static ZSTD_parameters
|
||||
ZSTD_getParams_internal(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize, ZSTD_CParamMode_e mode)
|
||||
{
|
||||
ZSTD_parameters params;
|
||||
ZSTD_compressionParameters const cParams = ZSTD_getCParams_internal(compressionLevel, srcSizeHint, dictSize, mode);
|
||||
DEBUGLOG(5, "ZSTD_getParams (cLevel=%i)", compressionLevel);
|
||||
return ZSTD_rust_params_makeParams(cParams);
|
||||
ZSTD_memset(¶ms, 0, sizeof(params));
|
||||
params.cParams = cParams;
|
||||
params.fParams.contentSizeFlag = 1;
|
||||
return params;
|
||||
}
|
||||
|
||||
/*! ZSTD_getParams() :
|
||||
|
||||
+2108
-4
@@ -17,7 +17,2111 @@
|
||||
#include "../common/compiler.h"
|
||||
#include "../common/error_private.h"
|
||||
|
||||
/* Implementation moved to Rust (rust/src/legacy/zstd_v01.rs).
|
||||
* The frozen v0.1 decoder, including its embedded FSE/Huff0 snapshot and the
|
||||
* ZSTDv01_Dctx streaming state, now lives entirely in Rust; C code only ever
|
||||
* holds an opaque ZSTDv01_Dctx pointer. */
|
||||
|
||||
/******************************************
|
||||
* Static allocation
|
||||
******************************************/
|
||||
/* You can statically allocate FSE CTable/DTable as a table of unsigned using below macro */
|
||||
#define FSE_DTABLE_SIZE_U32(maxTableLog) (1 + (1<<maxTableLog))
|
||||
|
||||
/* You can statically allocate Huff0 DTable as a table of unsigned short using below macro */
|
||||
#define HUF_DTABLE_SIZE_U16(maxTableLog) (1 + (1<<maxTableLog))
|
||||
#define HUF_CREATE_STATIC_DTABLE(DTable, maxTableLog) \
|
||||
unsigned short DTable[HUF_DTABLE_SIZE_U16(maxTableLog)] = { maxTableLog }
|
||||
|
||||
|
||||
/******************************************
|
||||
* Error Management
|
||||
******************************************/
|
||||
#define FSE_LIST_ERRORS(ITEM) \
|
||||
ITEM(FSE_OK_NoError) ITEM(FSE_ERROR_GENERIC) \
|
||||
ITEM(FSE_ERROR_tableLog_tooLarge) ITEM(FSE_ERROR_maxSymbolValue_tooLarge) ITEM(FSE_ERROR_maxSymbolValue_tooSmall) \
|
||||
ITEM(FSE_ERROR_dstSize_tooSmall) ITEM(FSE_ERROR_srcSize_wrong)\
|
||||
ITEM(FSE_ERROR_corruptionDetected) \
|
||||
ITEM(FSE_ERROR_maxCode)
|
||||
|
||||
#define FSE_GENERATE_ENUM(ENUM) ENUM,
|
||||
typedef enum { FSE_LIST_ERRORS(FSE_GENERATE_ENUM) } FSE_errorCodes; /* enum is exposed, to detect & handle specific errors; compare function result to -enum value */
|
||||
|
||||
|
||||
/******************************************
|
||||
* FSE symbol compression API
|
||||
******************************************/
|
||||
/*
|
||||
This API consists of small unitary functions, which highly benefit from being inlined.
|
||||
You will want to enable link-time-optimization to ensure these functions are properly inlined in your binary.
|
||||
Visual seems to do it automatically.
|
||||
For gcc or clang, you'll need to add -flto flag at compilation and linking stages.
|
||||
If none of these solutions is applicable, include "fse.c" directly.
|
||||
*/
|
||||
|
||||
typedef unsigned FSE_CTable; /* don't allocate that. It's just a way to be more restrictive than void* */
|
||||
typedef unsigned FSE_DTable; /* don't allocate that. It's just a way to be more restrictive than void* */
|
||||
|
||||
typedef struct
|
||||
{
|
||||
size_t bitContainer;
|
||||
int bitPos;
|
||||
char* startPtr;
|
||||
char* ptr;
|
||||
char* endPtr;
|
||||
} FSE_CStream_t;
|
||||
|
||||
typedef struct
|
||||
{
|
||||
ptrdiff_t value;
|
||||
const void* stateTable;
|
||||
const void* symbolTT;
|
||||
unsigned stateLog;
|
||||
} FSE_CState_t;
|
||||
|
||||
typedef struct
|
||||
{
|
||||
size_t bitContainer;
|
||||
unsigned bitsConsumed;
|
||||
const char* ptr;
|
||||
const char* start;
|
||||
} FSE_DStream_t;
|
||||
|
||||
typedef struct
|
||||
{
|
||||
size_t state;
|
||||
const void* table; /* precise table may vary, depending on U16 */
|
||||
} FSE_DState_t;
|
||||
|
||||
typedef enum { FSE_DStream_unfinished = 0,
|
||||
FSE_DStream_endOfBuffer = 1,
|
||||
FSE_DStream_completed = 2,
|
||||
FSE_DStream_tooFar = 3 } FSE_DStream_status; /* result of FSE_reloadDStream() */
|
||||
/* 1,2,4,8 would be better for bitmap combinations, but slows down performance a bit ... ?! */
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Tuning parameters
|
||||
****************************************************************/
|
||||
/* MEMORY_USAGE :
|
||||
* Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
|
||||
* Increasing memory usage improves compression ratio
|
||||
* Reduced memory usage can improve speed, due to cache effect
|
||||
* Recommended max value is 14, for 16KB, which nicely fits into Intel x86 L1 cache */
|
||||
#define FSE_MAX_MEMORY_USAGE 14
|
||||
#define FSE_DEFAULT_MEMORY_USAGE 13
|
||||
|
||||
/* FSE_MAX_SYMBOL_VALUE :
|
||||
* Maximum symbol value authorized.
|
||||
* Required for proper stack allocation */
|
||||
#define FSE_MAX_SYMBOL_VALUE 255
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* template functions type & suffix
|
||||
****************************************************************/
|
||||
#define FSE_FUNCTION_TYPE BYTE
|
||||
#define FSE_FUNCTION_EXTENSION
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Byte symbol type
|
||||
****************************************************************/
|
||||
typedef struct
|
||||
{
|
||||
unsigned short newState;
|
||||
unsigned char symbol;
|
||||
unsigned char nbBits;
|
||||
} FSE_decode_t; /* size == U32 */
|
||||
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Compiler specifics
|
||||
****************************************************************/
|
||||
#ifdef _MSC_VER /* Visual Studio */
|
||||
# define FORCE_INLINE static __forceinline
|
||||
# include <intrin.h> /* For Visual 2005 */
|
||||
# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
|
||||
# pragma warning(disable : 4214) /* disable: C4214: non-int bitfields */
|
||||
#else
|
||||
# define GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
|
||||
# if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */
|
||||
# ifdef __GNUC__
|
||||
# define FORCE_INLINE static inline __attribute__((always_inline))
|
||||
# else
|
||||
# define FORCE_INLINE static inline
|
||||
# endif
|
||||
# else
|
||||
# define FORCE_INLINE static
|
||||
# endif /* __STDC_VERSION__ */
|
||||
#endif
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Includes
|
||||
****************************************************************/
|
||||
#include <stdlib.h> /* malloc, free, qsort */
|
||||
#include <string.h> /* memcpy, memset */
|
||||
#include <stdio.h> /* printf (debug) */
|
||||
|
||||
|
||||
#ifndef MEM_ACCESS_MODULE
|
||||
#define MEM_ACCESS_MODULE
|
||||
/****************************************************************
|
||||
* Basic Types
|
||||
*****************************************************************/
|
||||
#if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */
|
||||
# include <stdint.h>
|
||||
typedef uint8_t BYTE;
|
||||
typedef uint16_t U16;
|
||||
typedef int16_t S16;
|
||||
typedef uint32_t U32;
|
||||
typedef int32_t S32;
|
||||
typedef uint64_t U64;
|
||||
typedef int64_t S64;
|
||||
#else
|
||||
typedef unsigned char BYTE;
|
||||
typedef unsigned short U16;
|
||||
typedef signed short S16;
|
||||
typedef unsigned int U32;
|
||||
typedef signed int S32;
|
||||
typedef unsigned long long U64;
|
||||
typedef signed long long S64;
|
||||
#endif
|
||||
|
||||
#endif /* MEM_ACCESS_MODULE */
|
||||
|
||||
/****************************************************************
|
||||
* Memory I/O
|
||||
*****************************************************************/
|
||||
|
||||
static unsigned FSE_32bits(void)
|
||||
{
|
||||
return sizeof(void*)==4;
|
||||
}
|
||||
|
||||
static unsigned FSE_isLittleEndian(void)
|
||||
{
|
||||
const union { U32 i; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */
|
||||
return one.c[0];
|
||||
}
|
||||
|
||||
static U16 FSE_read16(const void* memPtr)
|
||||
{
|
||||
U16 val; memcpy(&val, memPtr, sizeof(val)); return val;
|
||||
}
|
||||
|
||||
static U32 FSE_read32(const void* memPtr)
|
||||
{
|
||||
U32 val; memcpy(&val, memPtr, sizeof(val)); return val;
|
||||
}
|
||||
|
||||
static U64 FSE_read64(const void* memPtr)
|
||||
{
|
||||
U64 val; memcpy(&val, memPtr, sizeof(val)); return val;
|
||||
}
|
||||
|
||||
static U16 FSE_readLE16(const void* memPtr)
|
||||
{
|
||||
if (FSE_isLittleEndian())
|
||||
return FSE_read16(memPtr);
|
||||
else
|
||||
{
|
||||
const BYTE* p = (const BYTE*)memPtr;
|
||||
return (U16)(p[0] + (p[1]<<8));
|
||||
}
|
||||
}
|
||||
|
||||
static U32 FSE_readLE32(const void* memPtr)
|
||||
{
|
||||
if (FSE_isLittleEndian())
|
||||
return FSE_read32(memPtr);
|
||||
else
|
||||
{
|
||||
const BYTE* p = (const BYTE*)memPtr;
|
||||
return (U32)((U32)p[0] + ((U32)p[1]<<8) + ((U32)p[2]<<16) + ((U32)p[3]<<24));
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
static U64 FSE_readLE64(const void* memPtr)
|
||||
{
|
||||
if (FSE_isLittleEndian())
|
||||
return FSE_read64(memPtr);
|
||||
else
|
||||
{
|
||||
const BYTE* p = (const BYTE*)memPtr;
|
||||
return (U64)((U64)p[0] + ((U64)p[1]<<8) + ((U64)p[2]<<16) + ((U64)p[3]<<24)
|
||||
+ ((U64)p[4]<<32) + ((U64)p[5]<<40) + ((U64)p[6]<<48) + ((U64)p[7]<<56));
|
||||
}
|
||||
}
|
||||
|
||||
static size_t FSE_readLEST(const void* memPtr)
|
||||
{
|
||||
if (FSE_32bits())
|
||||
return (size_t)FSE_readLE32(memPtr);
|
||||
else
|
||||
return (size_t)FSE_readLE64(memPtr);
|
||||
}
|
||||
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Constants
|
||||
*****************************************************************/
|
||||
#define FSE_MAX_TABLELOG (FSE_MAX_MEMORY_USAGE-2)
|
||||
#define FSE_MAX_TABLESIZE (1U<<FSE_MAX_TABLELOG)
|
||||
#define FSE_MAXTABLESIZE_MASK (FSE_MAX_TABLESIZE-1)
|
||||
#define FSE_DEFAULT_TABLELOG (FSE_DEFAULT_MEMORY_USAGE-2)
|
||||
#define FSE_MIN_TABLELOG 5
|
||||
|
||||
#define FSE_TABLELOG_ABSOLUTE_MAX 15
|
||||
#if FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX
|
||||
#error "FSE_MAX_TABLELOG > FSE_TABLELOG_ABSOLUTE_MAX is not supported"
|
||||
#endif
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Error Management
|
||||
****************************************************************/
|
||||
#define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; } /* use only *after* variable declarations */
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Complex types
|
||||
****************************************************************/
|
||||
typedef struct
|
||||
{
|
||||
int deltaFindState;
|
||||
U32 deltaNbBits;
|
||||
} FSE_symbolCompressionTransform; /* total 8 bytes */
|
||||
|
||||
typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
|
||||
|
||||
/****************************************************************
|
||||
* Internal functions
|
||||
****************************************************************/
|
||||
FORCE_INLINE unsigned FSE_highbit32 (U32 val)
|
||||
{
|
||||
# if defined(_MSC_VER) /* Visual */
|
||||
unsigned long r;
|
||||
return _BitScanReverse(&r, val) ? (unsigned)r : 0;
|
||||
# elif defined(__GNUC__) && (GCC_VERSION >= 304) /* GCC Intrinsic */
|
||||
return __builtin_clz (val) ^ 31;
|
||||
# else /* Software version */
|
||||
static const unsigned DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
|
||||
U32 v = val;
|
||||
unsigned r;
|
||||
v |= v >> 1;
|
||||
v |= v >> 2;
|
||||
v |= v >> 4;
|
||||
v |= v >> 8;
|
||||
v |= v >> 16;
|
||||
r = DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27];
|
||||
return r;
|
||||
# endif
|
||||
}
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Templates
|
||||
****************************************************************/
|
||||
/*
|
||||
designed to be included
|
||||
for type-specific functions (template emulation in C)
|
||||
Objective is to write these functions only once, for improved maintenance
|
||||
*/
|
||||
|
||||
/* safety checks */
|
||||
#ifndef FSE_FUNCTION_EXTENSION
|
||||
# error "FSE_FUNCTION_EXTENSION must be defined"
|
||||
#endif
|
||||
#ifndef FSE_FUNCTION_TYPE
|
||||
# error "FSE_FUNCTION_TYPE must be defined"
|
||||
#endif
|
||||
|
||||
/* Function names */
|
||||
#define FSE_CAT(X,Y) X##Y
|
||||
#define FSE_FUNCTION_NAME(X,Y) FSE_CAT(X,Y)
|
||||
#define FSE_TYPE_NAME(X,Y) FSE_CAT(X,Y)
|
||||
|
||||
|
||||
|
||||
static U32 FSE_tableStep(U32 tableSize) { return (tableSize>>1) + (tableSize>>3) + 3; }
|
||||
|
||||
#define FSE_DECODE_TYPE FSE_decode_t
|
||||
|
||||
|
||||
typedef struct {
|
||||
U16 tableLog;
|
||||
U16 fastMode;
|
||||
} FSE_DTableHeader; /* sizeof U32 */
|
||||
|
||||
static size_t FSE_buildDTable
|
||||
(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
|
||||
{
|
||||
void* ptr = dt;
|
||||
FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
|
||||
FSE_DECODE_TYPE* const tableDecode = (FSE_DECODE_TYPE*)(ptr) + 1; /* because dt is unsigned, 32-bits aligned on 32-bits */
|
||||
const U32 tableSize = 1 << tableLog;
|
||||
const U32 tableMask = tableSize-1;
|
||||
const U32 step = FSE_tableStep(tableSize);
|
||||
U16 symbolNext[FSE_MAX_SYMBOL_VALUE+1];
|
||||
U32 position = 0;
|
||||
U32 highThreshold = tableSize-1;
|
||||
const S16 largeLimit= (S16)(1 << (tableLog-1));
|
||||
U32 noLarge = 1;
|
||||
U32 s;
|
||||
|
||||
/* Sanity Checks */
|
||||
if (maxSymbolValue > FSE_MAX_SYMBOL_VALUE) return (size_t)-FSE_ERROR_maxSymbolValue_tooLarge;
|
||||
if (tableLog > FSE_MAX_TABLELOG) return (size_t)-FSE_ERROR_tableLog_tooLarge;
|
||||
|
||||
/* Init, lay down lowprob symbols */
|
||||
DTableH[0].tableLog = (U16)tableLog;
|
||||
for (s=0; s<=maxSymbolValue; s++)
|
||||
{
|
||||
if (normalizedCounter[s]==-1)
|
||||
{
|
||||
tableDecode[highThreshold--].symbol = (FSE_FUNCTION_TYPE)s;
|
||||
symbolNext[s] = 1;
|
||||
}
|
||||
else
|
||||
{
|
||||
if (normalizedCounter[s] >= largeLimit) noLarge=0;
|
||||
symbolNext[s] = normalizedCounter[s];
|
||||
}
|
||||
}
|
||||
|
||||
/* Spread symbols */
|
||||
for (s=0; s<=maxSymbolValue; s++)
|
||||
{
|
||||
int i;
|
||||
for (i=0; i<normalizedCounter[s]; i++)
|
||||
{
|
||||
tableDecode[position].symbol = (FSE_FUNCTION_TYPE)s;
|
||||
position = (position + step) & tableMask;
|
||||
while (position > highThreshold) position = (position + step) & tableMask; /* lowprob area */
|
||||
}
|
||||
}
|
||||
|
||||
if (position!=0) return (size_t)-FSE_ERROR_GENERIC; /* position must reach all cells once, otherwise normalizedCounter is incorrect */
|
||||
|
||||
/* Build Decoding table */
|
||||
{
|
||||
U32 i;
|
||||
for (i=0; i<tableSize; i++)
|
||||
{
|
||||
FSE_FUNCTION_TYPE symbol = (FSE_FUNCTION_TYPE)(tableDecode[i].symbol);
|
||||
U16 nextState = symbolNext[symbol]++;
|
||||
tableDecode[i].nbBits = (BYTE) (tableLog - FSE_highbit32 ((U32)nextState) );
|
||||
tableDecode[i].newState = (U16) ( (nextState << tableDecode[i].nbBits) - tableSize);
|
||||
}
|
||||
}
|
||||
|
||||
DTableH->fastMode = (U16)noLarge;
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
/******************************************
|
||||
* FSE byte symbol
|
||||
******************************************/
|
||||
#ifndef FSE_COMMONDEFS_ONLY
|
||||
|
||||
static unsigned FSE_isError(size_t code) { return (code > (size_t)(-FSE_ERROR_maxCode)); }
|
||||
|
||||
static short FSE_abs(short a)
|
||||
{
|
||||
return a<0? -a : a;
|
||||
}
|
||||
|
||||
|
||||
/****************************************************************
|
||||
* Header bitstream management
|
||||
****************************************************************/
|
||||
static size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSVPtr, unsigned* tableLogPtr,
|
||||
const void* headerBuffer, size_t hbSize)
|
||||
{
|
||||
const BYTE* const istart = (const BYTE*) headerBuffer;
|
||||
const BYTE* const iend = istart + hbSize;
|
||||
const BYTE* ip = istart;
|
||||
int nbBits;
|
||||
int remaining;
|
||||
int threshold;
|
||||
U32 bitStream;
|
||||
int bitCount;
|
||||
unsigned charnum = 0;
|
||||
int previous0 = 0;
|
||||
|
||||
if (hbSize < 4) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
bitStream = FSE_readLE32(ip);
|
||||
nbBits = (bitStream & 0xF) + FSE_MIN_TABLELOG; /* extract tableLog */
|
||||
if (nbBits > FSE_TABLELOG_ABSOLUTE_MAX) return (size_t)-FSE_ERROR_tableLog_tooLarge;
|
||||
bitStream >>= 4;
|
||||
bitCount = 4;
|
||||
*tableLogPtr = nbBits;
|
||||
remaining = (1<<nbBits)+1;
|
||||
threshold = 1<<nbBits;
|
||||
nbBits++;
|
||||
|
||||
while ((remaining>1) && (charnum<=*maxSVPtr))
|
||||
{
|
||||
if (previous0)
|
||||
{
|
||||
unsigned n0 = charnum;
|
||||
while ((bitStream & 0xFFFF) == 0xFFFF)
|
||||
{
|
||||
n0+=24;
|
||||
if (ip < iend-5)
|
||||
{
|
||||
ip+=2;
|
||||
bitStream = FSE_readLE32(ip) >> bitCount;
|
||||
}
|
||||
else
|
||||
{
|
||||
bitStream >>= 16;
|
||||
bitCount+=16;
|
||||
}
|
||||
}
|
||||
while ((bitStream & 3) == 3)
|
||||
{
|
||||
n0+=3;
|
||||
bitStream>>=2;
|
||||
bitCount+=2;
|
||||
}
|
||||
n0 += bitStream & 3;
|
||||
bitCount += 2;
|
||||
if (n0 > *maxSVPtr) return (size_t)-FSE_ERROR_maxSymbolValue_tooSmall;
|
||||
while (charnum < n0) normalizedCounter[charnum++] = 0;
|
||||
if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4))
|
||||
{
|
||||
ip += bitCount>>3;
|
||||
bitCount &= 7;
|
||||
bitStream = FSE_readLE32(ip) >> bitCount;
|
||||
}
|
||||
else
|
||||
bitStream >>= 2;
|
||||
}
|
||||
{
|
||||
const short max = (short)((2*threshold-1)-remaining);
|
||||
short count;
|
||||
|
||||
if ((bitStream & (threshold-1)) < (U32)max)
|
||||
{
|
||||
count = (short)(bitStream & (threshold-1));
|
||||
bitCount += nbBits-1;
|
||||
}
|
||||
else
|
||||
{
|
||||
count = (short)(bitStream & (2*threshold-1));
|
||||
if (count >= threshold) count -= max;
|
||||
bitCount += nbBits;
|
||||
}
|
||||
|
||||
count--; /* extra accuracy */
|
||||
remaining -= FSE_abs(count);
|
||||
normalizedCounter[charnum++] = count;
|
||||
previous0 = !count;
|
||||
while (remaining < threshold)
|
||||
{
|
||||
nbBits--;
|
||||
threshold >>= 1;
|
||||
}
|
||||
|
||||
{
|
||||
if ((ip <= iend-7) || (ip + (bitCount>>3) <= iend-4))
|
||||
{
|
||||
ip += bitCount>>3;
|
||||
bitCount &= 7;
|
||||
}
|
||||
else
|
||||
{
|
||||
bitCount -= (int)(8 * (iend - 4 - ip));
|
||||
ip = iend - 4;
|
||||
}
|
||||
bitStream = FSE_readLE32(ip) >> (bitCount & 31);
|
||||
}
|
||||
}
|
||||
}
|
||||
if (remaining != 1) return (size_t)-FSE_ERROR_GENERIC;
|
||||
*maxSVPtr = charnum-1;
|
||||
|
||||
ip += (bitCount+7)>>3;
|
||||
if ((size_t)(ip-istart) > hbSize) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
return ip-istart;
|
||||
}
|
||||
|
||||
|
||||
/*********************************************************
|
||||
* Decompression (Byte symbols)
|
||||
*********************************************************/
|
||||
static size_t FSE_buildDTable_rle (FSE_DTable* dt, BYTE symbolValue)
|
||||
{
|
||||
void* ptr = dt;
|
||||
FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
|
||||
FSE_decode_t* const cell = (FSE_decode_t*)(ptr) + 1; /* because dt is unsigned */
|
||||
|
||||
DTableH->tableLog = 0;
|
||||
DTableH->fastMode = 0;
|
||||
|
||||
cell->newState = 0;
|
||||
cell->symbol = symbolValue;
|
||||
cell->nbBits = 0;
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
static size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits)
|
||||
{
|
||||
void* ptr = dt;
|
||||
FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
|
||||
FSE_decode_t* const dinfo = (FSE_decode_t*)(ptr) + 1; /* because dt is unsigned */
|
||||
const unsigned tableSize = 1 << nbBits;
|
||||
const unsigned tableMask = tableSize - 1;
|
||||
const unsigned maxSymbolValue = tableMask;
|
||||
unsigned s;
|
||||
|
||||
/* Sanity checks */
|
||||
if (nbBits < 1) return (size_t)-FSE_ERROR_GENERIC; /* min size */
|
||||
|
||||
/* Build Decoding Table */
|
||||
DTableH->tableLog = (U16)nbBits;
|
||||
DTableH->fastMode = 1;
|
||||
for (s=0; s<=maxSymbolValue; s++)
|
||||
{
|
||||
dinfo[s].newState = 0;
|
||||
dinfo[s].symbol = (BYTE)s;
|
||||
dinfo[s].nbBits = (BYTE)nbBits;
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
|
||||
/* FSE_initDStream
|
||||
* Initialize a FSE_DStream_t.
|
||||
* srcBuffer must point at the beginning of an FSE block.
|
||||
* The function result is the size of the FSE_block (== srcSize).
|
||||
* If srcSize is too small, the function will return an errorCode;
|
||||
*/
|
||||
static size_t FSE_initDStream(FSE_DStream_t* bitD, const void* srcBuffer, size_t srcSize)
|
||||
{
|
||||
if (srcSize < 1) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
|
||||
if (srcSize >= sizeof(size_t))
|
||||
{
|
||||
U32 contain32;
|
||||
bitD->start = (const char*)srcBuffer;
|
||||
bitD->ptr = (const char*)srcBuffer + srcSize - sizeof(size_t);
|
||||
bitD->bitContainer = FSE_readLEST(bitD->ptr);
|
||||
contain32 = ((const BYTE*)srcBuffer)[srcSize-1];
|
||||
if (contain32 == 0) return (size_t)-FSE_ERROR_GENERIC; /* stop bit not present */
|
||||
bitD->bitsConsumed = 8 - FSE_highbit32(contain32);
|
||||
}
|
||||
else
|
||||
{
|
||||
U32 contain32;
|
||||
bitD->start = (const char*)srcBuffer;
|
||||
bitD->ptr = bitD->start;
|
||||
bitD->bitContainer = *(const BYTE*)(bitD->start);
|
||||
switch(srcSize)
|
||||
{
|
||||
case 7: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[6]) << (sizeof(size_t)*8 - 16);
|
||||
/* fallthrough */
|
||||
case 6: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[5]) << (sizeof(size_t)*8 - 24);
|
||||
/* fallthrough */
|
||||
case 5: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[4]) << (sizeof(size_t)*8 - 32);
|
||||
/* fallthrough */
|
||||
case 4: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[3]) << 24;
|
||||
/* fallthrough */
|
||||
case 3: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[2]) << 16;
|
||||
/* fallthrough */
|
||||
case 2: bitD->bitContainer += (size_t)(((const BYTE*)(bitD->start))[1]) << 8;
|
||||
/* fallthrough */
|
||||
default:;
|
||||
}
|
||||
contain32 = ((const BYTE*)srcBuffer)[srcSize-1];
|
||||
if (contain32 == 0) return (size_t)-FSE_ERROR_GENERIC; /* stop bit not present */
|
||||
bitD->bitsConsumed = 8 - FSE_highbit32(contain32);
|
||||
bitD->bitsConsumed += (U32)(sizeof(size_t) - srcSize)*8;
|
||||
}
|
||||
|
||||
return srcSize;
|
||||
}
|
||||
|
||||
|
||||
/*!FSE_lookBits
|
||||
* Provides next n bits from the bitContainer.
|
||||
* bitContainer is not modified (bits are still present for next read/look)
|
||||
* On 32-bits, maxNbBits==25
|
||||
* On 64-bits, maxNbBits==57
|
||||
* return : value extracted.
|
||||
*/
|
||||
static size_t FSE_lookBits(FSE_DStream_t* bitD, U32 nbBits)
|
||||
{
|
||||
const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
|
||||
return ((bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> 1) >> ((bitMask-nbBits) & bitMask);
|
||||
}
|
||||
|
||||
static size_t FSE_lookBitsFast(FSE_DStream_t* bitD, U32 nbBits) /* only if nbBits >= 1 !! */
|
||||
{
|
||||
const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
|
||||
return (bitD->bitContainer << (bitD->bitsConsumed & bitMask)) >> (((bitMask+1)-nbBits) & bitMask);
|
||||
}
|
||||
|
||||
static void FSE_skipBits(FSE_DStream_t* bitD, U32 nbBits)
|
||||
{
|
||||
bitD->bitsConsumed += nbBits;
|
||||
}
|
||||
|
||||
|
||||
/*!FSE_readBits
|
||||
* Read next n bits from the bitContainer.
|
||||
* On 32-bits, don't read more than maxNbBits==25
|
||||
* On 64-bits, don't read more than maxNbBits==57
|
||||
* Use the fast variant *only* if n >= 1.
|
||||
* return : value extracted.
|
||||
*/
|
||||
static size_t FSE_readBits(FSE_DStream_t* bitD, U32 nbBits)
|
||||
{
|
||||
size_t value = FSE_lookBits(bitD, nbBits);
|
||||
FSE_skipBits(bitD, nbBits);
|
||||
return value;
|
||||
}
|
||||
|
||||
static size_t FSE_readBitsFast(FSE_DStream_t* bitD, U32 nbBits) /* only if nbBits >= 1 !! */
|
||||
{
|
||||
size_t value = FSE_lookBitsFast(bitD, nbBits);
|
||||
FSE_skipBits(bitD, nbBits);
|
||||
return value;
|
||||
}
|
||||
|
||||
static unsigned FSE_reloadDStream(FSE_DStream_t* bitD)
|
||||
{
|
||||
if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8)) /* should never happen */
|
||||
return FSE_DStream_tooFar;
|
||||
|
||||
if (bitD->ptr >= bitD->start + sizeof(bitD->bitContainer))
|
||||
{
|
||||
bitD->ptr -= bitD->bitsConsumed >> 3;
|
||||
bitD->bitsConsumed &= 7;
|
||||
bitD->bitContainer = FSE_readLEST(bitD->ptr);
|
||||
return FSE_DStream_unfinished;
|
||||
}
|
||||
if (bitD->ptr == bitD->start)
|
||||
{
|
||||
if (bitD->bitsConsumed < sizeof(bitD->bitContainer)*8) return FSE_DStream_endOfBuffer;
|
||||
return FSE_DStream_completed;
|
||||
}
|
||||
{
|
||||
U32 nbBytes = bitD->bitsConsumed >> 3;
|
||||
U32 result = FSE_DStream_unfinished;
|
||||
if (bitD->ptr - nbBytes < bitD->start)
|
||||
{
|
||||
nbBytes = (U32)(bitD->ptr - bitD->start); /* ptr > start */
|
||||
result = FSE_DStream_endOfBuffer;
|
||||
}
|
||||
bitD->ptr -= nbBytes;
|
||||
bitD->bitsConsumed -= nbBytes*8;
|
||||
bitD->bitContainer = FSE_readLEST(bitD->ptr); /* reminder : srcSize > sizeof(bitD) */
|
||||
return result;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
static void FSE_initDState(FSE_DState_t* DStatePtr, FSE_DStream_t* bitD, const FSE_DTable* dt)
|
||||
{
|
||||
const void* ptr = dt;
|
||||
const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)ptr;
|
||||
DStatePtr->state = FSE_readBits(bitD, DTableH->tableLog);
|
||||
FSE_reloadDStream(bitD);
|
||||
DStatePtr->table = dt + 1;
|
||||
}
|
||||
|
||||
static BYTE FSE_decodeSymbol(FSE_DState_t* DStatePtr, FSE_DStream_t* bitD)
|
||||
{
|
||||
const FSE_decode_t DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
|
||||
const U32 nbBits = DInfo.nbBits;
|
||||
BYTE symbol = DInfo.symbol;
|
||||
size_t lowBits = FSE_readBits(bitD, nbBits);
|
||||
|
||||
DStatePtr->state = DInfo.newState + lowBits;
|
||||
return symbol;
|
||||
}
|
||||
|
||||
static BYTE FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, FSE_DStream_t* bitD)
|
||||
{
|
||||
const FSE_decode_t DInfo = ((const FSE_decode_t*)(DStatePtr->table))[DStatePtr->state];
|
||||
const U32 nbBits = DInfo.nbBits;
|
||||
BYTE symbol = DInfo.symbol;
|
||||
size_t lowBits = FSE_readBitsFast(bitD, nbBits);
|
||||
|
||||
DStatePtr->state = DInfo.newState + lowBits;
|
||||
return symbol;
|
||||
}
|
||||
|
||||
/* FSE_endOfDStream
|
||||
Tells if bitD has reached end of bitStream or not */
|
||||
|
||||
static unsigned FSE_endOfDStream(const FSE_DStream_t* bitD)
|
||||
{
|
||||
return ((bitD->ptr == bitD->start) && (bitD->bitsConsumed == sizeof(bitD->bitContainer)*8));
|
||||
}
|
||||
|
||||
static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
|
||||
{
|
||||
return DStatePtr->state == 0;
|
||||
}
|
||||
|
||||
|
||||
FORCE_INLINE size_t FSE_decompress_usingDTable_generic(
|
||||
void* dst, size_t maxDstSize,
|
||||
const void* cSrc, size_t cSrcSize,
|
||||
const FSE_DTable* dt, const unsigned fast)
|
||||
{
|
||||
BYTE* const ostart = (BYTE*) dst;
|
||||
BYTE* op = ostart;
|
||||
BYTE* const omax = op + maxDstSize;
|
||||
BYTE* const olimit = omax-3;
|
||||
|
||||
FSE_DStream_t bitD;
|
||||
FSE_DState_t state1;
|
||||
FSE_DState_t state2;
|
||||
size_t errorCode;
|
||||
|
||||
/* Init */
|
||||
errorCode = FSE_initDStream(&bitD, cSrc, cSrcSize); /* replaced last arg by maxCompressed Size */
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
|
||||
FSE_initDState(&state1, &bitD, dt);
|
||||
FSE_initDState(&state2, &bitD, dt);
|
||||
|
||||
#define FSE_GETSYMBOL(statePtr) fast ? FSE_decodeSymbolFast(statePtr, &bitD) : FSE_decodeSymbol(statePtr, &bitD)
|
||||
|
||||
/* 4 symbols per loop */
|
||||
for ( ; (FSE_reloadDStream(&bitD)==FSE_DStream_unfinished) && (op<olimit) ; op+=4)
|
||||
{
|
||||
op[0] = FSE_GETSYMBOL(&state1);
|
||||
|
||||
if (FSE_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8) /* This test must be static */
|
||||
FSE_reloadDStream(&bitD);
|
||||
|
||||
op[1] = FSE_GETSYMBOL(&state2);
|
||||
|
||||
if (FSE_MAX_TABLELOG*4+7 > sizeof(bitD.bitContainer)*8) /* This test must be static */
|
||||
{ if (FSE_reloadDStream(&bitD) > FSE_DStream_unfinished) { op+=2; break; } }
|
||||
|
||||
op[2] = FSE_GETSYMBOL(&state1);
|
||||
|
||||
if (FSE_MAX_TABLELOG*2+7 > sizeof(bitD.bitContainer)*8) /* This test must be static */
|
||||
FSE_reloadDStream(&bitD);
|
||||
|
||||
op[3] = FSE_GETSYMBOL(&state2);
|
||||
}
|
||||
|
||||
/* tail */
|
||||
/* note : FSE_reloadDStream(&bitD) >= FSE_DStream_partiallyFilled; Ends at exactly FSE_DStream_completed */
|
||||
while (1)
|
||||
{
|
||||
if ( (FSE_reloadDStream(&bitD)>FSE_DStream_completed) || (op==omax) || (FSE_endOfDStream(&bitD) && (fast || FSE_endOfDState(&state1))) )
|
||||
break;
|
||||
|
||||
*op++ = FSE_GETSYMBOL(&state1);
|
||||
|
||||
if ( (FSE_reloadDStream(&bitD)>FSE_DStream_completed) || (op==omax) || (FSE_endOfDStream(&bitD) && (fast || FSE_endOfDState(&state2))) )
|
||||
break;
|
||||
|
||||
*op++ = FSE_GETSYMBOL(&state2);
|
||||
}
|
||||
|
||||
/* end ? */
|
||||
if (FSE_endOfDStream(&bitD) && FSE_endOfDState(&state1) && FSE_endOfDState(&state2))
|
||||
return op-ostart;
|
||||
|
||||
if (op==omax) return (size_t)-FSE_ERROR_dstSize_tooSmall; /* dst buffer is full, but cSrc unfinished */
|
||||
|
||||
return (size_t)-FSE_ERROR_corruptionDetected;
|
||||
}
|
||||
|
||||
|
||||
static size_t FSE_decompress_usingDTable(void* dst, size_t originalSize,
|
||||
const void* cSrc, size_t cSrcSize,
|
||||
const FSE_DTable* dt)
|
||||
{
|
||||
FSE_DTableHeader DTableH;
|
||||
memcpy(&DTableH, dt, sizeof(DTableH)); /* memcpy() into local variable, to avoid strict aliasing warning */
|
||||
|
||||
/* select fast mode (static) */
|
||||
if (DTableH.fastMode) return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 1);
|
||||
return FSE_decompress_usingDTable_generic(dst, originalSize, cSrc, cSrcSize, dt, 0);
|
||||
}
|
||||
|
||||
|
||||
static size_t FSE_decompress(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize)
|
||||
{
|
||||
const BYTE* const istart = (const BYTE*)cSrc;
|
||||
const BYTE* ip = istart;
|
||||
short counting[FSE_MAX_SYMBOL_VALUE+1];
|
||||
DTable_max_t dt; /* Static analyzer seems unable to understand this table will be properly initialized later */
|
||||
unsigned tableLog;
|
||||
unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
|
||||
size_t errorCode;
|
||||
|
||||
if (cSrcSize<2) return (size_t)-FSE_ERROR_srcSize_wrong; /* too small input size */
|
||||
|
||||
/* normal FSE decoding mode */
|
||||
errorCode = FSE_readNCount (counting, &maxSymbolValue, &tableLog, istart, cSrcSize);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
if (errorCode >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong; /* too small input size */
|
||||
ip += errorCode;
|
||||
cSrcSize -= errorCode;
|
||||
|
||||
errorCode = FSE_buildDTable (dt, counting, maxSymbolValue, tableLog);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
|
||||
/* always return, even if it is an error code */
|
||||
return FSE_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, dt);
|
||||
}
|
||||
|
||||
|
||||
|
||||
/* *******************************************************
|
||||
* Huff0 : Huffman block compression
|
||||
*********************************************************/
|
||||
#define HUF_MAX_SYMBOL_VALUE 255
|
||||
#define HUF_DEFAULT_TABLELOG 12 /* used by default, when not specified */
|
||||
#define HUF_MAX_TABLELOG 12 /* max possible tableLog; for allocation purpose; can be modified */
|
||||
#define HUF_ABSOLUTEMAX_TABLELOG 16 /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
|
||||
#if (HUF_MAX_TABLELOG > HUF_ABSOLUTEMAX_TABLELOG)
|
||||
# error "HUF_MAX_TABLELOG is too large !"
|
||||
#endif
|
||||
|
||||
typedef struct HUF_CElt_s {
|
||||
U16 val;
|
||||
BYTE nbBits;
|
||||
} HUF_CElt ;
|
||||
|
||||
typedef struct nodeElt_s {
|
||||
U32 count;
|
||||
U16 parent;
|
||||
BYTE byte;
|
||||
BYTE nbBits;
|
||||
} nodeElt;
|
||||
|
||||
|
||||
/* *******************************************************
|
||||
* Huff0 : Huffman block decompression
|
||||
*********************************************************/
|
||||
typedef struct {
|
||||
BYTE byte;
|
||||
BYTE nbBits;
|
||||
} HUF_DElt;
|
||||
|
||||
static size_t HUF_readDTable (U16* DTable, const void* src, size_t srcSize)
|
||||
{
|
||||
BYTE huffWeight[HUF_MAX_SYMBOL_VALUE + 1];
|
||||
U32 rankVal[HUF_ABSOLUTEMAX_TABLELOG + 1]; /* large enough for values from 0 to 16 */
|
||||
U32 weightTotal;
|
||||
U32 maxBits;
|
||||
const BYTE* ip = (const BYTE*) src;
|
||||
size_t iSize;
|
||||
size_t oSize;
|
||||
U32 n;
|
||||
U32 nextRankStart;
|
||||
void* ptr = DTable+1;
|
||||
HUF_DElt* const dt = (HUF_DElt*)ptr;
|
||||
|
||||
if (!srcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
iSize = ip[0];
|
||||
|
||||
FSE_STATIC_ASSERT(sizeof(HUF_DElt) == sizeof(U16)); /* if compilation fails here, assertion is false */
|
||||
//memset(huffWeight, 0, sizeof(huffWeight)); /* should not be necessary, but some analyzer complain ... */
|
||||
if (iSize >= 128) /* special header */
|
||||
{
|
||||
if (iSize >= (242)) /* RLE */
|
||||
{
|
||||
static int l[14] = { 1, 2, 3, 4, 7, 8, 15, 16, 31, 32, 63, 64, 127, 128 };
|
||||
oSize = l[iSize-242];
|
||||
memset(huffWeight, 1, sizeof(huffWeight));
|
||||
iSize = 0;
|
||||
}
|
||||
else /* Incompressible */
|
||||
{
|
||||
oSize = iSize - 127;
|
||||
iSize = ((oSize+1)/2);
|
||||
if (iSize+1 > srcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
ip += 1;
|
||||
for (n=0; n<oSize; n+=2)
|
||||
{
|
||||
huffWeight[n] = ip[n/2] >> 4;
|
||||
huffWeight[n+1] = ip[n/2] & 15;
|
||||
}
|
||||
}
|
||||
}
|
||||
else /* header compressed with FSE (normal case) */
|
||||
{
|
||||
if (iSize+1 > srcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
oSize = FSE_decompress(huffWeight, HUF_MAX_SYMBOL_VALUE, ip+1, iSize); /* max 255 values decoded, last one is implied */
|
||||
if (FSE_isError(oSize)) return oSize;
|
||||
}
|
||||
|
||||
/* collect weight stats */
|
||||
memset(rankVal, 0, sizeof(rankVal));
|
||||
weightTotal = 0;
|
||||
for (n=0; n<oSize; n++)
|
||||
{
|
||||
if (huffWeight[n] >= HUF_ABSOLUTEMAX_TABLELOG) return (size_t)-FSE_ERROR_corruptionDetected;
|
||||
rankVal[huffWeight[n]]++;
|
||||
weightTotal += (1 << huffWeight[n]) >> 1;
|
||||
}
|
||||
if (weightTotal == 0) return (size_t)-FSE_ERROR_corruptionDetected;
|
||||
|
||||
/* get last non-null symbol weight (implied, total must be 2^n) */
|
||||
maxBits = FSE_highbit32(weightTotal) + 1;
|
||||
if (maxBits > DTable[0]) return (size_t)-FSE_ERROR_tableLog_tooLarge; /* DTable is too small */
|
||||
DTable[0] = (U16)maxBits;
|
||||
{
|
||||
U32 total = 1 << maxBits;
|
||||
U32 rest = total - weightTotal;
|
||||
U32 verif = 1 << FSE_highbit32(rest);
|
||||
U32 lastWeight = FSE_highbit32(rest) + 1;
|
||||
if (verif != rest) return (size_t)-FSE_ERROR_corruptionDetected; /* last value must be a clean power of 2 */
|
||||
huffWeight[oSize] = (BYTE)lastWeight;
|
||||
rankVal[lastWeight]++;
|
||||
}
|
||||
|
||||
/* check tree construction validity */
|
||||
if ((rankVal[1] < 2) || (rankVal[1] & 1)) return (size_t)-FSE_ERROR_corruptionDetected; /* by construction : at least 2 elts of rank 1, must be even */
|
||||
|
||||
/* Prepare ranks */
|
||||
nextRankStart = 0;
|
||||
for (n=1; n<=maxBits; n++)
|
||||
{
|
||||
U32 current = nextRankStart;
|
||||
nextRankStart += (rankVal[n] << (n-1));
|
||||
rankVal[n] = current;
|
||||
}
|
||||
|
||||
/* fill DTable */
|
||||
for (n=0; n<=oSize; n++)
|
||||
{
|
||||
const U32 w = huffWeight[n];
|
||||
const U32 length = (1 << w) >> 1;
|
||||
U32 i;
|
||||
HUF_DElt D;
|
||||
D.byte = (BYTE)n; D.nbBits = (BYTE)(maxBits + 1 - w);
|
||||
for (i = rankVal[w]; i < rankVal[w] + length; i++)
|
||||
dt[i] = D;
|
||||
rankVal[w] += length;
|
||||
}
|
||||
|
||||
return iSize+1;
|
||||
}
|
||||
|
||||
|
||||
static BYTE HUF_decodeSymbol(FSE_DStream_t* Dstream, const HUF_DElt* dt, const U32 dtLog)
|
||||
{
|
||||
const size_t val = FSE_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */
|
||||
const BYTE c = dt[val].byte;
|
||||
FSE_skipBits(Dstream, dt[val].nbBits);
|
||||
return c;
|
||||
}
|
||||
|
||||
static size_t HUF_decompress_usingDTable( /* -3% slower when non static */
|
||||
void* dst, size_t maxDstSize,
|
||||
const void* cSrc, size_t cSrcSize,
|
||||
const U16* DTable)
|
||||
{
|
||||
if (cSrcSize < 6) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
{
|
||||
BYTE* const ostart = (BYTE*) dst;
|
||||
BYTE* op = ostart;
|
||||
BYTE* const omax = op + maxDstSize;
|
||||
BYTE* const olimit = maxDstSize < 15 ? op : omax-15;
|
||||
|
||||
const void* ptr = DTable;
|
||||
const HUF_DElt* const dt = (const HUF_DElt*)(ptr)+1;
|
||||
const U32 dtLog = DTable[0];
|
||||
size_t errorCode;
|
||||
U32 reloadStatus;
|
||||
|
||||
/* Init */
|
||||
|
||||
const U16* jumpTable = (const U16*)cSrc;
|
||||
const size_t length1 = FSE_readLE16(jumpTable);
|
||||
const size_t length2 = FSE_readLE16(jumpTable+1);
|
||||
const size_t length3 = FSE_readLE16(jumpTable+2);
|
||||
const size_t length4 = cSrcSize - 6 - length1 - length2 - length3; /* check coherency !! */
|
||||
const char* const start1 = (const char*)(cSrc) + 6;
|
||||
const char* const start2 = start1 + length1;
|
||||
const char* const start3 = start2 + length2;
|
||||
const char* const start4 = start3 + length3;
|
||||
FSE_DStream_t bitD1, bitD2, bitD3, bitD4;
|
||||
|
||||
if (length1+length2+length3+6 >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
|
||||
errorCode = FSE_initDStream(&bitD1, start1, length1);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
errorCode = FSE_initDStream(&bitD2, start2, length2);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
errorCode = FSE_initDStream(&bitD3, start3, length3);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
errorCode = FSE_initDStream(&bitD4, start4, length4);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
|
||||
reloadStatus=FSE_reloadDStream(&bitD2);
|
||||
|
||||
/* 16 symbols per loop */
|
||||
for ( ; (reloadStatus<FSE_DStream_completed) && (op<olimit); /* D2-3-4 are supposed to be synchronized and finish together */
|
||||
op+=16, reloadStatus = FSE_reloadDStream(&bitD2) | FSE_reloadDStream(&bitD3) | FSE_reloadDStream(&bitD4), FSE_reloadDStream(&bitD1))
|
||||
{
|
||||
#define HUF_DECODE_SYMBOL_0(n, Dstream) \
|
||||
op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog);
|
||||
|
||||
#define HUF_DECODE_SYMBOL_1(n, Dstream) \
|
||||
op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
|
||||
if (FSE_32bits() && (HUF_MAX_TABLELOG>12)) FSE_reloadDStream(&Dstream)
|
||||
|
||||
#define HUF_DECODE_SYMBOL_2(n, Dstream) \
|
||||
op[n] = HUF_decodeSymbol(&Dstream, dt, dtLog); \
|
||||
if (FSE_32bits()) FSE_reloadDStream(&Dstream)
|
||||
|
||||
HUF_DECODE_SYMBOL_1( 0, bitD1);
|
||||
HUF_DECODE_SYMBOL_1( 1, bitD2);
|
||||
HUF_DECODE_SYMBOL_1( 2, bitD3);
|
||||
HUF_DECODE_SYMBOL_1( 3, bitD4);
|
||||
HUF_DECODE_SYMBOL_2( 4, bitD1);
|
||||
HUF_DECODE_SYMBOL_2( 5, bitD2);
|
||||
HUF_DECODE_SYMBOL_2( 6, bitD3);
|
||||
HUF_DECODE_SYMBOL_2( 7, bitD4);
|
||||
HUF_DECODE_SYMBOL_1( 8, bitD1);
|
||||
HUF_DECODE_SYMBOL_1( 9, bitD2);
|
||||
HUF_DECODE_SYMBOL_1(10, bitD3);
|
||||
HUF_DECODE_SYMBOL_1(11, bitD4);
|
||||
HUF_DECODE_SYMBOL_0(12, bitD1);
|
||||
HUF_DECODE_SYMBOL_0(13, bitD2);
|
||||
HUF_DECODE_SYMBOL_0(14, bitD3);
|
||||
HUF_DECODE_SYMBOL_0(15, bitD4);
|
||||
}
|
||||
|
||||
if (reloadStatus!=FSE_DStream_completed) /* not complete : some bitStream might be FSE_DStream_unfinished */
|
||||
return (size_t)-FSE_ERROR_corruptionDetected;
|
||||
|
||||
/* tail */
|
||||
{
|
||||
/* bitTail = bitD1; */ /* *much* slower : -20% !??! */
|
||||
FSE_DStream_t bitTail;
|
||||
bitTail.ptr = bitD1.ptr;
|
||||
bitTail.bitsConsumed = bitD1.bitsConsumed;
|
||||
bitTail.bitContainer = bitD1.bitContainer; /* required in case of FSE_DStream_endOfBuffer */
|
||||
bitTail.start = start1;
|
||||
for ( ; (FSE_reloadDStream(&bitTail) < FSE_DStream_completed) && (op<omax) ; op++)
|
||||
{
|
||||
HUF_DECODE_SYMBOL_0(0, bitTail);
|
||||
}
|
||||
|
||||
if (FSE_endOfDStream(&bitTail))
|
||||
return op-ostart;
|
||||
}
|
||||
|
||||
if (op==omax) return (size_t)-FSE_ERROR_dstSize_tooSmall; /* dst buffer is full, but cSrc unfinished */
|
||||
|
||||
return (size_t)-FSE_ERROR_corruptionDetected;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
static size_t HUF_decompress (void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize)
|
||||
{
|
||||
HUF_CREATE_STATIC_DTABLE(DTable, HUF_MAX_TABLELOG);
|
||||
const BYTE* ip = (const BYTE*) cSrc;
|
||||
size_t errorCode;
|
||||
|
||||
errorCode = HUF_readDTable (DTable, cSrc, cSrcSize);
|
||||
if (FSE_isError(errorCode)) return errorCode;
|
||||
if (errorCode >= cSrcSize) return (size_t)-FSE_ERROR_srcSize_wrong;
|
||||
ip += errorCode;
|
||||
cSrcSize -= errorCode;
|
||||
|
||||
return HUF_decompress_usingDTable (dst, maxDstSize, ip, cSrcSize, DTable);
|
||||
}
|
||||
|
||||
|
||||
#endif /* FSE_COMMONDEFS_ONLY */
|
||||
|
||||
/*
|
||||
zstd - standard compression library
|
||||
Copyright (C) 2014-2015, Yann Collet.
|
||||
|
||||
BSD 2-Clause License (https://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.
|
||||
|
||||
You can contact the author at :
|
||||
- zstd source repository : https://github.com/Cyan4973/zstd
|
||||
- ztsd public forum : https://groups.google.com/forum/#!forum/lz4c
|
||||
*/
|
||||
|
||||
/****************************************************************
|
||||
* Tuning parameters
|
||||
*****************************************************************/
|
||||
/* MEMORY_USAGE :
|
||||
* Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
|
||||
* Increasing memory usage improves compression ratio
|
||||
* Reduced memory usage can improve speed, due to cache effect */
|
||||
#define ZSTD_MEMORY_USAGE 17
|
||||
|
||||
|
||||
/**************************************
|
||||
CPU Feature Detection
|
||||
**************************************/
|
||||
/*
|
||||
* Automated efficient unaligned memory access detection
|
||||
* Based on known hardware architectures
|
||||
* This list will be updated thanks to feedbacks
|
||||
*/
|
||||
#if defined(CPU_HAS_EFFICIENT_UNALIGNED_MEMORY_ACCESS) \
|
||||
|| defined(__ARM_FEATURE_UNALIGNED) \
|
||||
|| defined(__i386__) || defined(__x86_64__) \
|
||||
|| defined(_M_IX86) || defined(_M_X64) \
|
||||
|| defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_8__) \
|
||||
|| (defined(_M_ARM) && (_M_ARM >= 7))
|
||||
# define ZSTD_UNALIGNED_ACCESS 1
|
||||
#else
|
||||
# define ZSTD_UNALIGNED_ACCESS 0
|
||||
#endif
|
||||
|
||||
|
||||
/********************************************************
|
||||
* Includes
|
||||
*********************************************************/
|
||||
#include <stdlib.h> /* calloc */
|
||||
#include <string.h> /* memcpy, memmove */
|
||||
#include <stdio.h> /* debug : printf */
|
||||
|
||||
|
||||
/********************************************************
|
||||
* Compiler specifics
|
||||
*********************************************************/
|
||||
#ifdef __AVX2__
|
||||
# include <immintrin.h> /* AVX2 intrinsics */
|
||||
#endif
|
||||
|
||||
#ifdef _MSC_VER /* Visual Studio */
|
||||
# include <intrin.h> /* For Visual 2005 */
|
||||
# pragma warning(disable : 4127) /* disable: C4127: conditional expression is constant */
|
||||
# pragma warning(disable : 4324) /* disable: C4324: padded structure */
|
||||
#endif
|
||||
|
||||
|
||||
#ifndef MEM_ACCESS_MODULE
|
||||
#define MEM_ACCESS_MODULE
|
||||
/********************************************************
|
||||
* Basic Types
|
||||
*********************************************************/
|
||||
#if defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L /* C99 */
|
||||
# if defined(_AIX)
|
||||
# include <inttypes.h>
|
||||
# else
|
||||
# include <stdint.h> /* intptr_t */
|
||||
# endif
|
||||
typedef uint8_t BYTE;
|
||||
typedef uint16_t U16;
|
||||
typedef int16_t S16;
|
||||
typedef uint32_t U32;
|
||||
typedef int32_t S32;
|
||||
typedef uint64_t U64;
|
||||
#else
|
||||
typedef unsigned char BYTE;
|
||||
typedef unsigned short U16;
|
||||
typedef signed short S16;
|
||||
typedef unsigned int U32;
|
||||
typedef signed int S32;
|
||||
typedef unsigned long long U64;
|
||||
#endif
|
||||
|
||||
#endif /* MEM_ACCESS_MODULE */
|
||||
|
||||
|
||||
/********************************************************
|
||||
* Constants
|
||||
*********************************************************/
|
||||
static const U32 ZSTD_magicNumber = 0xFD2FB51E; /* 3rd version : seqNb header */
|
||||
|
||||
#define HASH_LOG (ZSTD_MEMORY_USAGE - 2)
|
||||
#define HASH_TABLESIZE (1 << HASH_LOG)
|
||||
#define HASH_MASK (HASH_TABLESIZE - 1)
|
||||
|
||||
#define KNUTH 2654435761
|
||||
|
||||
#define BIT7 128
|
||||
#define BIT6 64
|
||||
#define BIT5 32
|
||||
#define BIT4 16
|
||||
|
||||
#define KB *(1 <<10)
|
||||
#define MB *(1 <<20)
|
||||
#define GB *(1U<<30)
|
||||
|
||||
#define BLOCKSIZE (128 KB) /* define, for static allocation */
|
||||
|
||||
#define WORKPLACESIZE (BLOCKSIZE*3)
|
||||
#define MINMATCH 4
|
||||
#define MLbits 7
|
||||
#define LLbits 6
|
||||
#define Offbits 5
|
||||
#define MaxML ((1<<MLbits )-1)
|
||||
#define MaxLL ((1<<LLbits )-1)
|
||||
#define MaxOff ((1<<Offbits)-1)
|
||||
#define LitFSELog 11
|
||||
#define MLFSELog 10
|
||||
#define LLFSELog 10
|
||||
#define OffFSELog 9
|
||||
#define MAX(a,b) ((a)<(b)?(b):(a))
|
||||
#define MaxSeq MAX(MaxLL, MaxML)
|
||||
|
||||
#define LITERAL_NOENTROPY 63
|
||||
#define COMMAND_NOENTROPY 7 /* to remove */
|
||||
|
||||
#define ZSTD_CONTENTSIZE_ERROR (0ULL - 2)
|
||||
|
||||
static const size_t ZSTD_blockHeaderSize = 3;
|
||||
static const size_t ZSTD_frameHeaderSize = 4;
|
||||
|
||||
|
||||
/********************************************************
|
||||
* Memory operations
|
||||
*********************************************************/
|
||||
static unsigned ZSTD_32bits(void) { return sizeof(void*)==4; }
|
||||
|
||||
static unsigned ZSTD_isLittleEndian(void)
|
||||
{
|
||||
const union { U32 i; BYTE c[4]; } one = { 1 }; /* don't use static : performance detrimental */
|
||||
return one.c[0];
|
||||
}
|
||||
|
||||
static U16 ZSTD_read16(const void* p) { U16 r; memcpy(&r, p, sizeof(r)); return r; }
|
||||
|
||||
static void ZSTD_copy4(void* dst, const void* src) { memcpy(dst, src, 4); }
|
||||
|
||||
static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
|
||||
|
||||
#define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; }
|
||||
|
||||
static void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length)
|
||||
{
|
||||
const BYTE* ip = (const BYTE*)src;
|
||||
BYTE* op = (BYTE*)dst;
|
||||
BYTE* const oend = op + length;
|
||||
while (op < oend) COPY8(op, ip);
|
||||
}
|
||||
|
||||
static U16 ZSTD_readLE16(const void* memPtr)
|
||||
{
|
||||
if (ZSTD_isLittleEndian()) return ZSTD_read16(memPtr);
|
||||
else
|
||||
{
|
||||
const BYTE* p = (const BYTE*)memPtr;
|
||||
return (U16)((U16)p[0] + ((U16)p[1]<<8));
|
||||
}
|
||||
}
|
||||
|
||||
static U32 ZSTD_readLE24(const void* memPtr)
|
||||
{
|
||||
return ZSTD_readLE16(memPtr) + (((const BYTE*)memPtr)[2] << 16);
|
||||
}
|
||||
|
||||
static U32 ZSTD_readBE32(const void* memPtr)
|
||||
{
|
||||
const BYTE* p = (const BYTE*)memPtr;
|
||||
return (U32)(((U32)p[0]<<24) + ((U32)p[1]<<16) + ((U32)p[2]<<8) + ((U32)p[3]<<0));
|
||||
}
|
||||
|
||||
|
||||
/**************************************
|
||||
* Local structures
|
||||
***************************************/
|
||||
typedef struct ZSTD_Cctx_s ZSTD_Cctx;
|
||||
|
||||
typedef enum { bt_compressed, bt_raw, bt_rle, bt_end } blockType_t;
|
||||
|
||||
typedef struct
|
||||
{
|
||||
blockType_t blockType;
|
||||
U32 origSize;
|
||||
} blockProperties_t;
|
||||
|
||||
typedef struct {
|
||||
void* buffer;
|
||||
U32* offsetStart;
|
||||
U32* offset;
|
||||
BYTE* offCodeStart;
|
||||
BYTE* offCode;
|
||||
BYTE* litStart;
|
||||
BYTE* lit;
|
||||
BYTE* litLengthStart;
|
||||
BYTE* litLength;
|
||||
BYTE* matchLengthStart;
|
||||
BYTE* matchLength;
|
||||
BYTE* dumpsStart;
|
||||
BYTE* dumps;
|
||||
} SeqStore_t;
|
||||
|
||||
|
||||
typedef struct ZSTD_Cctx_s
|
||||
{
|
||||
const BYTE* base;
|
||||
U32 current;
|
||||
U32 nextUpdate;
|
||||
SeqStore_t seqStore;
|
||||
#ifdef __AVX2__
|
||||
__m256i hashTable[HASH_TABLESIZE>>3];
|
||||
#else
|
||||
U32 hashTable[HASH_TABLESIZE];
|
||||
#endif
|
||||
BYTE buffer[WORKPLACESIZE];
|
||||
} cctxi_t;
|
||||
|
||||
|
||||
|
||||
|
||||
/**************************************
|
||||
* Error Management
|
||||
**************************************/
|
||||
/* published entry point */
|
||||
unsigned ZSTDv01_isError(size_t code) { return ERR_isError(code); }
|
||||
|
||||
|
||||
/**************************************
|
||||
* Tool functions
|
||||
**************************************/
|
||||
#define ZSTD_VERSION_MAJOR 0 /* for breaking interface changes */
|
||||
#define ZSTD_VERSION_MINOR 1 /* for new (non-breaking) interface capabilities */
|
||||
#define ZSTD_VERSION_RELEASE 3 /* for tweaks, bug-fixes, or development */
|
||||
#define ZSTD_VERSION_NUMBER (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
|
||||
|
||||
/**************************************************************
|
||||
* Decompression code
|
||||
**************************************************************/
|
||||
|
||||
static size_t ZSTDv01_getcBlockSize(const void* src, size_t srcSize, blockProperties_t* bpPtr)
|
||||
{
|
||||
const BYTE* const in = (const BYTE* const)src;
|
||||
BYTE headerFlags;
|
||||
U32 cSize;
|
||||
|
||||
if (srcSize < 3) return ERROR(srcSize_wrong);
|
||||
|
||||
headerFlags = *in;
|
||||
cSize = in[2] + (in[1]<<8) + ((in[0] & 7)<<16);
|
||||
|
||||
bpPtr->blockType = (blockType_t)(headerFlags >> 6);
|
||||
bpPtr->origSize = (bpPtr->blockType == bt_rle) ? cSize : 0;
|
||||
|
||||
if (bpPtr->blockType == bt_end) return 0;
|
||||
if (bpPtr->blockType == bt_rle) return 1;
|
||||
return cSize;
|
||||
}
|
||||
|
||||
|
||||
static size_t ZSTD_copyUncompressedBlock(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
|
||||
{
|
||||
if (srcSize > maxDstSize) return ERROR(dstSize_tooSmall);
|
||||
if (srcSize > 0) {
|
||||
memcpy(dst, src, srcSize);
|
||||
}
|
||||
return srcSize;
|
||||
}
|
||||
|
||||
|
||||
static size_t ZSTD_decompressLiterals(void* ctx,
|
||||
void* dst, size_t maxDstSize,
|
||||
const void* src, size_t srcSize)
|
||||
{
|
||||
BYTE* op = (BYTE*)dst;
|
||||
BYTE* const oend = op + maxDstSize;
|
||||
const BYTE* ip = (const BYTE*)src;
|
||||
size_t errorCode;
|
||||
size_t litSize;
|
||||
|
||||
/* check : minimum 2, for litSize, +1, for content */
|
||||
if (srcSize <= 3) return ERROR(corruption_detected);
|
||||
|
||||
litSize = ip[1] + (ip[0]<<8);
|
||||
litSize += ((ip[-3] >> 3) & 7) << 16; /* mmmmh.... */
|
||||
op = oend - litSize;
|
||||
|
||||
(void)ctx;
|
||||
if (litSize > maxDstSize) return ERROR(dstSize_tooSmall);
|
||||
errorCode = HUF_decompress(op, litSize, ip+2, srcSize-2);
|
||||
if (FSE_isError(errorCode)) return ERROR(GENERIC);
|
||||
return litSize;
|
||||
}
|
||||
|
||||
|
||||
static size_t ZSTDv01_decodeLiteralsBlock(void* ctx,
|
||||
void* dst, size_t maxDstSize,
|
||||
const BYTE** litStart, size_t* litSize,
|
||||
const void* src, size_t srcSize)
|
||||
{
|
||||
const BYTE* const istart = (const BYTE* const)src;
|
||||
const BYTE* ip = istart;
|
||||
BYTE* const ostart = (BYTE* const)dst;
|
||||
BYTE* const oend = ostart + maxDstSize;
|
||||
blockProperties_t litbp;
|
||||
|
||||
size_t litcSize = ZSTDv01_getcBlockSize(src, srcSize, &litbp);
|
||||
if (ZSTDv01_isError(litcSize)) return litcSize;
|
||||
if (litcSize > srcSize - ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
|
||||
ip += ZSTD_blockHeaderSize;
|
||||
|
||||
switch(litbp.blockType)
|
||||
{
|
||||
case bt_raw:
|
||||
*litStart = ip;
|
||||
ip += litcSize;
|
||||
*litSize = litcSize;
|
||||
break;
|
||||
case bt_rle:
|
||||
{
|
||||
size_t rleSize = litbp.origSize;
|
||||
if (rleSize>maxDstSize) return ERROR(dstSize_tooSmall);
|
||||
if (!srcSize) return ERROR(srcSize_wrong);
|
||||
if (rleSize > 0) {
|
||||
memset(oend - rleSize, *ip, rleSize);
|
||||
}
|
||||
*litStart = oend - rleSize;
|
||||
*litSize = rleSize;
|
||||
ip++;
|
||||
break;
|
||||
}
|
||||
case bt_compressed:
|
||||
{
|
||||
size_t decodedLitSize = ZSTD_decompressLiterals(ctx, dst, maxDstSize, ip, litcSize);
|
||||
if (ZSTDv01_isError(decodedLitSize)) return decodedLitSize;
|
||||
*litStart = oend - decodedLitSize;
|
||||
*litSize = decodedLitSize;
|
||||
ip += litcSize;
|
||||
break;
|
||||
}
|
||||
case bt_end:
|
||||
default:
|
||||
return ERROR(GENERIC);
|
||||
}
|
||||
|
||||
return ip-istart;
|
||||
}
|
||||
|
||||
|
||||
static size_t ZSTDv01_decodeSeqHeaders(int* nbSeq, const BYTE** dumpsPtr, size_t* dumpsLengthPtr,
|
||||
FSE_DTable* DTableLL, FSE_DTable* DTableML, FSE_DTable* DTableOffb,
|
||||
const void* src, size_t srcSize)
|
||||
{
|
||||
const BYTE* const istart = (const BYTE* const)src;
|
||||
const BYTE* ip = istart;
|
||||
const BYTE* const iend = istart + srcSize;
|
||||
U32 LLtype, Offtype, MLtype;
|
||||
U32 LLlog, Offlog, MLlog;
|
||||
size_t dumpsLength;
|
||||
|
||||
/* check */
|
||||
if (srcSize < 5) return ERROR(srcSize_wrong);
|
||||
|
||||
/* SeqHead */
|
||||
*nbSeq = ZSTD_readLE16(ip); ip+=2;
|
||||
LLtype = *ip >> 6;
|
||||
Offtype = (*ip >> 4) & 3;
|
||||
MLtype = (*ip >> 2) & 3;
|
||||
if (*ip & 2)
|
||||
{
|
||||
dumpsLength = ip[2];
|
||||
dumpsLength += ip[1] << 8;
|
||||
ip += 3;
|
||||
}
|
||||
else
|
||||
{
|
||||
dumpsLength = ip[1];
|
||||
dumpsLength += (ip[0] & 1) << 8;
|
||||
ip += 2;
|
||||
}
|
||||
*dumpsPtr = ip;
|
||||
ip += dumpsLength;
|
||||
*dumpsLengthPtr = dumpsLength;
|
||||
|
||||
/* check */
|
||||
if (ip > iend-3) return ERROR(srcSize_wrong); /* min : all 3 are "raw", hence no header, but at least xxLog bits per type */
|
||||
|
||||
/* sequences */
|
||||
{
|
||||
S16 norm[MaxML+1]; /* assumption : MaxML >= MaxLL and MaxOff */
|
||||
size_t headerSize;
|
||||
|
||||
/* Build DTables */
|
||||
switch(LLtype)
|
||||
{
|
||||
case bt_rle :
|
||||
LLlog = 0;
|
||||
FSE_buildDTable_rle(DTableLL, *ip++); break;
|
||||
case bt_raw :
|
||||
LLlog = LLbits;
|
||||
FSE_buildDTable_raw(DTableLL, LLbits); break;
|
||||
default :
|
||||
{ U32 max = MaxLL;
|
||||
headerSize = FSE_readNCount(norm, &max, &LLlog, ip, iend-ip);
|
||||
if (FSE_isError(headerSize)) return ERROR(GENERIC);
|
||||
if (LLlog > LLFSELog) return ERROR(corruption_detected);
|
||||
ip += headerSize;
|
||||
FSE_buildDTable(DTableLL, norm, max, LLlog);
|
||||
} }
|
||||
|
||||
switch(Offtype)
|
||||
{
|
||||
case bt_rle :
|
||||
Offlog = 0;
|
||||
if (ip > iend-2) return ERROR(srcSize_wrong); /* min : "raw", hence no header, but at least xxLog bits */
|
||||
FSE_buildDTable_rle(DTableOffb, *ip++); break;
|
||||
case bt_raw :
|
||||
Offlog = Offbits;
|
||||
FSE_buildDTable_raw(DTableOffb, Offbits); break;
|
||||
default :
|
||||
{ U32 max = MaxOff;
|
||||
headerSize = FSE_readNCount(norm, &max, &Offlog, ip, iend-ip);
|
||||
if (FSE_isError(headerSize)) return ERROR(GENERIC);
|
||||
if (Offlog > OffFSELog) return ERROR(corruption_detected);
|
||||
ip += headerSize;
|
||||
FSE_buildDTable(DTableOffb, norm, max, Offlog);
|
||||
} }
|
||||
|
||||
switch(MLtype)
|
||||
{
|
||||
case bt_rle :
|
||||
MLlog = 0;
|
||||
if (ip > iend-2) return ERROR(srcSize_wrong); /* min : "raw", hence no header, but at least xxLog bits */
|
||||
FSE_buildDTable_rle(DTableML, *ip++); break;
|
||||
case bt_raw :
|
||||
MLlog = MLbits;
|
||||
FSE_buildDTable_raw(DTableML, MLbits); break;
|
||||
default :
|
||||
{ U32 max = MaxML;
|
||||
headerSize = FSE_readNCount(norm, &max, &MLlog, ip, iend-ip);
|
||||
if (FSE_isError(headerSize)) return ERROR(GENERIC);
|
||||
if (MLlog > MLFSELog) return ERROR(corruption_detected);
|
||||
ip += headerSize;
|
||||
FSE_buildDTable(DTableML, norm, max, MLlog);
|
||||
} } }
|
||||
|
||||
return ip-istart;
|
||||
}
|
||||
|
||||
|
||||
typedef struct {
|
||||
size_t litLength;
|
||||
size_t offset;
|
||||
size_t matchLength;
|
||||
} seq_t;
|
||||
|
||||
typedef struct {
|
||||
FSE_DStream_t DStream;
|
||||
FSE_DState_t stateLL;
|
||||
FSE_DState_t stateOffb;
|
||||
FSE_DState_t stateML;
|
||||
size_t prevOffset;
|
||||
const BYTE* dumps;
|
||||
const BYTE* dumpsEnd;
|
||||
} seqState_t;
|
||||
|
||||
|
||||
static void ZSTD_decodeSequence(seq_t* seq, seqState_t* seqState)
|
||||
{
|
||||
size_t litLength;
|
||||
size_t prevOffset;
|
||||
size_t offset;
|
||||
size_t matchLength;
|
||||
const BYTE* dumps = seqState->dumps;
|
||||
const BYTE* const de = seqState->dumpsEnd;
|
||||
|
||||
/* Literal length */
|
||||
litLength = FSE_decodeSymbol(&(seqState->stateLL), &(seqState->DStream));
|
||||
prevOffset = litLength ? seq->offset : seqState->prevOffset;
|
||||
seqState->prevOffset = seq->offset;
|
||||
if (litLength == MaxLL)
|
||||
{
|
||||
const U32 add = dumps<de ? *dumps++ : 0;
|
||||
if (add < 255) litLength += add;
|
||||
else
|
||||
{
|
||||
if (dumps<=(de-3))
|
||||
{
|
||||
litLength = ZSTD_readLE24(dumps);
|
||||
dumps += 3;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/* Offset */
|
||||
{
|
||||
U32 offsetCode, nbBits;
|
||||
offsetCode = FSE_decodeSymbol(&(seqState->stateOffb), &(seqState->DStream));
|
||||
if (ZSTD_32bits()) FSE_reloadDStream(&(seqState->DStream));
|
||||
nbBits = offsetCode - 1;
|
||||
if (offsetCode==0) nbBits = 0; /* cmove */
|
||||
offset = ((size_t)1 << (nbBits & ((sizeof(offset)*8)-1))) + FSE_readBits(&(seqState->DStream), nbBits);
|
||||
if (ZSTD_32bits()) FSE_reloadDStream(&(seqState->DStream));
|
||||
if (offsetCode==0) offset = prevOffset;
|
||||
}
|
||||
|
||||
/* MatchLength */
|
||||
matchLength = FSE_decodeSymbol(&(seqState->stateML), &(seqState->DStream));
|
||||
if (matchLength == MaxML)
|
||||
{
|
||||
const U32 add = dumps<de ? *dumps++ : 0;
|
||||
if (add < 255) matchLength += add;
|
||||
else
|
||||
{
|
||||
if (dumps<=(de-3))
|
||||
{
|
||||
matchLength = ZSTD_readLE24(dumps);
|
||||
dumps += 3;
|
||||
}
|
||||
}
|
||||
}
|
||||
matchLength += MINMATCH;
|
||||
|
||||
/* save result */
|
||||
seq->litLength = litLength;
|
||||
seq->offset = offset;
|
||||
seq->matchLength = matchLength;
|
||||
seqState->dumps = dumps;
|
||||
}
|
||||
|
||||
|
||||
static size_t ZSTD_execSequence(BYTE* op,
|
||||
seq_t sequence,
|
||||
const BYTE** litPtr, const BYTE* const litLimit,
|
||||
BYTE* const base, BYTE* const oend)
|
||||
{
|
||||
static const int dec32table[] = {0, 1, 2, 1, 4, 4, 4, 4}; /* added */
|
||||
static const int dec64table[] = {8, 8, 8, 7, 8, 9,10,11}; /* subtracted */
|
||||
const BYTE* const ostart = op;
|
||||
BYTE* const oLitEnd = op + sequence.litLength;
|
||||
const size_t litLength = sequence.litLength;
|
||||
BYTE* const endMatch = op + litLength + sequence.matchLength; /* risk : address space overflow (32-bits) */
|
||||
const BYTE* const litEnd = *litPtr + litLength;
|
||||
|
||||
/* checks */
|
||||
size_t const seqLength = sequence.litLength + sequence.matchLength;
|
||||
|
||||
if (seqLength > (size_t)(oend - op)) return ERROR(dstSize_tooSmall);
|
||||
if (sequence.litLength > (size_t)(litLimit - *litPtr)) return ERROR(corruption_detected);
|
||||
/* Now we know there are no overflow in literal nor match lengths, can use pointer checks */
|
||||
if (sequence.offset > (U32)(oLitEnd - base)) return ERROR(corruption_detected);
|
||||
|
||||
if (endMatch > oend) return ERROR(dstSize_tooSmall); /* overwrite beyond dst buffer */
|
||||
if (litEnd > litLimit) return ERROR(corruption_detected); /* overRead beyond lit buffer */
|
||||
if (sequence.matchLength > (size_t)(*litPtr-op)) return ERROR(dstSize_tooSmall); /* overwrite literal segment */
|
||||
|
||||
/* copy Literals */
|
||||
ZSTD_memmove(op, *litPtr, sequence.litLength); /* note : v0.1 seems to allow scenarios where output or input are close to end of buffer */
|
||||
|
||||
op += litLength;
|
||||
*litPtr = litEnd; /* update for next sequence */
|
||||
|
||||
/* check : last match must be at a minimum distance of 8 from end of dest buffer */
|
||||
if (oend-op < 8) return ERROR(dstSize_tooSmall);
|
||||
|
||||
/* copy Match */
|
||||
{
|
||||
const U32 overlapRisk = (((size_t)(litEnd - endMatch)) < 12);
|
||||
const BYTE* match = op - sequence.offset; /* possible underflow at op - offset ? */
|
||||
size_t qutt = 12;
|
||||
U64 saved[2];
|
||||
|
||||
/* check */
|
||||
if (match < base) return ERROR(corruption_detected);
|
||||
if (sequence.offset > (size_t)base) return ERROR(corruption_detected);
|
||||
|
||||
/* save beginning of literal sequence, in case of write overlap */
|
||||
if (overlapRisk)
|
||||
{
|
||||
if ((endMatch + qutt) > oend) qutt = oend-endMatch;
|
||||
memcpy(saved, endMatch, qutt);
|
||||
}
|
||||
|
||||
if (sequence.offset < 8)
|
||||
{
|
||||
const int dec64 = dec64table[sequence.offset];
|
||||
op[0] = match[0];
|
||||
op[1] = match[1];
|
||||
op[2] = match[2];
|
||||
op[3] = match[3];
|
||||
match += dec32table[sequence.offset];
|
||||
ZSTD_copy4(op+4, match);
|
||||
match -= dec64;
|
||||
} else { ZSTD_copy8(op, match); }
|
||||
op += 8; match += 8;
|
||||
|
||||
if (endMatch > oend-(16-MINMATCH))
|
||||
{
|
||||
if (op < oend-8)
|
||||
{
|
||||
ZSTD_wildcopy(op, match, (oend-8) - op);
|
||||
match += (oend-8) - op;
|
||||
op = oend-8;
|
||||
}
|
||||
while (op<endMatch) *op++ = *match++;
|
||||
}
|
||||
else
|
||||
ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8); /* works even if matchLength < 8 */
|
||||
|
||||
/* restore, in case of overlap */
|
||||
if (overlapRisk) memcpy(endMatch, saved, qutt);
|
||||
}
|
||||
|
||||
return endMatch-ostart;
|
||||
}
|
||||
|
||||
typedef struct ZSTDv01_Dctx_s
|
||||
{
|
||||
U32 LLTable[FSE_DTABLE_SIZE_U32(LLFSELog)];
|
||||
U32 OffTable[FSE_DTABLE_SIZE_U32(OffFSELog)];
|
||||
U32 MLTable[FSE_DTABLE_SIZE_U32(MLFSELog)];
|
||||
void* previousDstEnd;
|
||||
void* base;
|
||||
size_t expected;
|
||||
blockType_t bType;
|
||||
U32 phase;
|
||||
} dctx_t;
|
||||
|
||||
|
||||
static size_t ZSTD_decompressSequences(
|
||||
void* ctx,
|
||||
void* dst, size_t maxDstSize,
|
||||
const void* seqStart, size_t seqSize,
|
||||
const BYTE* litStart, size_t litSize)
|
||||
{
|
||||
dctx_t* dctx = (dctx_t*)ctx;
|
||||
const BYTE* ip = (const BYTE*)seqStart;
|
||||
const BYTE* const iend = ip + seqSize;
|
||||
BYTE* const ostart = (BYTE* const)dst;
|
||||
BYTE* op = ostart;
|
||||
BYTE* const oend = ostart + maxDstSize;
|
||||
size_t errorCode, dumpsLength;
|
||||
const BYTE* litPtr = litStart;
|
||||
const BYTE* const litEnd = litStart + litSize;
|
||||
int nbSeq;
|
||||
const BYTE* dumps;
|
||||
U32* DTableLL = dctx->LLTable;
|
||||
U32* DTableML = dctx->MLTable;
|
||||
U32* DTableOffb = dctx->OffTable;
|
||||
BYTE* const base = (BYTE*) (dctx->base);
|
||||
|
||||
/* Build Decoding Tables */
|
||||
errorCode = ZSTDv01_decodeSeqHeaders(&nbSeq, &dumps, &dumpsLength,
|
||||
DTableLL, DTableML, DTableOffb,
|
||||
ip, iend-ip);
|
||||
if (ZSTDv01_isError(errorCode)) return errorCode;
|
||||
ip += errorCode;
|
||||
|
||||
/* Regen sequences */
|
||||
{
|
||||
seq_t sequence;
|
||||
seqState_t seqState;
|
||||
|
||||
memset(&sequence, 0, sizeof(sequence));
|
||||
seqState.dumps = dumps;
|
||||
seqState.dumpsEnd = dumps + dumpsLength;
|
||||
seqState.prevOffset = 1;
|
||||
errorCode = FSE_initDStream(&(seqState.DStream), ip, iend-ip);
|
||||
if (FSE_isError(errorCode)) return ERROR(corruption_detected);
|
||||
FSE_initDState(&(seqState.stateLL), &(seqState.DStream), DTableLL);
|
||||
FSE_initDState(&(seqState.stateOffb), &(seqState.DStream), DTableOffb);
|
||||
FSE_initDState(&(seqState.stateML), &(seqState.DStream), DTableML);
|
||||
|
||||
for ( ; (FSE_reloadDStream(&(seqState.DStream)) <= FSE_DStream_completed) && (nbSeq>0) ; )
|
||||
{
|
||||
size_t oneSeqSize;
|
||||
nbSeq--;
|
||||
ZSTD_decodeSequence(&sequence, &seqState);
|
||||
oneSeqSize = ZSTD_execSequence(op, sequence, &litPtr, litEnd, base, oend);
|
||||
if (ZSTDv01_isError(oneSeqSize)) return oneSeqSize;
|
||||
op += oneSeqSize;
|
||||
}
|
||||
|
||||
/* check if reached exact end */
|
||||
if ( !FSE_endOfDStream(&(seqState.DStream)) ) return ERROR(corruption_detected); /* requested too much : data is corrupted */
|
||||
if (nbSeq<0) return ERROR(corruption_detected); /* requested too many sequences : data is corrupted */
|
||||
|
||||
/* last literal segment */
|
||||
{
|
||||
size_t lastLLSize = litEnd - litPtr;
|
||||
if (op+lastLLSize > oend) return ERROR(dstSize_tooSmall);
|
||||
if (lastLLSize > 0) {
|
||||
if (op != litPtr) memmove(op, litPtr, lastLLSize);
|
||||
op += lastLLSize;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return op-ostart;
|
||||
}
|
||||
|
||||
|
||||
static size_t ZSTD_decompressBlock(
|
||||
void* ctx,
|
||||
void* dst, size_t maxDstSize,
|
||||
const void* src, size_t srcSize)
|
||||
{
|
||||
/* blockType == blockCompressed, srcSize is trusted */
|
||||
const BYTE* ip = (const BYTE*)src;
|
||||
const BYTE* litPtr = NULL;
|
||||
size_t litSize = 0;
|
||||
size_t errorCode;
|
||||
|
||||
/* Decode literals sub-block */
|
||||
errorCode = ZSTDv01_decodeLiteralsBlock(ctx, dst, maxDstSize, &litPtr, &litSize, src, srcSize);
|
||||
if (ZSTDv01_isError(errorCode)) return errorCode;
|
||||
ip += errorCode;
|
||||
srcSize -= errorCode;
|
||||
|
||||
return ZSTD_decompressSequences(ctx, dst, maxDstSize, ip, srcSize, litPtr, litSize);
|
||||
}
|
||||
|
||||
|
||||
size_t ZSTDv01_decompressDCtx(void* ctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
|
||||
{
|
||||
const BYTE* ip = (const BYTE*)src;
|
||||
const BYTE* iend = ip + srcSize;
|
||||
BYTE* const ostart = (BYTE* const)dst;
|
||||
BYTE* op = ostart;
|
||||
BYTE* const oend = ostart + maxDstSize;
|
||||
size_t remainingSize = srcSize;
|
||||
U32 magicNumber;
|
||||
size_t errorCode=0;
|
||||
blockProperties_t blockProperties;
|
||||
|
||||
/* Frame Header */
|
||||
if (srcSize < ZSTD_frameHeaderSize+ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
|
||||
magicNumber = ZSTD_readBE32(src);
|
||||
if (magicNumber != ZSTD_magicNumber) return ERROR(prefix_unknown);
|
||||
ip += ZSTD_frameHeaderSize; remainingSize -= ZSTD_frameHeaderSize;
|
||||
|
||||
/* Loop on each block */
|
||||
while (1)
|
||||
{
|
||||
size_t blockSize = ZSTDv01_getcBlockSize(ip, iend-ip, &blockProperties);
|
||||
if (ZSTDv01_isError(blockSize)) return blockSize;
|
||||
|
||||
ip += ZSTD_blockHeaderSize;
|
||||
remainingSize -= ZSTD_blockHeaderSize;
|
||||
if (blockSize > remainingSize) return ERROR(srcSize_wrong);
|
||||
|
||||
switch(blockProperties.blockType)
|
||||
{
|
||||
case bt_compressed:
|
||||
errorCode = ZSTD_decompressBlock(ctx, op, oend-op, ip, blockSize);
|
||||
break;
|
||||
case bt_raw :
|
||||
errorCode = ZSTD_copyUncompressedBlock(op, oend-op, ip, blockSize);
|
||||
break;
|
||||
case bt_rle :
|
||||
return ERROR(GENERIC); /* not yet supported */
|
||||
break;
|
||||
case bt_end :
|
||||
/* end of frame */
|
||||
if (remainingSize) return ERROR(srcSize_wrong);
|
||||
break;
|
||||
default:
|
||||
return ERROR(GENERIC);
|
||||
}
|
||||
if (blockSize == 0) break; /* bt_end */
|
||||
|
||||
if (ZSTDv01_isError(errorCode)) return errorCode;
|
||||
op += errorCode;
|
||||
ip += blockSize;
|
||||
remainingSize -= blockSize;
|
||||
}
|
||||
|
||||
return op-ostart;
|
||||
}
|
||||
|
||||
size_t ZSTDv01_decompress(void* dst, size_t maxDstSize, const void* src, size_t srcSize)
|
||||
{
|
||||
dctx_t ctx;
|
||||
ctx.base = dst;
|
||||
return ZSTDv01_decompressDCtx(&ctx, dst, maxDstSize, src, srcSize);
|
||||
}
|
||||
|
||||
/* ZSTD_errorFrameSizeInfoLegacy() :
|
||||
assumes `cSize` and `dBound` are _not_ NULL */
|
||||
static void ZSTD_errorFrameSizeInfoLegacy(size_t* cSize, unsigned long long* dBound, size_t ret)
|
||||
{
|
||||
*cSize = ret;
|
||||
*dBound = ZSTD_CONTENTSIZE_ERROR;
|
||||
}
|
||||
|
||||
void ZSTDv01_findFrameSizeInfoLegacy(const void *src, size_t srcSize, size_t* cSize, unsigned long long* dBound)
|
||||
{
|
||||
const BYTE* ip = (const BYTE*)src;
|
||||
size_t remainingSize = srcSize;
|
||||
size_t nbBlocks = 0;
|
||||
U32 magicNumber;
|
||||
blockProperties_t blockProperties;
|
||||
|
||||
/* Frame Header */
|
||||
if (srcSize < ZSTD_frameHeaderSize+ZSTD_blockHeaderSize) {
|
||||
ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, ERROR(srcSize_wrong));
|
||||
return;
|
||||
}
|
||||
magicNumber = ZSTD_readBE32(src);
|
||||
if (magicNumber != ZSTD_magicNumber) {
|
||||
ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, ERROR(prefix_unknown));
|
||||
return;
|
||||
}
|
||||
ip += ZSTD_frameHeaderSize; remainingSize -= ZSTD_frameHeaderSize;
|
||||
|
||||
/* Loop on each block */
|
||||
while (1)
|
||||
{
|
||||
size_t blockSize = ZSTDv01_getcBlockSize(ip, remainingSize, &blockProperties);
|
||||
if (ZSTDv01_isError(blockSize)) {
|
||||
ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, blockSize);
|
||||
return;
|
||||
}
|
||||
|
||||
ip += ZSTD_blockHeaderSize;
|
||||
remainingSize -= ZSTD_blockHeaderSize;
|
||||
if (blockSize > remainingSize) {
|
||||
ZSTD_errorFrameSizeInfoLegacy(cSize, dBound, ERROR(srcSize_wrong));
|
||||
return;
|
||||
}
|
||||
|
||||
if (blockSize == 0) break; /* bt_end */
|
||||
|
||||
ip += blockSize;
|
||||
remainingSize -= blockSize;
|
||||
nbBlocks++;
|
||||
}
|
||||
|
||||
*cSize = ip - (const BYTE*)src;
|
||||
*dBound = nbBlocks * BLOCKSIZE;
|
||||
}
|
||||
|
||||
/*******************************
|
||||
* Streaming Decompression API
|
||||
*******************************/
|
||||
|
||||
size_t ZSTDv01_resetDCtx(ZSTDv01_Dctx* dctx)
|
||||
{
|
||||
dctx->expected = ZSTD_frameHeaderSize;
|
||||
dctx->phase = 0;
|
||||
dctx->previousDstEnd = NULL;
|
||||
dctx->base = NULL;
|
||||
return 0;
|
||||
}
|
||||
|
||||
ZSTDv01_Dctx* ZSTDv01_createDCtx(void)
|
||||
{
|
||||
ZSTDv01_Dctx* dctx = (ZSTDv01_Dctx*)malloc(sizeof(ZSTDv01_Dctx));
|
||||
if (dctx==NULL) return NULL;
|
||||
ZSTDv01_resetDCtx(dctx);
|
||||
return dctx;
|
||||
}
|
||||
|
||||
size_t ZSTDv01_freeDCtx(ZSTDv01_Dctx* dctx)
|
||||
{
|
||||
free(dctx);
|
||||
return 0;
|
||||
}
|
||||
|
||||
size_t ZSTDv01_nextSrcSizeToDecompress(ZSTDv01_Dctx* dctx)
|
||||
{
|
||||
return ((dctx_t*)dctx)->expected;
|
||||
}
|
||||
|
||||
size_t ZSTDv01_decompressContinue(ZSTDv01_Dctx* dctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
|
||||
{
|
||||
dctx_t* ctx = (dctx_t*)dctx;
|
||||
|
||||
/* Sanity check */
|
||||
if (srcSize != ctx->expected) return ERROR(srcSize_wrong);
|
||||
if (dst != ctx->previousDstEnd) /* not contiguous */
|
||||
ctx->base = dst;
|
||||
|
||||
/* Decompress : frame header */
|
||||
if (ctx->phase == 0)
|
||||
{
|
||||
/* Check frame magic header */
|
||||
U32 magicNumber = ZSTD_readBE32(src);
|
||||
if (magicNumber != ZSTD_magicNumber) return ERROR(prefix_unknown);
|
||||
ctx->phase = 1;
|
||||
ctx->expected = ZSTD_blockHeaderSize;
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* Decompress : block header */
|
||||
if (ctx->phase == 1)
|
||||
{
|
||||
blockProperties_t bp;
|
||||
size_t blockSize = ZSTDv01_getcBlockSize(src, ZSTD_blockHeaderSize, &bp);
|
||||
if (ZSTDv01_isError(blockSize)) return blockSize;
|
||||
if (bp.blockType == bt_end)
|
||||
{
|
||||
ctx->expected = 0;
|
||||
ctx->phase = 0;
|
||||
}
|
||||
else
|
||||
{
|
||||
ctx->expected = blockSize;
|
||||
ctx->bType = bp.blockType;
|
||||
ctx->phase = 2;
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* Decompress : block content */
|
||||
{
|
||||
size_t rSize;
|
||||
switch(ctx->bType)
|
||||
{
|
||||
case bt_compressed:
|
||||
rSize = ZSTD_decompressBlock(ctx, dst, maxDstSize, src, srcSize);
|
||||
break;
|
||||
case bt_raw :
|
||||
rSize = ZSTD_copyUncompressedBlock(dst, maxDstSize, src, srcSize);
|
||||
break;
|
||||
case bt_rle :
|
||||
return ERROR(GENERIC); /* not yet handled */
|
||||
break;
|
||||
case bt_end : /* should never happen (filtered at phase 1) */
|
||||
rSize = 0;
|
||||
break;
|
||||
default:
|
||||
return ERROR(GENERIC);
|
||||
}
|
||||
ctx->phase = 1;
|
||||
ctx->expected = ZSTD_blockHeaderSize;
|
||||
if (ZSTDv01_isError(rSize)) return rSize;
|
||||
ctx->previousDstEnd = (void*)( ((char*)dst) + rSize);
|
||||
return rSize;
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
+13
-34
@@ -32,8 +32,7 @@ RUST_CLI_MANIFEST := $(RUST_CLI_DIR)/Cargo.toml
|
||||
RUST_SOURCES := $(RUST_MANIFEST) $(RUST_DIR)/Cargo.lock \
|
||||
$(shell find $(RUST_DIR)/src -type f -name '*.rs' -print)
|
||||
RUST_CLI_SOURCES := $(RUST_CLI_MANIFEST) $(RUST_CLI_DIR)/Cargo.lock \
|
||||
$(RUST_CLI_DIR)/src/lib.rs $(RUST_DIR)/src/zstd_cli.rs \
|
||||
$(RUST_DIR)/src/timefn.rs $(RUST_DIR)/src/benchfn.rs
|
||||
$(RUST_CLI_DIR)/src/lib.rs $(RUST_DIR)/src/zstd_cli.rs
|
||||
|
||||
# Keep Rust's HUF implementation in lockstep with libzstd.mk's C selection.
|
||||
# Forced modes may arrive as libzstd.mk variables or as direct -D flags in
|
||||
@@ -56,33 +55,16 @@ endif
|
||||
endif
|
||||
|
||||
RUST_HUF_FEATURE :=
|
||||
RUST_HUF_MODE := default
|
||||
RUST_BUILD_CONFIG := default
|
||||
ifneq ($(RUST_HUF_FORCE_X1),0)
|
||||
RUST_HUF_FEATURE := huf-force-decompress-x1
|
||||
RUST_HUF_MODE := huf-force-decompress-x1
|
||||
RUST_BUILD_CONFIG := huf-force-decompress-x1
|
||||
endif
|
||||
ifneq ($(RUST_HUF_FORCE_X2),0)
|
||||
RUST_HUF_FEATURE := huf-force-decompress-x2
|
||||
RUST_HUF_MODE := huf-force-decompress-x2
|
||||
RUST_BUILD_CONFIG := huf-force-decompress-x2
|
||||
endif
|
||||
|
||||
# The full-featured zstd program mirrors libzstd.mk's legacy file selection:
|
||||
# ZSTD_LEGACY_SUPPORT=N compiles v0.N .. v0.7, so the Rust archive enables the
|
||||
# matching per-version features. The archive directory encodes the level so
|
||||
# builds for different legacy levels never share cached Rust outputs. The
|
||||
# compress-only, decompress-only, and CLI archives further below are used
|
||||
# solely by ZSTD_LEGACY_SUPPORT=0 program variants and stay legacy-free.
|
||||
empty :=
|
||||
space := $(empty) $(empty)
|
||||
comma := ,
|
||||
RUST_LEGACY_FEATURES :=
|
||||
ifneq ($(ZSTD_LEGACY_SUPPORT), 0)
|
||||
ifeq ($(shell test $(ZSTD_LEGACY_SUPPORT) -lt 8; echo $$?), 0)
|
||||
RUST_LEGACY_FEATURES := $(addprefix legacy-v0,$(wordlist $(ZSTD_LEGACY_SUPPORT),7,1 2 3 4 5 6 7))
|
||||
endif
|
||||
endif
|
||||
RUST_BUILD_CONFIG := $(RUST_HUF_MODE)-legacy$(ZSTD_LEGACY_SUPPORT)
|
||||
|
||||
RUST_TARGET_DIR := $(RUST_DIR)/target/$(RUST_BUILD_CONFIG)
|
||||
RUST_STATICLIB := $(RUST_TARGET_DIR)/release/libzstd_rs.a
|
||||
RUST_TARGET_32 ?= i686-unknown-linux-gnu
|
||||
@@ -93,9 +75,6 @@ RUST_CARGO_FLAGS += --features compression,decompression,dict-builder
|
||||
ifneq ($(RUST_HUF_FEATURE),)
|
||||
RUST_CARGO_FLAGS += --features $(RUST_HUF_FEATURE)
|
||||
endif
|
||||
ifneq ($(RUST_LEGACY_FEATURES),)
|
||||
RUST_CARGO_FLAGS += --features $(subst $(space),$(comma),$(strip $(RUST_LEGACY_FEATURES)))
|
||||
endif
|
||||
|
||||
$(RUST_STATICLIB): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_CARGO_FLAGS)
|
||||
@@ -103,13 +82,13 @@ $(RUST_STATICLIB): $(RUST_SOURCES)
|
||||
$(RUST_STATICLIB_32): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_CARGO_FLAGS) --target $(RUST_TARGET_32)
|
||||
|
||||
RUST_CLI_BUILD_CONFIG := cli-c1-d1-$(RUST_HUF_MODE)
|
||||
RUST_CLI_BUILD_CONFIG := cli-c1-d1-$(RUST_BUILD_CONFIG)
|
||||
RUST_CLI_TARGET_DIR := $(RUST_DIR)/target/$(RUST_CLI_BUILD_CONFIG)
|
||||
RUST_CLI_STATICLIB := $(RUST_CLI_TARGET_DIR)/release/libzstd_cli_rs.a
|
||||
RUST_CLI_STATICLIB_32 := $(RUST_CLI_TARGET_DIR)/$(RUST_TARGET_32)/release/libzstd_cli_rs.a
|
||||
RUST_CLI_CARGO_FLAGS := --manifest-path $(RUST_CLI_MANIFEST) --release \
|
||||
--target-dir $(RUST_CLI_TARGET_DIR) \
|
||||
--no-default-features --features cli,compression,decompression
|
||||
--no-default-features --features compression,decompression
|
||||
|
||||
$(RUST_CLI_STATICLIB): $(RUST_CLI_SOURCES)
|
||||
$(CARGO) build $(RUST_CLI_CARGO_FLAGS)
|
||||
@@ -117,7 +96,7 @@ $(RUST_CLI_STATICLIB): $(RUST_CLI_SOURCES)
|
||||
$(RUST_CLI_STATICLIB_32): $(RUST_CLI_SOURCES)
|
||||
$(CARGO) build $(RUST_CLI_CARGO_FLAGS) --target $(RUST_TARGET_32)
|
||||
|
||||
RUST_DECOMPRESS_BUILD_CONFIG := lib-c0-d1-b0-$(RUST_HUF_MODE)
|
||||
RUST_DECOMPRESS_BUILD_CONFIG := lib-c0-d1-b0-$(RUST_BUILD_CONFIG)
|
||||
RUST_DECOMPRESS_TARGET_DIR := $(RUST_DIR)/target/$(RUST_DECOMPRESS_BUILD_CONFIG)
|
||||
RUST_DECOMPRESS_STATICLIB := $(RUST_DECOMPRESS_TARGET_DIR)/release/libzstd_rs.a
|
||||
RUST_DECOMPRESS_CARGO_FLAGS := --manifest-path $(RUST_MANIFEST) --release \
|
||||
@@ -130,17 +109,17 @@ endif
|
||||
$(RUST_DECOMPRESS_STATICLIB): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_DECOMPRESS_CARGO_FLAGS)
|
||||
|
||||
RUST_DECOMPRESS_CLI_BUILD_CONFIG := cli-c0-d1-$(RUST_HUF_MODE)
|
||||
RUST_DECOMPRESS_CLI_BUILD_CONFIG := cli-c0-d1-$(RUST_BUILD_CONFIG)
|
||||
RUST_DECOMPRESS_CLI_TARGET_DIR := $(RUST_DIR)/target/$(RUST_DECOMPRESS_CLI_BUILD_CONFIG)
|
||||
RUST_DECOMPRESS_CLI_STATICLIB := $(RUST_DECOMPRESS_CLI_TARGET_DIR)/release/libzstd_cli_rs.a
|
||||
RUST_DECOMPRESS_CLI_CARGO_FLAGS := --manifest-path $(RUST_CLI_MANIFEST) --release \
|
||||
--target-dir $(RUST_DECOMPRESS_CLI_TARGET_DIR) \
|
||||
--no-default-features --features cli,decompression
|
||||
--no-default-features --features decompression
|
||||
|
||||
$(RUST_DECOMPRESS_CLI_STATICLIB): $(RUST_CLI_SOURCES)
|
||||
$(CARGO) build $(RUST_DECOMPRESS_CLI_CARGO_FLAGS)
|
||||
|
||||
RUST_COMPRESS_BUILD_CONFIG := lib-c1-d0-b0-$(RUST_HUF_MODE)
|
||||
RUST_COMPRESS_BUILD_CONFIG := lib-c1-d0-b0-$(RUST_BUILD_CONFIG)
|
||||
RUST_COMPRESS_TARGET_DIR := $(RUST_DIR)/target/$(RUST_COMPRESS_BUILD_CONFIG)
|
||||
RUST_COMPRESS_STATICLIB := $(RUST_COMPRESS_TARGET_DIR)/release/libzstd_rs.a
|
||||
RUST_COMPRESS_CARGO_FLAGS := --manifest-path $(RUST_MANIFEST) --release \
|
||||
@@ -150,7 +129,7 @@ RUST_COMPRESS_CARGO_FLAGS := --manifest-path $(RUST_MANIFEST) --release \
|
||||
$(RUST_COMPRESS_STATICLIB): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_COMPRESS_CARGO_FLAGS)
|
||||
|
||||
RUST_DICTBUILDER_BUILD_CONFIG := lib-c1-d0-b1-$(RUST_HUF_MODE)
|
||||
RUST_DICTBUILDER_BUILD_CONFIG := lib-c1-d0-b1-$(RUST_BUILD_CONFIG)
|
||||
RUST_DICTBUILDER_TARGET_DIR := $(RUST_DIR)/target/$(RUST_DICTBUILDER_BUILD_CONFIG)
|
||||
RUST_DICTBUILDER_STATICLIB := $(RUST_DICTBUILDER_TARGET_DIR)/release/libzstd_rs.a
|
||||
RUST_DICTBUILDER_CARGO_FLAGS := --manifest-path $(RUST_MANIFEST) --release \
|
||||
@@ -161,12 +140,12 @@ RUST_DICTBUILDER_CARGO_FLAGS := --manifest-path $(RUST_MANIFEST) --release \
|
||||
$(RUST_DICTBUILDER_STATICLIB): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_DICTBUILDER_CARGO_FLAGS)
|
||||
|
||||
RUST_COMPRESS_CLI_BUILD_CONFIG := cli-c1-d0-$(RUST_HUF_MODE)
|
||||
RUST_COMPRESS_CLI_BUILD_CONFIG := cli-c1-d0-$(RUST_BUILD_CONFIG)
|
||||
RUST_COMPRESS_CLI_TARGET_DIR := $(RUST_DIR)/target/$(RUST_COMPRESS_CLI_BUILD_CONFIG)
|
||||
RUST_COMPRESS_CLI_STATICLIB := $(RUST_COMPRESS_CLI_TARGET_DIR)/release/libzstd_cli_rs.a
|
||||
RUST_COMPRESS_CLI_CARGO_FLAGS := --manifest-path $(RUST_CLI_MANIFEST) --release \
|
||||
--target-dir $(RUST_COMPRESS_CLI_TARGET_DIR) \
|
||||
--no-default-features --features cli,compression
|
||||
--no-default-features --features compression
|
||||
|
||||
$(RUST_COMPRESS_CLI_STATICLIB): $(RUST_CLI_SOURCES)
|
||||
$(CARGO) build $(RUST_COMPRESS_CLI_CARGO_FLAGS)
|
||||
|
||||
+243
-17
@@ -8,23 +8,249 @@
|
||||
* You may select, at your option, one of the above-listed licenses.
|
||||
*/
|
||||
|
||||
/* The implementation lives in rust/src/benchfn.rs, built into the Rust CLI
|
||||
* static archive. This translation unit stays in the original source lists so
|
||||
* build configuration keeps working while the implementation is in Rust. */
|
||||
|
||||
#include <stddef.h> /* size_t, offsetof */
|
||||
|
||||
/* *************************************
|
||||
* Includes
|
||||
***************************************/
|
||||
#include <stdlib.h> /* malloc, free */
|
||||
#include <string.h> /* memset */
|
||||
#include <assert.h> /* assert */
|
||||
|
||||
#include "timefn.h" /* UTIL_time_t, UTIL_getTime */
|
||||
#include "benchfn.h"
|
||||
|
||||
/* BMK_runTime_t and BMK_runOutcome_t are returned by value across the C/Rust
|
||||
* boundary, and BMK_benchParams_t is passed by value. The Rust #[repr(C)]
|
||||
* definitions mirror the offsets pinned here. */
|
||||
typedef char BMK_staticAssert_runTimeSumOffset[
|
||||
(offsetof(BMK_runTime_t, sumOfReturn) == sizeof(double)) ? 1 : -1];
|
||||
typedef char BMK_staticAssert_outcomeResultOffset[
|
||||
(offsetof(BMK_runOutcome_t, error_result_never_ever_use_directly)
|
||||
== sizeof(BMK_runTime_t)) ? 1 : -1];
|
||||
typedef char BMK_staticAssert_outcomeTagOffset[
|
||||
(offsetof(BMK_runOutcome_t, error_tag_never_ever_use_directly)
|
||||
== sizeof(BMK_runTime_t) + sizeof(size_t)) ? 1 : -1];
|
||||
typedef char BMK_staticAssert_shellAlignment[
|
||||
(sizeof(BMK_timedFnState_shell) == BMK_TIMEDFNSTATE_SIZE) ? 1 : -1];
|
||||
|
||||
/* *************************************
|
||||
* Constants
|
||||
***************************************/
|
||||
#define TIMELOOP_MICROSEC SEC_TO_MICRO /* 1 second */
|
||||
#define TIMELOOP_NANOSEC (1*1000000000ULL) /* 1 second */
|
||||
|
||||
#define KB *(1 <<10)
|
||||
#define MB *(1 <<20)
|
||||
#define GB *(1U<<30)
|
||||
|
||||
|
||||
/* *************************************
|
||||
* Debug errors
|
||||
***************************************/
|
||||
#if defined(DEBUG) && (DEBUG >= 1)
|
||||
# include <stdio.h> /* fprintf */
|
||||
# define DISPLAY(...) fprintf(stderr, __VA_ARGS__)
|
||||
# define DEBUGOUTPUT(...) { if (DEBUG) DISPLAY(__VA_ARGS__); }
|
||||
#else
|
||||
# define DEBUGOUTPUT(...)
|
||||
#endif
|
||||
|
||||
|
||||
/* error without displaying */
|
||||
#define RETURN_QUIET_ERROR(retValue, ...) { \
|
||||
DEBUGOUTPUT("%s: %i: \n", __FILE__, __LINE__); \
|
||||
DEBUGOUTPUT("Error : "); \
|
||||
DEBUGOUTPUT(__VA_ARGS__); \
|
||||
DEBUGOUTPUT(" \n"); \
|
||||
return retValue; \
|
||||
}
|
||||
|
||||
/* Abort execution if a condition is not met */
|
||||
#define CONTROL(c) { if (!(c)) { DEBUGOUTPUT("error: %s \n", #c); abort(); } }
|
||||
|
||||
|
||||
/* *************************************
|
||||
* Benchmarking an arbitrary function
|
||||
***************************************/
|
||||
|
||||
int BMK_isSuccessful_runOutcome(BMK_runOutcome_t outcome)
|
||||
{
|
||||
return outcome.error_tag_never_ever_use_directly == 0;
|
||||
}
|
||||
|
||||
/* warning : this function will stop program execution if outcome is invalid !
|
||||
* check outcome validity first, using BMK_isValid_runResult() */
|
||||
BMK_runTime_t BMK_extract_runTime(BMK_runOutcome_t outcome)
|
||||
{
|
||||
CONTROL(outcome.error_tag_never_ever_use_directly == 0);
|
||||
return outcome.internal_never_ever_use_directly;
|
||||
}
|
||||
|
||||
size_t BMK_extract_errorResult(BMK_runOutcome_t outcome)
|
||||
{
|
||||
CONTROL(outcome.error_tag_never_ever_use_directly != 0);
|
||||
return outcome.error_result_never_ever_use_directly;
|
||||
}
|
||||
|
||||
static BMK_runOutcome_t BMK_runOutcome_error(size_t errorResult)
|
||||
{
|
||||
BMK_runOutcome_t b;
|
||||
memset(&b, 0, sizeof(b));
|
||||
b.error_tag_never_ever_use_directly = 1;
|
||||
b.error_result_never_ever_use_directly = errorResult;
|
||||
return b;
|
||||
}
|
||||
|
||||
static BMK_runOutcome_t BMK_setValid_runTime(BMK_runTime_t runTime)
|
||||
{
|
||||
BMK_runOutcome_t outcome;
|
||||
outcome.error_tag_never_ever_use_directly = 0;
|
||||
outcome.internal_never_ever_use_directly = runTime;
|
||||
return outcome;
|
||||
}
|
||||
|
||||
|
||||
/* initFn will be measured once, benchFn will be measured `nbLoops` times */
|
||||
/* initFn is optional, provide NULL if none */
|
||||
/* benchFn must return a size_t value that errorFn can interpret */
|
||||
/* takes # of blocks and list of size & stuff for each. */
|
||||
/* can report result of benchFn for each block into blockResult. */
|
||||
/* blockResult is optional, provide NULL if this information is not required */
|
||||
/* note : time per loop can be reported as zero if run time < timer resolution */
|
||||
BMK_runOutcome_t BMK_benchFunction(BMK_benchParams_t p,
|
||||
unsigned nbLoops)
|
||||
{
|
||||
nbLoops += !nbLoops; /* minimum nbLoops is 1 */
|
||||
|
||||
/* init */
|
||||
{ size_t i;
|
||||
for(i = 0; i < p.blockCount; i++) {
|
||||
memset(p.dstBuffers[i], 0xE5, p.dstCapacities[i]); /* warm up and erase result buffer */
|
||||
} }
|
||||
|
||||
/* benchmark */
|
||||
{ size_t dstSize = 0;
|
||||
UTIL_time_t const clockStart = UTIL_getTime();
|
||||
unsigned loopNb, blockNb;
|
||||
if (p.initFn != NULL) p.initFn(p.initPayload);
|
||||
for (loopNb = 0; loopNb < nbLoops; loopNb++) {
|
||||
for (blockNb = 0; blockNb < p.blockCount; blockNb++) {
|
||||
size_t const res = p.benchFn(p.srcBuffers[blockNb], p.srcSizes[blockNb],
|
||||
p.dstBuffers[blockNb], p.dstCapacities[blockNb],
|
||||
p.benchPayload);
|
||||
if (loopNb == 0) {
|
||||
if (p.blockResults != NULL) p.blockResults[blockNb] = res;
|
||||
if ((p.errorFn != NULL) && (p.errorFn(res))) {
|
||||
RETURN_QUIET_ERROR(BMK_runOutcome_error(res),
|
||||
"Function benchmark failed on block %u (of size %u) with error %i",
|
||||
blockNb, (unsigned)p.srcSizes[blockNb], (int)res);
|
||||
}
|
||||
dstSize += res;
|
||||
} }
|
||||
} /* for (loopNb = 0; loopNb < nbLoops; loopNb++) */
|
||||
|
||||
{ PTime const totalTime = UTIL_clockSpanNano(clockStart);
|
||||
BMK_runTime_t rt;
|
||||
rt.nanoSecPerRun = (double)totalTime / nbLoops;
|
||||
rt.sumOfReturn = dstSize;
|
||||
return BMK_setValid_runTime(rt);
|
||||
} }
|
||||
}
|
||||
|
||||
|
||||
/* ==== Benchmarking any function, providing intermediate results ==== */
|
||||
|
||||
struct BMK_timedFnState_s {
|
||||
PTime timeSpent_ns;
|
||||
PTime timeBudget_ns;
|
||||
PTime runBudget_ns;
|
||||
BMK_runTime_t fastestRun;
|
||||
unsigned nbLoops;
|
||||
UTIL_time_t coolTime;
|
||||
}; /* typedef'd to BMK_timedFnState_t within bench.h */
|
||||
|
||||
BMK_timedFnState_t* BMK_createTimedFnState(unsigned total_ms, unsigned run_ms)
|
||||
{
|
||||
BMK_timedFnState_t* const r = (BMK_timedFnState_t*)malloc(sizeof(*r));
|
||||
if (r == NULL) return NULL; /* malloc() error */
|
||||
BMK_resetTimedFnState(r, total_ms, run_ms);
|
||||
return r;
|
||||
}
|
||||
|
||||
void BMK_freeTimedFnState(BMK_timedFnState_t* state) { free(state); }
|
||||
|
||||
BMK_timedFnState_t*
|
||||
BMK_initStatic_timedFnState(void* buffer, size_t size, unsigned total_ms, unsigned run_ms)
|
||||
{
|
||||
typedef char check_size[ 2 * (sizeof(BMK_timedFnState_shell) >= sizeof(struct BMK_timedFnState_s)) - 1]; /* static assert : a compilation failure indicates that BMK_timedFnState_shell is not large enough */
|
||||
typedef struct { check_size c; BMK_timedFnState_t tfs; } tfs_align; /* force tfs to be aligned at its next best position */
|
||||
size_t const tfs_alignment = offsetof(tfs_align, tfs); /* provides the minimal alignment restriction for BMK_timedFnState_t */
|
||||
BMK_timedFnState_t* const r = (BMK_timedFnState_t*)buffer;
|
||||
if (buffer == NULL) return NULL;
|
||||
if (size < sizeof(struct BMK_timedFnState_s)) return NULL;
|
||||
if ((size_t)buffer % tfs_alignment) return NULL; /* buffer must be properly aligned */
|
||||
BMK_resetTimedFnState(r, total_ms, run_ms);
|
||||
return r;
|
||||
}
|
||||
|
||||
void BMK_resetTimedFnState(BMK_timedFnState_t* timedFnState, unsigned total_ms, unsigned run_ms)
|
||||
{
|
||||
if (!total_ms) total_ms = 1 ;
|
||||
if (!run_ms) run_ms = 1;
|
||||
if (run_ms > total_ms) run_ms = total_ms;
|
||||
timedFnState->timeSpent_ns = 0;
|
||||
timedFnState->timeBudget_ns = (PTime)total_ms * TIMELOOP_NANOSEC / 1000;
|
||||
timedFnState->runBudget_ns = (PTime)run_ms * TIMELOOP_NANOSEC / 1000;
|
||||
timedFnState->fastestRun.nanoSecPerRun = (double)TIMELOOP_NANOSEC * 2000000000; /* hopefully large enough : must be larger than any potential measurement */
|
||||
timedFnState->fastestRun.sumOfReturn = (size_t)(-1LL);
|
||||
timedFnState->nbLoops = 1;
|
||||
timedFnState->coolTime = UTIL_getTime();
|
||||
}
|
||||
|
||||
/* Tells if nb of seconds set in timedFnState for all runs is spent.
|
||||
* note : this function will return 1 if BMK_benchFunctionTimed() has actually errored. */
|
||||
int BMK_isCompleted_TimedFn(const BMK_timedFnState_t* timedFnState)
|
||||
{
|
||||
return (timedFnState->timeSpent_ns >= timedFnState->timeBudget_ns);
|
||||
}
|
||||
|
||||
|
||||
#undef MIN
|
||||
#define MIN(a,b) ( (a) < (b) ? (a) : (b) )
|
||||
|
||||
#define MINUSABLETIME (TIMELOOP_NANOSEC / 2) /* 0.5 seconds */
|
||||
|
||||
BMK_runOutcome_t BMK_benchTimedFn(BMK_timedFnState_t* cont,
|
||||
BMK_benchParams_t p)
|
||||
{
|
||||
PTime const runBudget_ns = cont->runBudget_ns;
|
||||
PTime const runTimeMin_ns = runBudget_ns / 2;
|
||||
int completed = 0;
|
||||
BMK_runTime_t bestRunTime = cont->fastestRun;
|
||||
|
||||
while (!completed) {
|
||||
BMK_runOutcome_t const runResult = BMK_benchFunction(p, cont->nbLoops);
|
||||
|
||||
if(!BMK_isSuccessful_runOutcome(runResult)) { /* error : move out */
|
||||
return runResult;
|
||||
}
|
||||
|
||||
{ BMK_runTime_t const newRunTime = BMK_extract_runTime(runResult);
|
||||
double const loopDuration_ns = newRunTime.nanoSecPerRun * cont->nbLoops;
|
||||
|
||||
cont->timeSpent_ns += (unsigned long long)loopDuration_ns;
|
||||
|
||||
/* estimate nbLoops for next run to last approximately 1 second */
|
||||
if (loopDuration_ns > ((double)runBudget_ns / 50)) {
|
||||
double const fastestRun_ns = MIN(bestRunTime.nanoSecPerRun, newRunTime.nanoSecPerRun);
|
||||
cont->nbLoops = (unsigned)((double)runBudget_ns / fastestRun_ns) + 1;
|
||||
} else {
|
||||
/* previous run was too short : blindly increase workload by x multiplier */
|
||||
const unsigned multiplier = 10;
|
||||
assert(cont->nbLoops < ((unsigned)-1) / multiplier); /* avoid overflow */
|
||||
cont->nbLoops *= multiplier;
|
||||
}
|
||||
|
||||
if(loopDuration_ns < (double)runTimeMin_ns) {
|
||||
/* don't report results for which benchmark run time was too small : increased risks of rounding errors */
|
||||
assert(completed == 0);
|
||||
continue;
|
||||
} else {
|
||||
if(newRunTime.nanoSecPerRun < bestRunTime.nanoSecPerRun) {
|
||||
bestRunTime = newRunTime;
|
||||
}
|
||||
completed = 1;
|
||||
}
|
||||
}
|
||||
} /* while (!completed) */
|
||||
|
||||
return BMK_setValid_runTime(bestRunTime);
|
||||
}
|
||||
|
||||
+155
-8
@@ -8,14 +8,161 @@
|
||||
* You may select, at your option, one of the above-listed licenses.
|
||||
*/
|
||||
|
||||
/* The implementation lives in rust/src/timefn.rs, built into the Rust CLI
|
||||
* static archive. This translation unit stays in the original source lists so
|
||||
* build configuration keeps working while the implementation is in Rust. */
|
||||
|
||||
/* === Dependencies === */
|
||||
|
||||
#include "timefn.h"
|
||||
#include "platform.h" /* set _POSIX_C_SOURCE */
|
||||
#include <time.h> /* CLOCK_MONOTONIC, TIME_UTC */
|
||||
|
||||
/* The Rust port mirrors this exact ABI: UTIL_time_t is returned by value and
|
||||
* must remain a plain 64-bit nanosecond counter. */
|
||||
typedef char UTIL_staticAssert_ptimeIs64Bit[(sizeof(PTime) == 8) ? 1 : -1];
|
||||
typedef char UTIL_staticAssert_timeIsPlainCounter[
|
||||
(sizeof(UTIL_time_t) == sizeof(PTime)) ? 1 : -1];
|
||||
/*-****************************************
|
||||
* Time functions
|
||||
******************************************/
|
||||
|
||||
#if defined(_WIN32) /* Windows */
|
||||
|
||||
#include <windows.h> /* LARGE_INTEGER */
|
||||
#include <stdlib.h> /* abort */
|
||||
#include <stdio.h> /* perror */
|
||||
|
||||
UTIL_time_t UTIL_getTime(void)
|
||||
{
|
||||
static LARGE_INTEGER ticksPerSecond;
|
||||
static int init = 0;
|
||||
if (!init) {
|
||||
if (!QueryPerformanceFrequency(&ticksPerSecond)) {
|
||||
perror("timefn::QueryPerformanceFrequency");
|
||||
abort();
|
||||
}
|
||||
init = 1;
|
||||
}
|
||||
{ UTIL_time_t r;
|
||||
LARGE_INTEGER x;
|
||||
QueryPerformanceCounter(&x);
|
||||
r.t = (PTime)(x.QuadPart * 1000000000ULL / ticksPerSecond.QuadPart);
|
||||
return r;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
#elif defined(__APPLE__) && defined(__MACH__)
|
||||
|
||||
#include <mach/mach_time.h> /* mach_timebase_info_data_t, mach_timebase_info, mach_absolute_time */
|
||||
|
||||
UTIL_time_t UTIL_getTime(void)
|
||||
{
|
||||
static mach_timebase_info_data_t rate;
|
||||
static int init = 0;
|
||||
if (!init) {
|
||||
mach_timebase_info(&rate);
|
||||
init = 1;
|
||||
}
|
||||
{ UTIL_time_t r;
|
||||
r.t = mach_absolute_time() * (PTime)rate.numer / (PTime)rate.denom;
|
||||
return r;
|
||||
}
|
||||
}
|
||||
|
||||
/* POSIX.1-2001 (optional) */
|
||||
#elif defined(CLOCK_MONOTONIC)
|
||||
|
||||
#include <stdlib.h> /* abort */
|
||||
#include <stdio.h> /* perror */
|
||||
|
||||
UTIL_time_t UTIL_getTime(void)
|
||||
{
|
||||
/* time must be initialized, othersize it may fail msan test.
|
||||
* No good reason, likely a limitation of timespec_get() for some target */
|
||||
struct timespec time = { 0, 0 };
|
||||
if (clock_gettime(CLOCK_MONOTONIC, &time) != 0) {
|
||||
perror("timefn::clock_gettime(CLOCK_MONOTONIC)");
|
||||
abort();
|
||||
}
|
||||
{ UTIL_time_t r;
|
||||
r.t = (PTime)time.tv_sec * 1000000000ULL + (PTime)time.tv_nsec;
|
||||
return r;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
/* C11 requires support of timespec_get().
|
||||
* However, FreeBSD 11 claims C11 compliance while lacking timespec_get().
|
||||
* Double confirm timespec_get() support by checking the definition of TIME_UTC.
|
||||
* However, some versions of Android manage to simultaneously define TIME_UTC
|
||||
* and lack timespec_get() support... */
|
||||
#elif (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 201112L) /* C11 */) \
|
||||
&& defined(TIME_UTC) && !defined(__ANDROID__)
|
||||
|
||||
#include <stdlib.h> /* abort */
|
||||
#include <stdio.h> /* perror */
|
||||
|
||||
UTIL_time_t UTIL_getTime(void)
|
||||
{
|
||||
/* time must be initialized, othersize it may fail msan test.
|
||||
* No good reason, likely a limitation of timespec_get() for some target */
|
||||
struct timespec time = { 0, 0 };
|
||||
if (timespec_get(&time, TIME_UTC) != TIME_UTC) {
|
||||
perror("timefn::timespec_get(TIME_UTC)");
|
||||
abort();
|
||||
}
|
||||
{ UTIL_time_t r;
|
||||
r.t = (PTime)time.tv_sec * 1000000000ULL + (PTime)time.tv_nsec;
|
||||
return r;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
#else /* relies on standard C90 (note : clock_t produces wrong measurements for multi-threaded workloads) */
|
||||
|
||||
UTIL_time_t UTIL_getTime(void)
|
||||
{
|
||||
UTIL_time_t r;
|
||||
r.t = (PTime)clock() * 1000000000ULL / CLOCKS_PER_SEC;
|
||||
return r;
|
||||
}
|
||||
|
||||
#define TIME_MT_MEASUREMENTS_NOT_SUPPORTED
|
||||
|
||||
#endif
|
||||
|
||||
/* ==== Common functions, valid for all time API ==== */
|
||||
|
||||
PTime UTIL_getSpanTimeNano(UTIL_time_t clockStart, UTIL_time_t clockEnd)
|
||||
{
|
||||
return clockEnd.t - clockStart.t;
|
||||
}
|
||||
|
||||
PTime UTIL_getSpanTimeMicro(UTIL_time_t begin, UTIL_time_t end)
|
||||
{
|
||||
return UTIL_getSpanTimeNano(begin, end) / 1000ULL;
|
||||
}
|
||||
|
||||
PTime UTIL_clockSpanMicro(UTIL_time_t clockStart )
|
||||
{
|
||||
UTIL_time_t const clockEnd = UTIL_getTime();
|
||||
return UTIL_getSpanTimeMicro(clockStart, clockEnd);
|
||||
}
|
||||
|
||||
PTime UTIL_clockSpanNano(UTIL_time_t clockStart )
|
||||
{
|
||||
UTIL_time_t const clockEnd = UTIL_getTime();
|
||||
return UTIL_getSpanTimeNano(clockStart, clockEnd);
|
||||
}
|
||||
|
||||
void UTIL_waitForNextTick(void)
|
||||
{
|
||||
UTIL_time_t const clockStart = UTIL_getTime();
|
||||
UTIL_time_t clockEnd;
|
||||
do {
|
||||
clockEnd = UTIL_getTime();
|
||||
} while (UTIL_getSpanTimeNano(clockStart, clockEnd) == 0);
|
||||
}
|
||||
|
||||
int UTIL_support_MT_measurements(void)
|
||||
{
|
||||
# if defined(TIME_MT_MEASUREMENTS_NOT_SUPPORTED)
|
||||
return 0;
|
||||
# else
|
||||
return 1;
|
||||
# endif
|
||||
}
|
||||
|
||||
@@ -10,67 +10,16 @@
|
||||
|
||||
/* The CLI parser and control flow live in rust/src/zstd_cli.rs. Keep this
|
||||
* translation unit as the stable C entry point used by program launchers. */
|
||||
#include <stddef.h> /* size_t */
|
||||
#define ZSTD_STATIC_LINKING_ONLY /* ZSTD_compressionParameters */
|
||||
#include "../lib/zstd.h"
|
||||
#ifndef ZSTD_NOBENCH
|
||||
# include "benchzstd.h" /* BMK_benchFilesAdvanced, BMK_syntheticTest */
|
||||
#endif
|
||||
|
||||
int ZSTD_rust_cli_main(int argCount, const char* const argv[]);
|
||||
const char* ZSTD_rust_cli_expected_version(void);
|
||||
int ZSTD_rust_cli_bench(const char* const* fileNames, unsigned nbFiles,
|
||||
const char* dictFileName,
|
||||
int startCLevel, int endCLevel,
|
||||
const ZSTD_compressionParameters* compressionParams,
|
||||
int displayLevel, unsigned nbSeconds,
|
||||
size_t blockSize, int nbWorkers);
|
||||
|
||||
const char* ZSTD_rust_cli_expected_version(void)
|
||||
{
|
||||
return ZSTD_VERSION_STRING;
|
||||
}
|
||||
|
||||
/* Benchmark bridge for the Rust CLI. Whether benchmarking exists is a C
|
||||
* preprocessor property (ZSTD_NOBENCH), so the decision stays in this shim:
|
||||
* the Rust frontend calls in unconditionally, and stripped program variants
|
||||
* never reference benchmark symbols.
|
||||
* @return the benchmark result code (>= 0), or -1 when unavailable. */
|
||||
int ZSTD_rust_cli_bench(const char* const* fileNames, unsigned nbFiles,
|
||||
const char* dictFileName,
|
||||
int startCLevel, int endCLevel,
|
||||
const ZSTD_compressionParameters* compressionParams,
|
||||
int displayLevel, unsigned nbSeconds,
|
||||
size_t blockSize, int nbWorkers)
|
||||
{
|
||||
#ifndef ZSTD_NOBENCH
|
||||
BMK_advancedParams_t advancedParams = BMK_initAdvancedParams();
|
||||
int startLevel = startCLevel;
|
||||
int endLevel = endCLevel;
|
||||
advancedParams.nbSeconds = nbSeconds;
|
||||
advancedParams.blockSize = blockSize;
|
||||
advancedParams.nbWorkers = nbWorkers;
|
||||
if (startLevel > ZSTD_maxCLevel()) startLevel = ZSTD_maxCLevel();
|
||||
if (endLevel > ZSTD_maxCLevel()) endLevel = ZSTD_maxCLevel();
|
||||
if (endLevel < startLevel) endLevel = startLevel;
|
||||
if (nbFiles == 0) {
|
||||
/* No input file: benchmark a synthetic sample (lorem generator). */
|
||||
return BMK_syntheticTest(-1.0, startLevel, endLevel,
|
||||
compressionParams, displayLevel,
|
||||
&advancedParams);
|
||||
}
|
||||
return BMK_benchFilesAdvanced(fileNames, nbFiles, dictFileName,
|
||||
startLevel, endLevel,
|
||||
compressionParams, displayLevel,
|
||||
&advancedParams);
|
||||
#else
|
||||
(void)fileNames; (void)nbFiles; (void)dictFileName;
|
||||
(void)startCLevel; (void)endCLevel; (void)compressionParams;
|
||||
(void)displayLevel; (void)nbSeconds; (void)blockSize; (void)nbWorkers;
|
||||
return -1;
|
||||
#endif
|
||||
}
|
||||
|
||||
int main(int argCount, const char* argv[])
|
||||
{
|
||||
return ZSTD_rust_cli_main(argCount, argv);
|
||||
|
||||
@@ -13,18 +13,6 @@ decompression = []
|
||||
dict-builder = []
|
||||
huf-force-decompress-x1 = []
|
||||
huf-force-decompress-x2 = []
|
||||
# Legacy-format decoders (zstd v0.1 .. v0.7). Never default features: the
|
||||
# build systems map ZSTD_LEGACY_SUPPORT=N to the features for versions >= N,
|
||||
# exactly mirroring which lib/legacy/zstd_v0N.c files the C build compiles.
|
||||
# A feature whose version is not yet ported to Rust gates nothing; the
|
||||
# original C file still provides that decoder.
|
||||
legacy-v01 = []
|
||||
legacy-v02 = []
|
||||
legacy-v03 = []
|
||||
legacy-v04 = []
|
||||
legacy-v05 = []
|
||||
legacy-v06 = []
|
||||
legacy-v07 = []
|
||||
|
||||
[dependencies]
|
||||
libc = "0.2"
|
||||
|
||||
+6
-63
@@ -34,12 +34,6 @@ zstd ABI:
|
||||
- `zstd_compress_frame` serializes frame headers, skippable frames, and the
|
||||
last empty block; it takes scalar frame parameters so the C-owned
|
||||
`ZSTD_CCtx_params` layout never crosses the language boundary.
|
||||
- `zstd_compress_params` owns the compression-level tables (formerly
|
||||
`clevels.h`), parameter bounds, clamping, validation, table selection,
|
||||
source/dictionary adjustment, and match-state/CDict size estimation.
|
||||
The C integration layer keeps the public `ZSTD_*` symbols and feeds the
|
||||
leaves configuration-owned scalars: the excluded-block-compressor
|
||||
strategy cascade, struct sizes, and sanitizer redzone policy.
|
||||
- `zstd_fast` and `zstd_double_fast` implement the single- and two-table
|
||||
fast block match finders, including attached and external dictionary paths.
|
||||
- `zstd_lazy` implements greedy, lazy, lazy2, and binary-tree matching,
|
||||
@@ -57,10 +51,6 @@ zstd ABI:
|
||||
- `pool` implements the bounded worker pool used by multithreaded compression.
|
||||
- Dictionary support
|
||||
- `zstd_ddict` owns, loads, copies, and references decode dictionaries.
|
||||
- Legacy decoding
|
||||
- `legacy` hosts one frozen module per historical format; `legacy::zstd_v01`
|
||||
ports the self-contained v0.1 decoder. Versions v0.2 through v0.7 are
|
||||
still C.
|
||||
- Block decompression
|
||||
- `zstd_decompress_block` decodes literal and sequence sections, maintains
|
||||
FSE/Huffman repeat state, and executes compressed-block sequences.
|
||||
@@ -73,58 +63,12 @@ zstd ABI:
|
||||
so library builds do not acquire program-only dependencies. The C
|
||||
`fileio` backend still owns file opening, safe replacement, sparse writes,
|
||||
metadata, and streaming I/O.
|
||||
- `timefn` provides the monotonic nanosecond clock behind `UTIL_time_t`,
|
||||
and `benchfn` owns the benchmark run/timing loop (`BMK_benchFunction`,
|
||||
`BMK_benchTimedFn`) used by the CLI benchmark mode and by C test tools.
|
||||
Both live in the `cli/` package, but C test binaries (fullbench, fuzzer,
|
||||
zstreamtest, paramgrill, ...) link a helpers-only build of that archive,
|
||||
produced without the package's `cli` feature, because the parser layer
|
||||
requires the C `fileio` backend that tests do not compile. Benchmark
|
||||
orchestration and reporting (`benchzstd.c`) remain C, reached from the
|
||||
Rust parser through the `ZSTD_NOBENCH`-gated bridge in `zstdcli.c`.
|
||||
|
||||
The optimal block matcher, high-level frame compression, dictionary-building
|
||||
except suffix-array construction, the legacy v0.2-v0.7 decoders, benchmark
|
||||
orchestration (`benchzstd`), and the CLI file-I/O backend are still C. They
|
||||
must move before the rewrite is complete. Keeping that boundary explicit
|
||||
prevents a passing hybrid build from being mistaken for the final all-Rust
|
||||
result.
|
||||
|
||||
## Legacy decoding
|
||||
|
||||
Each `lib/legacy/zstd_v0N.c` file is a frozen snapshot of the entropy coders
|
||||
and frame logic of one historical release. The Rust ports in `src/legacy/`
|
||||
keep that property: every version owns its own frozen FSE/Huff0 and frame
|
||||
logic, ported line by line, and must never reuse the modern entropy modules
|
||||
or share code with other legacy versions. Outputs and error codes must be
|
||||
byte-identical to the original C files. Their only shared dependency is the
|
||||
`errors` module, matching the C files' `error_private.h` include.
|
||||
|
||||
Cargo features `legacy-v01` .. `legacy-v07` gate the per-version modules and
|
||||
are never default features. The build systems derive the feature list from
|
||||
the C configuration:
|
||||
|
||||
- `lib/Makefile` and `programs/Makefile` map `ZSTD_LEGACY_SUPPORT=N` to the
|
||||
features for versions >= N (0 disables legacy), matching the
|
||||
`ZSTD_LEGACY_FILES` selection in `lib/libzstd.mk`.
|
||||
- `tests/Makefile` always enables all seven features because the test
|
||||
objects compile every `lib/legacy/*.c` file regardless of dispatch level.
|
||||
- `build/meson` maps `legacy_level` like the makefiles; `build/cmake`
|
||||
enables all seven whenever `ZSTD_LEGACY_SUPPORT` is on because it always
|
||||
compiles all seven C files.
|
||||
|
||||
Every build system also encodes the legacy selection in the Rust target
|
||||
directory name (for example `c1-d1-default-legacy5`), for the same reason the
|
||||
HUF mode is encoded there: a cached archive built for one configuration must
|
||||
never be linked into a build that expects another.
|
||||
|
||||
A feature whose version has not been ported yet gates nothing; the original
|
||||
C file still provides that decoder, so mixed C/Rust legacy levels link
|
||||
cleanly. Porting a version means adding `src/legacy/zstd_v0N.rs`, registering
|
||||
it in `src/legacy/mod.rs` behind its feature, and reducing
|
||||
`lib/legacy/zstd_v0N.c` to a declaration-only shim. For v0.1 the streaming
|
||||
`ZSTDv01_Dctx` state lives entirely in Rust: C code only ever holds an opaque
|
||||
pointer, so the C-side struct definition is gone.
|
||||
except suffix-array construction, legacy decoding callbacks, and the CLI
|
||||
file-I/O backend are still C. They must move before the rewrite is complete.
|
||||
Keeping that boundary explicit prevents a passing hybrid build from being
|
||||
mistaken for the final all-Rust result.
|
||||
|
||||
## Compatibility boundary
|
||||
|
||||
@@ -161,9 +105,8 @@ from `rust/cli` as well:
|
||||
```sh
|
||||
cargo clippy --all-targets -- -D warnings
|
||||
cargo test --all-targets
|
||||
cargo test --no-default-features --features cli,compression --all-targets
|
||||
cargo test --no-default-features --features cli,decompression --all-targets
|
||||
cargo test --no-default-features --all-targets
|
||||
cargo test --no-default-features --features compression --all-targets
|
||||
cargo test --no-default-features --features decompression --all-targets
|
||||
```
|
||||
|
||||
Then run original compatibility tests from the repository root, starting with
|
||||
|
||||
Generated
-9
@@ -2,15 +2,6 @@
|
||||
# It is not intended for manual editing.
|
||||
version = 4
|
||||
|
||||
[[package]]
|
||||
name = "libc"
|
||||
version = "0.2.186"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66"
|
||||
|
||||
[[package]]
|
||||
name = "zstd-cli-rs"
|
||||
version = "0.1.0"
|
||||
dependencies = [
|
||||
"libc",
|
||||
]
|
||||
|
||||
+1
-8
@@ -7,13 +7,6 @@ edition = "2021"
|
||||
crate-type = ["staticlib"]
|
||||
|
||||
[features]
|
||||
default = ["cli", "compression", "decompression"]
|
||||
# The command-line parser and dispatch layer, which requires the C fileio
|
||||
# backend at link time. Program archives enable it; C test binaries link a
|
||||
# helpers-only archive (timefn) built without it.
|
||||
cli = []
|
||||
default = ["compression", "decompression"]
|
||||
compression = []
|
||||
decompression = []
|
||||
|
||||
[dependencies]
|
||||
libc = "0.2"
|
||||
|
||||
@@ -1,7 +1,2 @@
|
||||
#[path = "../../src/benchfn.rs"]
|
||||
mod benchfn;
|
||||
#[path = "../../src/timefn.rs"]
|
||||
mod timefn;
|
||||
#[cfg(feature = "cli")]
|
||||
#[path = "../../src/zstd_cli.rs"]
|
||||
mod zstd_cli;
|
||||
|
||||
@@ -1,575 +0,0 @@
|
||||
#![allow(non_camel_case_types)]
|
||||
#![allow(non_snake_case)]
|
||||
#![allow(clippy::missing_safety_doc)]
|
||||
|
||||
//! Benchmark loop for arbitrary functions over a set of blocks.
|
||||
//!
|
||||
//! Port of `programs/benchfn.c`. `BMK_benchFunction` measures one batch of
|
||||
//! runs; `BMK_benchTimedFn` repeats batches, growing the loop count until a
|
||||
//! run lasts long enough to be reported reliably against `run_ms`, within a
|
||||
//! `total_ms` budget tracked by `BMK_timedFnState_t`.
|
||||
//!
|
||||
//! ABI notes: `BMK_runOutcome_t` and `BMK_runTime_t` are returned by value
|
||||
//! across the C boundary and `BMK_benchParams_t` is passed by value, so all
|
||||
//! three are `repr(C)` mirrors of the `benchfn.h` layout, pinned by asserts
|
||||
//! here and in the C shim. `BMK_timedFnState_t` is opaque to C, but
|
||||
//! `BMK_initStatic_timedFnState` guarantees it fits the 64-byte
|
||||
//! `BMK_timedFnState_shell`, and `BMK_createTimedFnState` uses `malloc` so
|
||||
//! creation and destruction stay interchangeable with C callers.
|
||||
|
||||
use std::os::raw::{c_int, c_uint, c_void};
|
||||
use std::ptr;
|
||||
|
||||
use crate::timefn::{PTime, UTIL_clockSpanNano, UTIL_getTime, UTIL_time_t};
|
||||
|
||||
const TIMELOOP_NANOSEC: PTime = 1_000_000_000;
|
||||
|
||||
/// Valid benchmark result (`BMK_runTime_t` in benchfn.h).
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
pub struct BMK_runTime_t {
|
||||
/// Time per iteration, over all blocks.
|
||||
pub nanoSecPerRun: f64,
|
||||
/// Sum of the benchmarked function's return values, first loop only.
|
||||
pub sumOfReturn: usize,
|
||||
}
|
||||
|
||||
/// Outcome variant of a benchmark run (`BMK_runOutcome_t` in benchfn.h):
|
||||
/// either a valid `BMK_runTime_t` or an error result. C callers treat it as
|
||||
/// opaque and use the accessor functions below.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
pub struct BMK_runOutcome_t {
|
||||
pub internal_never_ever_use_directly: BMK_runTime_t,
|
||||
pub error_result_never_ever_use_directly: usize,
|
||||
pub error_tag_never_ever_use_directly: c_int,
|
||||
}
|
||||
|
||||
// These mirror the static asserts in the programs/benchfn.c shim: the structs
|
||||
// cross the ABI by value, so field offsets must match the C header exactly.
|
||||
const _: () =
|
||||
assert!(std::mem::offset_of!(BMK_runTime_t, sumOfReturn) == std::mem::size_of::<f64>());
|
||||
const _: () = assert!(
|
||||
std::mem::offset_of!(BMK_runOutcome_t, error_result_never_ever_use_directly)
|
||||
== std::mem::size_of::<BMK_runTime_t>()
|
||||
);
|
||||
const _: () = assert!(
|
||||
std::mem::offset_of!(BMK_runOutcome_t, error_tag_never_ever_use_directly)
|
||||
== std::mem::size_of::<BMK_runTime_t>() + std::mem::size_of::<usize>()
|
||||
);
|
||||
|
||||
/// `size_t (*BMK_benchFn_t)(const void*, size_t, void*, size_t, void*)`
|
||||
pub type BMK_benchFn_t = Option<
|
||||
unsafe extern "C" fn(
|
||||
src: *const c_void,
|
||||
srcSize: usize,
|
||||
dst: *mut c_void,
|
||||
dstCapacity: usize,
|
||||
customPayload: *mut c_void,
|
||||
) -> usize,
|
||||
>;
|
||||
/// `size_t (*BMK_initFn_t)(void*)`
|
||||
pub type BMK_initFn_t = Option<unsafe extern "C" fn(initPayload: *mut c_void) -> usize>;
|
||||
/// `unsigned (*BMK_errorFn_t)(size_t)`
|
||||
pub type BMK_errorFn_t = Option<unsafe extern "C" fn(result: usize) -> c_uint>;
|
||||
|
||||
/// Parameters of `BMK_benchFunction`, passed by value (`BMK_benchParams_t`).
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy)]
|
||||
pub struct BMK_benchParams_t {
|
||||
pub benchFn: BMK_benchFn_t,
|
||||
pub benchPayload: *mut c_void,
|
||||
pub initFn: BMK_initFn_t,
|
||||
pub initPayload: *mut c_void,
|
||||
pub errorFn: BMK_errorFn_t,
|
||||
pub blockCount: usize,
|
||||
pub srcBuffers: *const *const c_void,
|
||||
pub srcSizes: *const usize,
|
||||
pub dstBuffers: *const *mut c_void,
|
||||
pub dstCapacities: *const usize,
|
||||
pub blockResults: *mut usize,
|
||||
}
|
||||
|
||||
/// Aborts, like benchfn.c's `CONTROL`, when an accessor is used on the wrong
|
||||
/// outcome variant.
|
||||
fn control(condition: bool) {
|
||||
if !condition {
|
||||
std::process::abort();
|
||||
}
|
||||
}
|
||||
|
||||
fn error_outcome(errorResult: usize) -> BMK_runOutcome_t {
|
||||
BMK_runOutcome_t {
|
||||
internal_never_ever_use_directly: BMK_runTime_t {
|
||||
nanoSecPerRun: 0.0,
|
||||
sumOfReturn: 0,
|
||||
},
|
||||
error_result_never_ever_use_directly: errorResult,
|
||||
error_tag_never_ever_use_directly: 1,
|
||||
}
|
||||
}
|
||||
|
||||
fn valid_outcome(runTime: BMK_runTime_t) -> BMK_runOutcome_t {
|
||||
BMK_runOutcome_t {
|
||||
internal_never_ever_use_directly: runTime,
|
||||
error_result_never_ever_use_directly: 0,
|
||||
error_tag_never_ever_use_directly: 0,
|
||||
}
|
||||
}
|
||||
|
||||
/// Tells if the outcome carries a valid measurement.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn BMK_isSuccessful_runOutcome(outcome: BMK_runOutcome_t) -> c_int {
|
||||
c_int::from(outcome.error_tag_never_ever_use_directly == 0)
|
||||
}
|
||||
|
||||
/// Extracts the measurement; aborts if the outcome is an error, so validity
|
||||
/// must be checked first with `BMK_isSuccessful_runOutcome`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn BMK_extract_runTime(outcome: BMK_runOutcome_t) -> BMK_runTime_t {
|
||||
control(outcome.error_tag_never_ever_use_directly == 0);
|
||||
outcome.internal_never_ever_use_directly
|
||||
}
|
||||
|
||||
/// Extracts the faulty `benchFn` return value; aborts if the outcome is
|
||||
/// valid, so failure must be checked first.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn BMK_extract_errorResult(outcome: BMK_runOutcome_t) -> usize {
|
||||
control(outcome.error_tag_never_ever_use_directly != 0);
|
||||
outcome.error_result_never_ever_use_directly
|
||||
}
|
||||
|
||||
/// Runs `initFn` once, then `benchFn` `nbLoops` times over every block, and
|
||||
/// reports the mean time per loop. On the first loop, per-block results are
|
||||
/// stored into `blockResults` (when provided) and checked with `errorFn`
|
||||
/// (when provided); the first failing block aborts the measurement and
|
||||
/// produces an error outcome carrying the faulty return value.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn BMK_benchFunction(
|
||||
p: BMK_benchParams_t,
|
||||
mut nbLoops: c_uint,
|
||||
) -> BMK_runOutcome_t {
|
||||
// Minimum nbLoops is 1.
|
||||
nbLoops += c_uint::from(nbLoops == 0);
|
||||
|
||||
// Warm up and erase the result buffers.
|
||||
for blockNb in 0..p.blockCount {
|
||||
unsafe {
|
||||
let dst = *p.dstBuffers.add(blockNb);
|
||||
ptr::write_bytes(dst.cast::<u8>(), 0xE5, *p.dstCapacities.add(blockNb));
|
||||
}
|
||||
}
|
||||
|
||||
let benchFn = p.benchFn.expect("benchFn is mandatory");
|
||||
let mut dstSize = 0usize;
|
||||
let clockStart = UTIL_getTime();
|
||||
if let Some(initFn) = p.initFn {
|
||||
unsafe { initFn(p.initPayload) };
|
||||
}
|
||||
for loopNb in 0..nbLoops {
|
||||
for blockNb in 0..p.blockCount {
|
||||
let res = unsafe {
|
||||
benchFn(
|
||||
*p.srcBuffers.add(blockNb),
|
||||
*p.srcSizes.add(blockNb),
|
||||
*p.dstBuffers.add(blockNb),
|
||||
*p.dstCapacities.add(blockNb),
|
||||
p.benchPayload,
|
||||
)
|
||||
};
|
||||
if loopNb == 0 {
|
||||
if !p.blockResults.is_null() {
|
||||
unsafe { *p.blockResults.add(blockNb) = res };
|
||||
}
|
||||
if let Some(errorFn) = p.errorFn {
|
||||
if unsafe { errorFn(res) } != 0 {
|
||||
return error_outcome(res);
|
||||
}
|
||||
}
|
||||
dstSize = dstSize.wrapping_add(res);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
let totalTime = UTIL_clockSpanNano(clockStart);
|
||||
valid_outcome(BMK_runTime_t {
|
||||
nanoSecPerRun: totalTime as f64 / f64::from(nbLoops),
|
||||
sumOfReturn: dstSize,
|
||||
})
|
||||
}
|
||||
|
||||
/// Benchmark session state (`struct BMK_timedFnState_s`), opaque to C.
|
||||
#[repr(C)]
|
||||
pub struct BMK_timedFnState_t {
|
||||
timeSpent_ns: PTime,
|
||||
timeBudget_ns: PTime,
|
||||
runBudget_ns: PTime,
|
||||
fastestRun: BMK_runTime_t,
|
||||
nbLoops: c_uint,
|
||||
coolTime: UTIL_time_t,
|
||||
}
|
||||
|
||||
/// `BMK_TIMEDFNSTATE_SIZE` in benchfn.h: capacity of the caller-provided
|
||||
/// `BMK_timedFnState_shell`, which the state must always fit.
|
||||
const BMK_TIMEDFNSTATE_SIZE: usize = 64;
|
||||
const _: () = assert!(std::mem::size_of::<BMK_timedFnState_t>() <= BMK_TIMEDFNSTATE_SIZE);
|
||||
// The shell aligns via a `long long` member; the state must not need more.
|
||||
const _: () = assert!(std::mem::align_of::<BMK_timedFnState_t>() <= std::mem::align_of::<u64>());
|
||||
|
||||
/// Allocates and initializes a benchmark session lasting a minimum of
|
||||
/// `total_ms`, paced at intervals of approximately `run_ms`. Uses `malloc`
|
||||
/// so ownership stays interchangeable with the original C implementation.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn BMK_createTimedFnState(
|
||||
total_ms: c_uint,
|
||||
run_ms: c_uint,
|
||||
) -> *mut BMK_timedFnState_t {
|
||||
let state = unsafe { libc::malloc(std::mem::size_of::<BMK_timedFnState_t>()) }
|
||||
.cast::<BMK_timedFnState_t>();
|
||||
if state.is_null() {
|
||||
return ptr::null_mut();
|
||||
}
|
||||
unsafe { BMK_resetTimedFnState(state, total_ms, run_ms) };
|
||||
state
|
||||
}
|
||||
|
||||
/// Releases a state obtained from `BMK_createTimedFnState`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn BMK_freeTimedFnState(state: *mut BMK_timedFnState_t) {
|
||||
unsafe { libc::free(state.cast()) };
|
||||
}
|
||||
|
||||
/// Places the session state into a caller-provided buffer, typically a
|
||||
/// `BMK_timedFnState_shell`. Returns NULL when the buffer is missing, too
|
||||
/// small, or misaligned.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn BMK_initStatic_timedFnState(
|
||||
buffer: *mut c_void,
|
||||
size: usize,
|
||||
total_ms: c_uint,
|
||||
run_ms: c_uint,
|
||||
) -> *mut BMK_timedFnState_t {
|
||||
if buffer.is_null() {
|
||||
return ptr::null_mut();
|
||||
}
|
||||
if size < std::mem::size_of::<BMK_timedFnState_t>() {
|
||||
return ptr::null_mut();
|
||||
}
|
||||
if !(buffer as usize).is_multiple_of(std::mem::align_of::<BMK_timedFnState_t>()) {
|
||||
return ptr::null_mut();
|
||||
}
|
||||
let state = buffer.cast::<BMK_timedFnState_t>();
|
||||
unsafe { BMK_resetTimedFnState(state, total_ms, run_ms) };
|
||||
state
|
||||
}
|
||||
|
||||
/// Re-arms a session for a new benchmark of `total_ms`, paced at `run_ms`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn BMK_resetTimedFnState(
|
||||
timedFnState: *mut BMK_timedFnState_t,
|
||||
total_ms: c_uint,
|
||||
run_ms: c_uint,
|
||||
) {
|
||||
let total_ms = if total_ms == 0 { 1 } else { total_ms };
|
||||
let mut run_ms = if run_ms == 0 { 1 } else { run_ms };
|
||||
if run_ms > total_ms {
|
||||
run_ms = total_ms;
|
||||
}
|
||||
let state = BMK_timedFnState_t {
|
||||
timeSpent_ns: 0,
|
||||
timeBudget_ns: PTime::from(total_ms) * TIMELOOP_NANOSEC / 1000,
|
||||
runBudget_ns: PTime::from(run_ms) * TIMELOOP_NANOSEC / 1000,
|
||||
fastestRun: BMK_runTime_t {
|
||||
// Must be larger than any potential measurement.
|
||||
nanoSecPerRun: TIMELOOP_NANOSEC as f64 * 2_000_000_000.0,
|
||||
sumOfReturn: usize::MAX,
|
||||
},
|
||||
nbLoops: 1,
|
||||
coolTime: UTIL_getTime(),
|
||||
};
|
||||
unsafe { timedFnState.write(state) };
|
||||
}
|
||||
|
||||
/// Tells if the total time budget of the session is spent. Also reports 1
|
||||
/// after `BMK_benchTimedFn` returned an error.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn BMK_isCompleted_TimedFn(timedFnState: *const BMK_timedFnState_t) -> c_int {
|
||||
let state = unsafe { &*timedFnState };
|
||||
c_int::from(state.timeSpent_ns >= state.timeBudget_ns)
|
||||
}
|
||||
|
||||
/// Runs one measurement supposed to last about `run_ms`, automatically
|
||||
/// scaling `nbLoops`. Runs shorter than half the run budget are re-tried
|
||||
/// with a larger workload instead of being reported, limiting rounding-error
|
||||
/// risks; the best (fastest) qualifying run is returned.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn BMK_benchTimedFn(
|
||||
cont: *mut BMK_timedFnState_t,
|
||||
p: BMK_benchParams_t,
|
||||
) -> BMK_runOutcome_t {
|
||||
let cont = unsafe { &mut *cont };
|
||||
let runBudget_ns = cont.runBudget_ns;
|
||||
let runTimeMin_ns = runBudget_ns / 2;
|
||||
let mut bestRunTime = cont.fastestRun;
|
||||
|
||||
loop {
|
||||
let runResult = unsafe { BMK_benchFunction(p, cont.nbLoops) };
|
||||
|
||||
if BMK_isSuccessful_runOutcome(runResult) == 0 {
|
||||
// Error: move out.
|
||||
return runResult;
|
||||
}
|
||||
|
||||
let newRunTime = BMK_extract_runTime(runResult);
|
||||
let loopDuration_ns = newRunTime.nanoSecPerRun * f64::from(cont.nbLoops);
|
||||
|
||||
cont.timeSpent_ns = cont.timeSpent_ns.wrapping_add(loopDuration_ns as PTime);
|
||||
|
||||
// Estimate nbLoops for the next run to last approximately run_ms.
|
||||
if loopDuration_ns > runBudget_ns as f64 / 50.0 {
|
||||
let fastestRun_ns = bestRunTime.nanoSecPerRun.min(newRunTime.nanoSecPerRun);
|
||||
cont.nbLoops = ((runBudget_ns as f64 / fastestRun_ns) as c_uint).wrapping_add(1);
|
||||
} else {
|
||||
// Previous run was too short: blindly increase workload by a
|
||||
// x10 multiplier.
|
||||
const MULTIPLIER: c_uint = 10;
|
||||
debug_assert!(cont.nbLoops < c_uint::MAX / MULTIPLIER); // avoid overflow
|
||||
cont.nbLoops = cont.nbLoops.wrapping_mul(MULTIPLIER);
|
||||
}
|
||||
|
||||
if loopDuration_ns < runTimeMin_ns as f64 {
|
||||
// Don't report results when the run time was too small, which
|
||||
// increases the risk of rounding errors.
|
||||
continue;
|
||||
}
|
||||
if newRunTime.nanoSecPerRun < bestRunTime.nanoSecPerRun {
|
||||
bestRunTime = newRunTime;
|
||||
}
|
||||
return valid_outcome(bestRunTime);
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
/// Test payload observed through `benchPayload`/`initPayload` pointers.
|
||||
#[derive(Default)]
|
||||
struct CallLog {
|
||||
bench_calls: usize,
|
||||
init_calls: usize,
|
||||
}
|
||||
|
||||
/// Counts invocations and reports `srcSize`, like a size-preserving codec.
|
||||
unsafe extern "C" fn counting_bench_fn(
|
||||
_src: *const c_void,
|
||||
srcSize: usize,
|
||||
_dst: *mut c_void,
|
||||
_dstCapacity: usize,
|
||||
payload: *mut c_void,
|
||||
) -> usize {
|
||||
let log = unsafe { &mut *payload.cast::<CallLog>() };
|
||||
log.bench_calls += 1;
|
||||
srcSize
|
||||
}
|
||||
|
||||
unsafe extern "C" fn counting_init_fn(payload: *mut c_void) -> usize {
|
||||
let log = unsafe { &mut *payload.cast::<CallLog>() };
|
||||
log.init_calls += 1;
|
||||
0
|
||||
}
|
||||
|
||||
/// Flags results of 5 bytes and above as errors.
|
||||
unsafe extern "C" fn error_on_5(result: usize) -> c_uint {
|
||||
c_uint::from(result >= 5)
|
||||
}
|
||||
|
||||
struct Fixture {
|
||||
srcs: Vec<Vec<u8>>,
|
||||
dsts: Vec<Vec<u8>>,
|
||||
src_ptrs: Vec<*const c_void>,
|
||||
src_sizes: Vec<usize>,
|
||||
dst_ptrs: Vec<*mut c_void>,
|
||||
dst_capacities: Vec<usize>,
|
||||
block_results: Vec<usize>,
|
||||
log: CallLog,
|
||||
}
|
||||
|
||||
impl Fixture {
|
||||
fn new(block_sizes: &[usize]) -> Box<Self> {
|
||||
let srcs: Vec<Vec<u8>> = block_sizes.iter().map(|size| vec![0u8; *size]).collect();
|
||||
let mut dsts: Vec<Vec<u8>> = block_sizes.iter().map(|size| vec![0u8; *size]).collect();
|
||||
let src_ptrs = srcs.iter().map(|src| src.as_ptr().cast()).collect();
|
||||
let src_sizes = srcs.iter().map(Vec::len).collect();
|
||||
let dst_ptrs = dsts.iter_mut().map(|dst| dst.as_mut_ptr().cast()).collect();
|
||||
let dst_capacities = dsts.iter().map(Vec::len).collect();
|
||||
let block_results = vec![0usize; block_sizes.len()];
|
||||
Box::new(Self {
|
||||
srcs,
|
||||
dsts,
|
||||
src_ptrs,
|
||||
src_sizes,
|
||||
dst_ptrs,
|
||||
dst_capacities,
|
||||
block_results,
|
||||
log: CallLog::default(),
|
||||
})
|
||||
}
|
||||
|
||||
fn params(&mut self, errorFn: BMK_errorFn_t) -> BMK_benchParams_t {
|
||||
let payload: *mut CallLog = &mut self.log;
|
||||
BMK_benchParams_t {
|
||||
benchFn: Some(counting_bench_fn),
|
||||
benchPayload: payload.cast(),
|
||||
initFn: Some(counting_init_fn),
|
||||
initPayload: payload.cast(),
|
||||
errorFn,
|
||||
blockCount: self.srcs.len(),
|
||||
srcBuffers: self.src_ptrs.as_ptr(),
|
||||
srcSizes: self.src_sizes.as_ptr(),
|
||||
dstBuffers: self.dst_ptrs.as_ptr(),
|
||||
dstCapacities: self.dst_capacities.as_ptr(),
|
||||
blockResults: self.block_results.as_mut_ptr(),
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn bench_function_accounts_loops_blocks_and_first_loop_results() {
|
||||
let mut fixture = Fixture::new(&[3, 8]);
|
||||
let params = fixture.params(None);
|
||||
|
||||
let outcome = unsafe { BMK_benchFunction(params, 4) };
|
||||
|
||||
assert_eq!(BMK_isSuccessful_runOutcome(outcome), 1);
|
||||
let run_time = BMK_extract_runTime(outcome);
|
||||
// benchFn ran nbLoops times over each block; initFn ran once.
|
||||
assert_eq!(fixture.log.bench_calls, 4 * 2);
|
||||
assert_eq!(fixture.log.init_calls, 1);
|
||||
// sumOfReturn and blockResults reflect the first loop only.
|
||||
assert_eq!(run_time.sumOfReturn, 3 + 8);
|
||||
assert_eq!(fixture.block_results, vec![3, 8]);
|
||||
assert!(run_time.nanoSecPerRun >= 0.0);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn bench_function_treats_zero_loops_as_one_and_warms_up_buffers() {
|
||||
let mut fixture = Fixture::new(&[4]);
|
||||
let params = fixture.params(None);
|
||||
|
||||
let outcome = unsafe { BMK_benchFunction(params, 0) };
|
||||
|
||||
assert_eq!(BMK_isSuccessful_runOutcome(outcome), 1);
|
||||
assert_eq!(fixture.log.bench_calls, 1);
|
||||
// The result buffer was erased with the 0xE5 warm-up pattern.
|
||||
assert_eq!(fixture.dsts[0], vec![0xE5; 4]);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn bench_function_reports_the_first_failing_block() {
|
||||
let mut fixture = Fixture::new(&[3, 5, 7]);
|
||||
let params = fixture.params(Some(error_on_5));
|
||||
|
||||
let outcome = unsafe { BMK_benchFunction(params, 10) };
|
||||
|
||||
assert_eq!(BMK_isSuccessful_runOutcome(outcome), 0);
|
||||
assert_eq!(BMK_extract_errorResult(outcome), 5);
|
||||
// Execution stopped at the failing block, before the third one.
|
||||
assert_eq!(fixture.log.bench_calls, 2);
|
||||
// blockResults were recorded up to and including the failure.
|
||||
assert_eq!(fixture.block_results[..2], [3, 5]);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn reset_clamps_budgets_and_rearms_the_loop_counter() {
|
||||
let state = BMK_createTimedFnState(0, 7);
|
||||
assert!(!state.is_null());
|
||||
|
||||
{
|
||||
let state = unsafe { &*state };
|
||||
// total_ms 0 becomes 1ms, and run_ms is clamped to total_ms.
|
||||
assert_eq!(state.timeBudget_ns, 1_000_000);
|
||||
assert_eq!(state.runBudget_ns, 1_000_000);
|
||||
assert_eq!(state.nbLoops, 1);
|
||||
assert_eq!(state.timeSpent_ns, 0);
|
||||
assert_eq!(state.fastestRun.sumOfReturn, usize::MAX);
|
||||
}
|
||||
assert_eq!(unsafe { BMK_isCompleted_TimedFn(state) }, 0);
|
||||
|
||||
unsafe { BMK_resetTimedFnState(state, 2_000, 500) };
|
||||
{
|
||||
let state = unsafe { &*state };
|
||||
assert_eq!(state.timeBudget_ns, 2_000_000_000);
|
||||
assert_eq!(state.runBudget_ns, 500_000_000);
|
||||
}
|
||||
|
||||
unsafe { BMK_freeTimedFnState(state) };
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn static_state_initialization_validates_its_buffer() {
|
||||
let mut shell = [0u64; BMK_TIMEDFNSTATE_SIZE / 8];
|
||||
let buffer: *mut c_void = shell.as_mut_ptr().cast();
|
||||
|
||||
// A properly sized and aligned buffer is accepted.
|
||||
let state = unsafe { BMK_initStatic_timedFnState(buffer, 64, 1_000, 100) };
|
||||
assert!(!state.is_null());
|
||||
assert_eq!(unsafe { BMK_isCompleted_TimedFn(state) }, 0);
|
||||
|
||||
// NULL, undersized, and misaligned buffers are rejected.
|
||||
let too_small = std::mem::size_of::<BMK_timedFnState_t>() - 1;
|
||||
unsafe {
|
||||
assert!(BMK_initStatic_timedFnState(ptr::null_mut(), 64, 1, 1).is_null());
|
||||
assert!(BMK_initStatic_timedFnState(buffer, too_small, 1, 1).is_null());
|
||||
assert!(
|
||||
BMK_initStatic_timedFnState(buffer.cast::<u8>().add(1).cast(), 63, 1, 1).is_null()
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn timed_runs_grow_the_workload_and_spend_the_budget() {
|
||||
let mut fixture = Fixture::new(&[16]);
|
||||
let params = fixture.params(None);
|
||||
let state = BMK_createTimedFnState(4, 2);
|
||||
assert!(!state.is_null());
|
||||
|
||||
let mut rounds = 0usize;
|
||||
while unsafe { BMK_isCompleted_TimedFn(state) } == 0 {
|
||||
let outcome = unsafe { BMK_benchTimedFn(state, params) };
|
||||
assert_eq!(BMK_isSuccessful_runOutcome(outcome), 1);
|
||||
let run_time = BMK_extract_runTime(outcome);
|
||||
assert_eq!(run_time.sumOfReturn, 16);
|
||||
rounds += 1;
|
||||
assert!(rounds < 1_000, "the time budget must eventually be spent");
|
||||
}
|
||||
|
||||
{
|
||||
let state = unsafe { &*state };
|
||||
// A reported run had to last at least runBudget/2, which is only
|
||||
// reachable for this trivial function with a grown loop counter.
|
||||
assert!(state.nbLoops > 1);
|
||||
assert!(state.timeSpent_ns >= state.timeBudget_ns);
|
||||
}
|
||||
// Every reported outcome came from a run of >= runBudget/2, and the
|
||||
// budget accounting matches BMK_isCompleted_TimedFn.
|
||||
assert!(rounds >= 1);
|
||||
assert!(fixture.log.bench_calls >= rounds);
|
||||
|
||||
unsafe { BMK_freeTimedFnState(state) };
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn timed_runs_propagate_errors_without_aborting() {
|
||||
let mut fixture = Fixture::new(&[9]);
|
||||
let params = fixture.params(Some(error_on_5));
|
||||
let state = BMK_createTimedFnState(1_000, 100);
|
||||
|
||||
let outcome = unsafe { BMK_benchTimedFn(state, params) };
|
||||
|
||||
assert_eq!(BMK_isSuccessful_runOutcome(outcome), 0);
|
||||
assert_eq!(BMK_extract_errorResult(outcome), 9);
|
||||
|
||||
unsafe { BMK_freeTimedFnState(state) };
|
||||
}
|
||||
}
|
||||
@@ -1,40 +0,0 @@
|
||||
//! Frozen decoders for the legacy zstd formats (v0.1 through v0.7).
|
||||
//!
|
||||
//! Frozen-decoder policy
|
||||
//! =====================
|
||||
//!
|
||||
//! Each `lib/legacy/zstd_v0N.c` translation unit is a self-contained snapshot
|
||||
//! of the entropy coders and frame logic of that historical release. The
|
||||
//! Rust ports mirror that property:
|
||||
//!
|
||||
//! - Every version keeps its own frozen FSE/Huff0 and frame logic. The
|
||||
//! modern `fse_decompress`, `huf_decompress`, `bitstream`, or `mem`
|
||||
//! modules must NOT be reused here, and legacy versions must not share
|
||||
//! code with each other, even where functions look identical. The legacy
|
||||
//! formats are frozen; the modern modules keep evolving.
|
||||
//! - Ports are line-by-line translations of the corresponding C file: same
|
||||
//! table layouts, same arithmetic, same error codes. Outputs must be
|
||||
//! byte-identical to the C implementation, including error behavior.
|
||||
//! - The only shared dependency is `crate::errors`, because the C files
|
||||
//! include `error_private.h` for the public `ZSTD_ErrorCode` values.
|
||||
//!
|
||||
//! Registration
|
||||
//! ============
|
||||
//!
|
||||
//! Cargo features `legacy-v01` .. `legacy-v07` are all declared in
|
||||
//! `Cargo.toml`. The build systems always pass the feature list derived
|
||||
//! from the C configuration (`ZSTD_LEGACY_SUPPORT=N` enables versions N
|
||||
//! and newer), so a feature may be enabled before its port exists. A version
|
||||
//! without a Rust module simply stays implemented by its C file.
|
||||
//!
|
||||
//! To port version `v0N`: add `zstd_v0N.rs` next to this file, reduce
|
||||
//! `lib/legacy/zstd_v0N.c` to a declaration-only shim, and register the
|
||||
//! module here with exactly one line:
|
||||
//!
|
||||
//! ```text
|
||||
//! #[cfg(feature = "legacy-v0N")]
|
||||
//! pub mod zstd_v0N;
|
||||
//! ```
|
||||
|
||||
#[cfg(feature = "legacy-v01")]
|
||||
pub mod zstd_v01;
|
||||
@@ -1,2354 +0,0 @@
|
||||
#![allow(non_snake_case)]
|
||||
|
||||
//! Frozen decoder for the zstd v0.1 format.
|
||||
//!
|
||||
//! This is a line-by-line port of `lib/legacy/zstd_v01.c`: the same table
|
||||
//! layouts, the same arithmetic, and the same error codes. The C file is a
|
||||
//! self-contained snapshot of the v0.1-era FSE and Huff0 coders, so this
|
||||
//! module deliberately reimplements them instead of reusing the modern
|
||||
//! `fse_decompress`/`huf_decompress` modules (see `legacy/mod.rs` for the
|
||||
//! frozen-decoder policy). The only shared dependency is `crate::errors`,
|
||||
//! mirroring the C file's `error_private.h` include.
|
||||
//!
|
||||
//! The streaming `ZSTDv01_Dctx` state lives entirely in Rust; C callers only
|
||||
//! ever hold an opaque pointer to it. It is allocated with `libc::malloc`
|
||||
//! and released with `libc::free`, exactly like the original C context.
|
||||
|
||||
use crate::errors::{ERR_isError, ZstdErrorCode, ERROR};
|
||||
use std::os::raw::{c_uint, c_void};
|
||||
use std::ptr;
|
||||
|
||||
/* ******************************************
|
||||
* Error management (frozen v0.1 FSE codes)
|
||||
********************************************/
|
||||
/* FSE_LIST_ERRORS in zstd_v01.c; values are returned as `(size_t)-code`. */
|
||||
const FSE_ERROR_GENERIC: usize = 1;
|
||||
const FSE_ERROR_TABLELOG_TOO_LARGE: usize = 2;
|
||||
const FSE_ERROR_MAX_SYMBOL_VALUE_TOO_LARGE: usize = 3;
|
||||
const FSE_ERROR_MAX_SYMBOL_VALUE_TOO_SMALL: usize = 4;
|
||||
const FSE_ERROR_DST_SIZE_TOO_SMALL: usize = 5;
|
||||
const FSE_ERROR_SRC_SIZE_WRONG: usize = 6;
|
||||
const FSE_ERROR_CORRUPTION_DETECTED: usize = 7;
|
||||
const FSE_ERROR_MAX_CODE: usize = 8;
|
||||
|
||||
#[inline]
|
||||
fn fse_error(code: usize) -> usize {
|
||||
code.wrapping_neg()
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn fse_is_error(code: usize) -> bool {
|
||||
code > fse_error(FSE_ERROR_MAX_CODE)
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* Tuning parameters (frozen)
|
||||
********************************************/
|
||||
const FSE_MAX_MEMORY_USAGE: u32 = 14;
|
||||
const FSE_MAX_SYMBOL_VALUE: u32 = 255;
|
||||
|
||||
const FSE_MAX_TABLELOG: u32 = FSE_MAX_MEMORY_USAGE - 2;
|
||||
const FSE_MIN_TABLELOG: u32 = 5;
|
||||
const FSE_TABLELOG_ABSOLUTE_MAX: u32 = 15;
|
||||
|
||||
const HUF_MAX_SYMBOL_VALUE: u32 = 255;
|
||||
const HUF_MAX_TABLELOG: u32 = 12;
|
||||
const HUF_ABSOLUTEMAX_TABLELOG: u32 = 16;
|
||||
|
||||
const IS_32BITS: bool = std::mem::size_of::<usize>() == 4;
|
||||
const USIZE_BITS: u32 = usize::BITS;
|
||||
|
||||
/* ******************************************
|
||||
* Memory I/O (FSE_read* / ZSTD_read* helpers)
|
||||
********************************************/
|
||||
#[inline]
|
||||
unsafe fn fse_read_le16(mem_ptr: *const u8) -> u16 {
|
||||
u16::from_le_bytes(ptr::read_unaligned(mem_ptr as *const [u8; 2]))
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_read_le32(mem_ptr: *const u8) -> u32 {
|
||||
u32::from_le_bytes(ptr::read_unaligned(mem_ptr as *const [u8; 4]))
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_read_le64(mem_ptr: *const u8) -> u64 {
|
||||
u64::from_le_bytes(ptr::read_unaligned(mem_ptr as *const [u8; 8]))
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_read_lest(mem_ptr: *const u8) -> usize {
|
||||
if IS_32BITS {
|
||||
fse_read_le32(mem_ptr) as usize
|
||||
} else {
|
||||
fse_read_le64(mem_ptr) as usize
|
||||
}
|
||||
}
|
||||
|
||||
/// `FSE_highbit32`; the caller guarantees `val != 0`, as in C.
|
||||
#[inline]
|
||||
fn fse_highbit32(val: u32) -> u32 {
|
||||
val.leading_zeros() ^ 31
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* FSE structures
|
||||
********************************************/
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy)]
|
||||
struct FseDecode {
|
||||
new_state: u16,
|
||||
symbol: u8,
|
||||
nb_bits: u8,
|
||||
}
|
||||
|
||||
#[repr(C)]
|
||||
struct FseDTableHeader {
|
||||
table_log: u16,
|
||||
fast_mode: u16,
|
||||
}
|
||||
|
||||
struct FseDStream {
|
||||
bit_container: usize,
|
||||
bits_consumed: u32,
|
||||
ptr: *const u8,
|
||||
start: *const u8,
|
||||
}
|
||||
|
||||
struct FseDState {
|
||||
state: usize,
|
||||
table: *const FseDecode,
|
||||
}
|
||||
|
||||
const FSE_DSTREAM_UNFINISHED: u32 = 0;
|
||||
const FSE_DSTREAM_END_OF_BUFFER: u32 = 1;
|
||||
const FSE_DSTREAM_COMPLETED: u32 = 2;
|
||||
const FSE_DSTREAM_TOO_FAR: u32 = 3;
|
||||
|
||||
#[inline]
|
||||
fn fse_table_step(table_size: u32) -> u32 {
|
||||
(table_size >> 1) + (table_size >> 3) + 3
|
||||
}
|
||||
|
||||
/* An FSE_DTable is an opaque u32 array: one header word followed by
|
||||
* `1 << tableLog` FseDecode entries, exactly as in C. */
|
||||
unsafe fn fse_build_dtable(
|
||||
dt: *mut u32,
|
||||
normalized_counter: *const i16,
|
||||
max_symbol_value: u32,
|
||||
table_log: u32,
|
||||
) -> usize {
|
||||
let dtable_h = dt as *mut FseDTableHeader;
|
||||
let table_decode = dt.add(1) as *mut FseDecode;
|
||||
|
||||
/* Sanity checks */
|
||||
if max_symbol_value > FSE_MAX_SYMBOL_VALUE {
|
||||
return fse_error(FSE_ERROR_MAX_SYMBOL_VALUE_TOO_LARGE);
|
||||
}
|
||||
if table_log > FSE_MAX_TABLELOG {
|
||||
return fse_error(FSE_ERROR_TABLELOG_TOO_LARGE);
|
||||
}
|
||||
|
||||
let table_size: u32 = 1 << table_log;
|
||||
let table_mask = table_size - 1;
|
||||
let step = fse_table_step(table_size);
|
||||
let mut symbol_next = [0u16; (FSE_MAX_SYMBOL_VALUE + 1) as usize];
|
||||
let mut position: u32 = 0;
|
||||
let mut high_threshold = table_size - 1;
|
||||
let large_limit = (1i32 << (table_log - 1)) as i16;
|
||||
let mut no_large: u32 = 1;
|
||||
|
||||
/* Init, lay down lowprob symbols */
|
||||
(*dtable_h).table_log = table_log as u16;
|
||||
for s in 0..=max_symbol_value {
|
||||
let count = *normalized_counter.add(s as usize);
|
||||
if count == -1 {
|
||||
(*table_decode.add(high_threshold as usize)).symbol = s as u8;
|
||||
high_threshold = high_threshold.wrapping_sub(1);
|
||||
symbol_next[s as usize] = 1;
|
||||
} else {
|
||||
if count >= large_limit {
|
||||
no_large = 0;
|
||||
}
|
||||
symbol_next[s as usize] = count as u16;
|
||||
}
|
||||
}
|
||||
|
||||
/* Spread symbols */
|
||||
for s in 0..=max_symbol_value {
|
||||
let count = *normalized_counter.add(s as usize);
|
||||
let mut i = 0i32;
|
||||
while i < count as i32 {
|
||||
(*table_decode.add(position as usize)).symbol = s as u8;
|
||||
position = (position + step) & table_mask;
|
||||
while position > high_threshold {
|
||||
position = (position + step) & table_mask; /* lowprob area */
|
||||
}
|
||||
i += 1;
|
||||
}
|
||||
}
|
||||
|
||||
if position != 0 {
|
||||
/* position must reach all cells once, otherwise normalizedCounter is incorrect */
|
||||
return fse_error(FSE_ERROR_GENERIC);
|
||||
}
|
||||
|
||||
/* Build Decoding table */
|
||||
for i in 0..table_size as usize {
|
||||
let symbol = (*table_decode.add(i)).symbol;
|
||||
let next_state = symbol_next[symbol as usize];
|
||||
symbol_next[symbol as usize] = next_state.wrapping_add(1);
|
||||
let nb_bits = (table_log - fse_highbit32(next_state as u32)) as u8;
|
||||
(*table_decode.add(i)).nb_bits = nb_bits;
|
||||
(*table_decode.add(i)).new_state =
|
||||
(((next_state as u32) << nb_bits).wrapping_sub(table_size)) as u16;
|
||||
}
|
||||
|
||||
(*dtable_h).fast_mode = no_large as u16;
|
||||
0
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* FSE header bitstream (FSE_readNCount)
|
||||
********************************************/
|
||||
unsafe fn fse_read_ncount(
|
||||
normalized_counter: *mut i16,
|
||||
max_sv_ptr: &mut u32,
|
||||
table_log_ptr: &mut u32,
|
||||
header_buffer: *const u8,
|
||||
hb_size: usize,
|
||||
) -> usize {
|
||||
let istart = header_buffer;
|
||||
let iend_addr = (istart as usize).wrapping_add(hb_size);
|
||||
let mut ip = istart;
|
||||
let mut charnum: u32 = 0;
|
||||
let mut previous0 = false;
|
||||
|
||||
if hb_size < 4 {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
let mut bit_stream = fse_read_le32(ip);
|
||||
let mut nb_bits: i32 = ((bit_stream & 0xF) + FSE_MIN_TABLELOG) as i32; /* extract tableLog */
|
||||
if nb_bits > FSE_TABLELOG_ABSOLUTE_MAX as i32 {
|
||||
return fse_error(FSE_ERROR_TABLELOG_TOO_LARGE);
|
||||
}
|
||||
bit_stream >>= 4;
|
||||
let mut bit_count: i32 = 4;
|
||||
*table_log_ptr = nb_bits as u32;
|
||||
let mut remaining: i32 = (1 << nb_bits) + 1;
|
||||
let mut threshold: i32 = 1 << nb_bits;
|
||||
nb_bits += 1;
|
||||
|
||||
while remaining > 1 && charnum <= *max_sv_ptr {
|
||||
if previous0 {
|
||||
let mut n0 = charnum;
|
||||
while (bit_stream & 0xFFFF) == 0xFFFF {
|
||||
n0 += 24;
|
||||
if (ip as usize) < iend_addr.wrapping_sub(5) {
|
||||
ip = ip.add(2);
|
||||
bit_stream = fse_read_le32(ip).wrapping_shr(bit_count as u32);
|
||||
} else {
|
||||
bit_stream >>= 16;
|
||||
bit_count += 16;
|
||||
}
|
||||
}
|
||||
while (bit_stream & 3) == 3 {
|
||||
n0 += 3;
|
||||
bit_stream >>= 2;
|
||||
bit_count += 2;
|
||||
}
|
||||
n0 += bit_stream & 3;
|
||||
bit_count += 2;
|
||||
if n0 > *max_sv_ptr {
|
||||
return fse_error(FSE_ERROR_MAX_SYMBOL_VALUE_TOO_SMALL);
|
||||
}
|
||||
while charnum < n0 {
|
||||
*normalized_counter.add(charnum as usize) = 0;
|
||||
charnum += 1;
|
||||
}
|
||||
if (ip as usize) <= iend_addr.wrapping_sub(7)
|
||||
|| (ip as usize).wrapping_add((bit_count >> 3) as usize)
|
||||
<= iend_addr.wrapping_sub(4)
|
||||
{
|
||||
ip = ip.add((bit_count >> 3) as usize);
|
||||
bit_count &= 7;
|
||||
bit_stream = fse_read_le32(ip).wrapping_shr(bit_count as u32);
|
||||
} else {
|
||||
bit_stream >>= 2;
|
||||
}
|
||||
}
|
||||
{
|
||||
let max: i16 = ((2 * threshold - 1) - remaining) as i16;
|
||||
let mut count: i16;
|
||||
|
||||
if (bit_stream & (threshold - 1) as u32) < max as i32 as u32 {
|
||||
count = (bit_stream & (threshold - 1) as u32) as u16 as i16;
|
||||
bit_count += nb_bits - 1;
|
||||
} else {
|
||||
count = (bit_stream & (2 * threshold - 1) as u32) as u16 as i16;
|
||||
if count as i32 >= threshold {
|
||||
count = ((count as i32) - (max as i32)) as i16;
|
||||
}
|
||||
bit_count += nb_bits;
|
||||
}
|
||||
|
||||
count = count.wrapping_sub(1); /* extra accuracy */
|
||||
remaining -= (count as i32).abs();
|
||||
*normalized_counter.add(charnum as usize) = count;
|
||||
charnum += 1;
|
||||
previous0 = count == 0;
|
||||
while remaining < threshold {
|
||||
nb_bits -= 1;
|
||||
threshold >>= 1;
|
||||
}
|
||||
|
||||
if (ip as usize) <= iend_addr.wrapping_sub(7)
|
||||
|| (ip as usize).wrapping_add((bit_count >> 3) as usize)
|
||||
<= iend_addr.wrapping_sub(4)
|
||||
{
|
||||
ip = ip.add((bit_count >> 3) as usize);
|
||||
bit_count &= 7;
|
||||
} else {
|
||||
bit_count -=
|
||||
(8 * (iend_addr.wrapping_sub(4) as isize - ip as usize as isize)) as i32;
|
||||
ip = (iend_addr - 4) as *const u8;
|
||||
}
|
||||
bit_stream = fse_read_le32(ip).wrapping_shr((bit_count & 31) as u32);
|
||||
}
|
||||
}
|
||||
if remaining != 1 {
|
||||
return fse_error(FSE_ERROR_GENERIC);
|
||||
}
|
||||
*max_sv_ptr = charnum - 1;
|
||||
|
||||
ip = ip.wrapping_offset(((bit_count + 7) >> 3) as isize);
|
||||
if (ip as usize).wrapping_sub(istart as usize) > hb_size {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
(ip as usize).wrapping_sub(istart as usize)
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* FSE decompression, byte symbols
|
||||
********************************************/
|
||||
unsafe fn fse_build_dtable_rle(dt: *mut u32, symbol_value: u8) -> usize {
|
||||
let dtable_h = dt as *mut FseDTableHeader;
|
||||
let cell = dt.add(1) as *mut FseDecode;
|
||||
|
||||
(*dtable_h).table_log = 0;
|
||||
(*dtable_h).fast_mode = 0;
|
||||
|
||||
(*cell).new_state = 0;
|
||||
(*cell).symbol = symbol_value;
|
||||
(*cell).nb_bits = 0;
|
||||
|
||||
0
|
||||
}
|
||||
|
||||
unsafe fn fse_build_dtable_raw(dt: *mut u32, nb_bits: u32) -> usize {
|
||||
let dtable_h = dt as *mut FseDTableHeader;
|
||||
let dinfo = dt.add(1) as *mut FseDecode;
|
||||
|
||||
/* Sanity checks */
|
||||
if nb_bits < 1 {
|
||||
return fse_error(FSE_ERROR_GENERIC); /* min size */
|
||||
}
|
||||
let table_size: u32 = 1 << nb_bits;
|
||||
let table_mask = table_size - 1;
|
||||
let max_symbol_value = table_mask;
|
||||
|
||||
(*dtable_h).table_log = nb_bits as u16;
|
||||
(*dtable_h).fast_mode = 1;
|
||||
for s in 0..=max_symbol_value {
|
||||
let cell = dinfo.add(s as usize);
|
||||
(*cell).new_state = 0;
|
||||
(*cell).symbol = s as u8;
|
||||
(*cell).nb_bits = nb_bits as u8;
|
||||
}
|
||||
|
||||
0
|
||||
}
|
||||
|
||||
unsafe fn fse_init_dstream(
|
||||
bit_d: &mut FseDStream,
|
||||
src_buffer: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
let word = std::mem::size_of::<usize>();
|
||||
if src_size < 1 {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
|
||||
if src_size >= word {
|
||||
bit_d.start = src_buffer;
|
||||
bit_d.ptr = src_buffer.add(src_size - word);
|
||||
bit_d.bit_container = fse_read_lest(bit_d.ptr);
|
||||
let contain32 = *src_buffer.add(src_size - 1) as u32;
|
||||
if contain32 == 0 {
|
||||
return fse_error(FSE_ERROR_GENERIC); /* stop bit not present */
|
||||
}
|
||||
bit_d.bits_consumed = 8 - fse_highbit32(contain32);
|
||||
} else {
|
||||
bit_d.start = src_buffer;
|
||||
bit_d.ptr = bit_d.start;
|
||||
bit_d.bit_container = *bit_d.start as usize;
|
||||
/* C switch with fallthrough over srcSize 7..2 */
|
||||
if src_size >= 7 {
|
||||
bit_d.bit_container += (*src_buffer.add(6) as usize) << (USIZE_BITS as usize - 16);
|
||||
}
|
||||
if src_size >= 6 {
|
||||
bit_d.bit_container += (*src_buffer.add(5) as usize) << (USIZE_BITS as usize - 24);
|
||||
}
|
||||
if src_size >= 5 {
|
||||
bit_d.bit_container += (*src_buffer.add(4) as usize) << (USIZE_BITS as usize - 32);
|
||||
}
|
||||
if src_size >= 4 {
|
||||
bit_d.bit_container += (*src_buffer.add(3) as usize) << 24;
|
||||
}
|
||||
if src_size >= 3 {
|
||||
bit_d.bit_container += (*src_buffer.add(2) as usize) << 16;
|
||||
}
|
||||
if src_size >= 2 {
|
||||
bit_d.bit_container += (*src_buffer.add(1) as usize) << 8;
|
||||
}
|
||||
let contain32 = *src_buffer.add(src_size - 1) as u32;
|
||||
if contain32 == 0 {
|
||||
return fse_error(FSE_ERROR_GENERIC); /* stop bit not present */
|
||||
}
|
||||
bit_d.bits_consumed = 8 - fse_highbit32(contain32);
|
||||
bit_d.bits_consumed += ((word - src_size) * 8) as u32;
|
||||
}
|
||||
|
||||
src_size
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_look_bits(bit_d: &FseDStream, nb_bits: u32) -> usize {
|
||||
let bit_mask = USIZE_BITS - 1;
|
||||
((bit_d.bit_container << (bit_d.bits_consumed & bit_mask)) >> 1)
|
||||
>> (bit_mask.wrapping_sub(nb_bits) & bit_mask)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_look_bits_fast(bit_d: &FseDStream, nb_bits: u32) -> usize {
|
||||
/* only if nb_bits >= 1 */
|
||||
let bit_mask = USIZE_BITS - 1;
|
||||
(bit_d.bit_container << (bit_d.bits_consumed & bit_mask))
|
||||
>> ((bit_mask + 1).wrapping_sub(nb_bits) & bit_mask)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn fse_skip_bits(bit_d: &mut FseDStream, nb_bits: u32) {
|
||||
bit_d.bits_consumed = bit_d.bits_consumed.wrapping_add(nb_bits);
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_read_bits(bit_d: &mut FseDStream, nb_bits: u32) -> usize {
|
||||
let value = fse_look_bits(bit_d, nb_bits);
|
||||
fse_skip_bits(bit_d, nb_bits);
|
||||
value
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn fse_read_bits_fast(bit_d: &mut FseDStream, nb_bits: u32) -> usize {
|
||||
/* only if nb_bits >= 1 */
|
||||
let value = fse_look_bits_fast(bit_d, nb_bits);
|
||||
fse_skip_bits(bit_d, nb_bits);
|
||||
value
|
||||
}
|
||||
|
||||
unsafe fn fse_reload_dstream(bit_d: &mut FseDStream) -> u32 {
|
||||
let word = std::mem::size_of::<usize>();
|
||||
if bit_d.bits_consumed > (word * 8) as u32 {
|
||||
/* should never happen */
|
||||
return FSE_DSTREAM_TOO_FAR;
|
||||
}
|
||||
|
||||
if (bit_d.ptr as usize) >= (bit_d.start as usize).wrapping_add(word) {
|
||||
bit_d.ptr = bit_d.ptr.sub((bit_d.bits_consumed >> 3) as usize);
|
||||
bit_d.bits_consumed &= 7;
|
||||
bit_d.bit_container = fse_read_lest(bit_d.ptr);
|
||||
return FSE_DSTREAM_UNFINISHED;
|
||||
}
|
||||
if bit_d.ptr == bit_d.start {
|
||||
if bit_d.bits_consumed < (word * 8) as u32 {
|
||||
return FSE_DSTREAM_END_OF_BUFFER;
|
||||
}
|
||||
return FSE_DSTREAM_COMPLETED;
|
||||
}
|
||||
{
|
||||
let mut nb_bytes = bit_d.bits_consumed >> 3;
|
||||
let mut result = FSE_DSTREAM_UNFINISHED;
|
||||
if (bit_d.ptr as usize).wrapping_sub(nb_bytes as usize) < (bit_d.start as usize) {
|
||||
nb_bytes = ((bit_d.ptr as usize) - (bit_d.start as usize)) as u32; /* ptr > start */
|
||||
result = FSE_DSTREAM_END_OF_BUFFER;
|
||||
}
|
||||
bit_d.ptr = bit_d.ptr.sub(nb_bytes as usize);
|
||||
bit_d.bits_consumed -= nb_bytes * 8;
|
||||
bit_d.bit_container = fse_read_lest(bit_d.ptr); /* reminder : srcSize > sizeof(bitD) */
|
||||
result
|
||||
}
|
||||
}
|
||||
|
||||
unsafe fn fse_init_dstate(d_state: &mut FseDState, bit_d: &mut FseDStream, dt: *const u32) {
|
||||
let dtable_h = dt as *const FseDTableHeader;
|
||||
d_state.state = fse_read_bits(bit_d, (*dtable_h).table_log as u32);
|
||||
fse_reload_dstream(bit_d);
|
||||
d_state.table = dt.add(1) as *const FseDecode;
|
||||
}
|
||||
|
||||
unsafe fn fse_decode_symbol(d_state: &mut FseDState, bit_d: &mut FseDStream) -> u8 {
|
||||
let d_info = *d_state.table.add(d_state.state);
|
||||
let low_bits = fse_read_bits(bit_d, d_info.nb_bits as u32);
|
||||
d_state.state = (d_info.new_state as usize).wrapping_add(low_bits);
|
||||
d_info.symbol
|
||||
}
|
||||
|
||||
unsafe fn fse_decode_symbol_fast(d_state: &mut FseDState, bit_d: &mut FseDStream) -> u8 {
|
||||
let d_info = *d_state.table.add(d_state.state);
|
||||
let low_bits = fse_read_bits_fast(bit_d, d_info.nb_bits as u32);
|
||||
d_state.state = (d_info.new_state as usize).wrapping_add(low_bits);
|
||||
d_info.symbol
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn fse_end_of_dstream(bit_d: &FseDStream) -> bool {
|
||||
bit_d.ptr == bit_d.start && bit_d.bits_consumed == USIZE_BITS
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn fse_end_of_dstate(d_state: &FseDState) -> bool {
|
||||
d_state.state == 0
|
||||
}
|
||||
|
||||
unsafe fn fse_decompress_using_dtable_generic(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
c_src: *const u8,
|
||||
c_src_size: usize,
|
||||
dt: *const u32,
|
||||
fast: bool,
|
||||
) -> usize {
|
||||
let ostart = dst;
|
||||
let mut op = ostart;
|
||||
let omax_addr = (op as usize).wrapping_add(max_dst_size);
|
||||
let olimit_addr = omax_addr.wrapping_sub(3);
|
||||
|
||||
let mut bit_d = FseDStream {
|
||||
bit_container: 0,
|
||||
bits_consumed: 0,
|
||||
ptr: ptr::null(),
|
||||
start: ptr::null(),
|
||||
};
|
||||
let mut state1 = FseDState {
|
||||
state: 0,
|
||||
table: ptr::null(),
|
||||
};
|
||||
let mut state2 = FseDState {
|
||||
state: 0,
|
||||
table: ptr::null(),
|
||||
};
|
||||
|
||||
/* Init */
|
||||
let error_code = fse_init_dstream(&mut bit_d, c_src, c_src_size);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
|
||||
fse_init_dstate(&mut state1, &mut bit_d, dt);
|
||||
fse_init_dstate(&mut state2, &mut bit_d, dt);
|
||||
|
||||
macro_rules! fse_getsymbol {
|
||||
($state:expr) => {
|
||||
if fast {
|
||||
fse_decode_symbol_fast($state, &mut bit_d)
|
||||
} else {
|
||||
fse_decode_symbol($state, &mut bit_d)
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
/* Constant conditions from the C source; on 64-bit both are false, on
|
||||
* 32-bit only the *4 variant reloads. */
|
||||
const RELOAD_2: bool = FSE_MAX_TABLELOG * 2 + 7 > USIZE_BITS;
|
||||
const RELOAD_4: bool = FSE_MAX_TABLELOG * 4 + 7 > USIZE_BITS;
|
||||
|
||||
/* 4 symbols per loop */
|
||||
while fse_reload_dstream(&mut bit_d) == FSE_DSTREAM_UNFINISHED && (op as usize) < olimit_addr {
|
||||
*op = fse_getsymbol!(&mut state1);
|
||||
|
||||
if RELOAD_2 {
|
||||
/* This test must be static */
|
||||
fse_reload_dstream(&mut bit_d);
|
||||
}
|
||||
|
||||
*op.add(1) = fse_getsymbol!(&mut state2);
|
||||
|
||||
if RELOAD_4 {
|
||||
/* This test must be static */
|
||||
if fse_reload_dstream(&mut bit_d) > FSE_DSTREAM_UNFINISHED {
|
||||
op = op.add(2);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
*op.add(2) = fse_getsymbol!(&mut state1);
|
||||
|
||||
if RELOAD_2 {
|
||||
/* This test must be static */
|
||||
fse_reload_dstream(&mut bit_d);
|
||||
}
|
||||
|
||||
*op.add(3) = fse_getsymbol!(&mut state2);
|
||||
op = op.add(4);
|
||||
}
|
||||
|
||||
/* tail */
|
||||
loop {
|
||||
if fse_reload_dstream(&mut bit_d) > FSE_DSTREAM_COMPLETED
|
||||
|| (op as usize) == omax_addr
|
||||
|| (fse_end_of_dstream(&bit_d) && (fast || fse_end_of_dstate(&state1)))
|
||||
{
|
||||
break;
|
||||
}
|
||||
|
||||
*op = fse_getsymbol!(&mut state1);
|
||||
op = op.add(1);
|
||||
|
||||
if fse_reload_dstream(&mut bit_d) > FSE_DSTREAM_COMPLETED
|
||||
|| (op as usize) == omax_addr
|
||||
|| (fse_end_of_dstream(&bit_d) && (fast || fse_end_of_dstate(&state2)))
|
||||
{
|
||||
break;
|
||||
}
|
||||
|
||||
*op = fse_getsymbol!(&mut state2);
|
||||
op = op.add(1);
|
||||
}
|
||||
|
||||
/* end ? */
|
||||
if fse_end_of_dstream(&bit_d) && fse_end_of_dstate(&state1) && fse_end_of_dstate(&state2) {
|
||||
return (op as usize) - (ostart as usize);
|
||||
}
|
||||
|
||||
if (op as usize) == omax_addr {
|
||||
/* dst buffer is full, but cSrc unfinished */
|
||||
return fse_error(FSE_ERROR_DST_SIZE_TOO_SMALL);
|
||||
}
|
||||
|
||||
fse_error(FSE_ERROR_CORRUPTION_DETECTED)
|
||||
}
|
||||
|
||||
unsafe fn fse_decompress_using_dtable(
|
||||
dst: *mut u8,
|
||||
original_size: usize,
|
||||
c_src: *const u8,
|
||||
c_src_size: usize,
|
||||
dt: *const u32,
|
||||
) -> usize {
|
||||
let fast_mode = (*(dt as *const FseDTableHeader)).fast_mode;
|
||||
|
||||
/* select fast mode (static) */
|
||||
if fast_mode != 0 {
|
||||
return fse_decompress_using_dtable_generic(
|
||||
dst,
|
||||
original_size,
|
||||
c_src,
|
||||
c_src_size,
|
||||
dt,
|
||||
true,
|
||||
);
|
||||
}
|
||||
fse_decompress_using_dtable_generic(dst, original_size, c_src, c_src_size, dt, false)
|
||||
}
|
||||
|
||||
unsafe fn fse_decompress(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
c_src: *const u8,
|
||||
c_src_size: usize,
|
||||
) -> usize {
|
||||
let istart = c_src;
|
||||
let mut ip = istart;
|
||||
let mut counting = [0i16; (FSE_MAX_SYMBOL_VALUE + 1) as usize];
|
||||
let mut dt = [0u32; 1 + (1 << FSE_MAX_TABLELOG)]; /* DTable_max_t */
|
||||
let mut table_log: u32 = 0;
|
||||
let mut max_symbol_value: u32 = FSE_MAX_SYMBOL_VALUE;
|
||||
let mut remaining_size = c_src_size;
|
||||
|
||||
if c_src_size < 2 {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG); /* too small input size */
|
||||
}
|
||||
|
||||
/* normal FSE decoding mode */
|
||||
let error_code = fse_read_ncount(
|
||||
counting.as_mut_ptr(),
|
||||
&mut max_symbol_value,
|
||||
&mut table_log,
|
||||
istart,
|
||||
c_src_size,
|
||||
);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
if error_code >= c_src_size {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG); /* too small input size */
|
||||
}
|
||||
ip = ip.add(error_code);
|
||||
remaining_size -= error_code;
|
||||
|
||||
let error_code = fse_build_dtable(
|
||||
dt.as_mut_ptr(),
|
||||
counting.as_ptr(),
|
||||
max_symbol_value,
|
||||
table_log,
|
||||
);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
|
||||
/* always return, even if it is an error code */
|
||||
fse_decompress_using_dtable(dst, max_dst_size, ip, remaining_size, dt.as_ptr())
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* Huff0 : Huffman block decompression
|
||||
********************************************/
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy)]
|
||||
struct HufDElt {
|
||||
byte: u8,
|
||||
nb_bits: u8,
|
||||
}
|
||||
|
||||
/* Loop shapes and arithmetic below intentionally mirror the frozen C. */
|
||||
#[allow(clippy::needless_range_loop, clippy::manual_div_ceil)]
|
||||
unsafe fn huf_read_dtable(dtable: *mut u16, src: *const u8, src_size: usize) -> usize {
|
||||
let mut huff_weight = [0u8; (HUF_MAX_SYMBOL_VALUE + 1) as usize];
|
||||
let mut rank_val = [0u32; (HUF_ABSOLUTEMAX_TABLELOG + 1) as usize];
|
||||
let ip = src;
|
||||
let dt = dtable.add(1) as *mut HufDElt;
|
||||
|
||||
if src_size == 0 {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
let mut i_size = *ip as usize;
|
||||
|
||||
let o_size: usize;
|
||||
if i_size >= 128 {
|
||||
/* special header */
|
||||
if i_size >= 242 {
|
||||
/* RLE */
|
||||
const L: [usize; 14] = [1, 2, 3, 4, 7, 8, 15, 16, 31, 32, 63, 64, 127, 128];
|
||||
o_size = L[i_size - 242];
|
||||
huff_weight = [1u8; (HUF_MAX_SYMBOL_VALUE + 1) as usize];
|
||||
i_size = 0;
|
||||
} else {
|
||||
/* Incompressible */
|
||||
o_size = i_size - 127;
|
||||
i_size = (o_size + 1) / 2;
|
||||
if i_size + 1 > src_size {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
let ip = ip.add(1);
|
||||
let mut n = 0usize;
|
||||
while n < o_size {
|
||||
huff_weight[n] = *ip.add(n / 2) >> 4;
|
||||
huff_weight[n + 1] = *ip.add(n / 2) & 15;
|
||||
n += 2;
|
||||
}
|
||||
}
|
||||
} else {
|
||||
/* header compressed with FSE (normal case) */
|
||||
if i_size + 1 > src_size {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
/* max 255 values decoded, last one is implied */
|
||||
let decoded = fse_decompress(
|
||||
huff_weight.as_mut_ptr(),
|
||||
HUF_MAX_SYMBOL_VALUE as usize,
|
||||
ip.add(1),
|
||||
i_size,
|
||||
);
|
||||
if fse_is_error(decoded) {
|
||||
return decoded;
|
||||
}
|
||||
o_size = decoded;
|
||||
}
|
||||
|
||||
/* collect weight stats */
|
||||
let mut weight_total: u32 = 0;
|
||||
for n in 0..o_size {
|
||||
if huff_weight[n] >= HUF_ABSOLUTEMAX_TABLELOG as u8 {
|
||||
return fse_error(FSE_ERROR_CORRUPTION_DETECTED);
|
||||
}
|
||||
rank_val[huff_weight[n] as usize] += 1;
|
||||
weight_total += (1u32 << huff_weight[n]) >> 1;
|
||||
}
|
||||
if weight_total == 0 {
|
||||
return fse_error(FSE_ERROR_CORRUPTION_DETECTED);
|
||||
}
|
||||
|
||||
/* get last non-null symbol weight (implied, total must be 2^n) */
|
||||
let max_bits = fse_highbit32(weight_total) + 1;
|
||||
if max_bits > *dtable as u32 {
|
||||
return fse_error(FSE_ERROR_TABLELOG_TOO_LARGE); /* DTable is too small */
|
||||
}
|
||||
*dtable = max_bits as u16;
|
||||
{
|
||||
let total = 1u32 << max_bits;
|
||||
let rest = total - weight_total;
|
||||
let verif = 1u32 << fse_highbit32(rest);
|
||||
let last_weight = fse_highbit32(rest) + 1;
|
||||
if verif != rest {
|
||||
return fse_error(FSE_ERROR_CORRUPTION_DETECTED); /* last value must be a clean power of 2 */
|
||||
}
|
||||
huff_weight[o_size] = last_weight as u8;
|
||||
rank_val[last_weight as usize] += 1;
|
||||
}
|
||||
|
||||
/* check tree construction validity */
|
||||
if rank_val[1] < 2 || (rank_val[1] & 1) != 0 {
|
||||
/* by construction : at least 2 elts of rank 1, must be even */
|
||||
return fse_error(FSE_ERROR_CORRUPTION_DETECTED);
|
||||
}
|
||||
|
||||
/* Prepare ranks */
|
||||
let mut next_rank_start: u32 = 0;
|
||||
for n in 1..=(max_bits as usize) {
|
||||
let current = next_rank_start;
|
||||
next_rank_start += rank_val[n] << (n - 1);
|
||||
rank_val[n] = current;
|
||||
}
|
||||
|
||||
/* fill DTable */
|
||||
for n in 0..=o_size {
|
||||
let w = huff_weight[n] as usize;
|
||||
let length = (1u32 << w) >> 1;
|
||||
let d = HufDElt {
|
||||
byte: n as u8,
|
||||
nb_bits: (max_bits + 1 - w as u32) as u8,
|
||||
};
|
||||
for i in rank_val[w]..(rank_val[w] + length) {
|
||||
*dt.add(i as usize) = d;
|
||||
}
|
||||
rank_val[w] += length;
|
||||
}
|
||||
|
||||
i_size + 1
|
||||
}
|
||||
|
||||
unsafe fn huf_decode_symbol(d_stream: &mut FseDStream, dt: *const HufDElt, dt_log: u32) -> u8 {
|
||||
let val = fse_look_bits_fast(d_stream, dt_log); /* note : dtLog >= 1 */
|
||||
let entry = *dt.add(val);
|
||||
fse_skip_bits(d_stream, entry.nb_bits as u32);
|
||||
entry.byte
|
||||
}
|
||||
|
||||
/* The C decode macros: SYMBOL_1 reloads on 32-bit only when
|
||||
* HUF_MAX_TABLELOG > 12 (never here); SYMBOL_2 reloads on 32-bit. */
|
||||
#[inline]
|
||||
unsafe fn huf_decode_symbol_1(
|
||||
op: *mut u8,
|
||||
d_stream: &mut FseDStream,
|
||||
dt: *const HufDElt,
|
||||
dt_log: u32,
|
||||
) {
|
||||
*op = huf_decode_symbol(d_stream, dt, dt_log);
|
||||
if IS_32BITS && HUF_MAX_TABLELOG > 12 {
|
||||
fse_reload_dstream(d_stream);
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn huf_decode_symbol_2(
|
||||
op: *mut u8,
|
||||
d_stream: &mut FseDStream,
|
||||
dt: *const HufDElt,
|
||||
dt_log: u32,
|
||||
) {
|
||||
*op = huf_decode_symbol(d_stream, dt, dt_log);
|
||||
if IS_32BITS {
|
||||
fse_reload_dstream(d_stream);
|
||||
}
|
||||
}
|
||||
|
||||
unsafe fn huf_decompress_using_dtable(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
c_src: *const u8,
|
||||
c_src_size: usize,
|
||||
dtable: *const u16,
|
||||
) -> usize {
|
||||
if c_src_size < 6 {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
|
||||
let ostart = dst;
|
||||
let mut op = ostart;
|
||||
let omax_addr = (op as usize).wrapping_add(max_dst_size);
|
||||
let olimit_addr = if max_dst_size < 15 {
|
||||
op as usize
|
||||
} else {
|
||||
omax_addr - 15
|
||||
};
|
||||
|
||||
let dt = dtable.add(1) as *const HufDElt;
|
||||
let dt_log = *dtable as u32;
|
||||
|
||||
/* Init */
|
||||
let length1 = fse_read_le16(c_src) as usize;
|
||||
let length2 = fse_read_le16(c_src.add(2)) as usize;
|
||||
let length3 = fse_read_le16(c_src.add(4)) as usize;
|
||||
let length4 = c_src_size
|
||||
.wrapping_sub(6)
|
||||
.wrapping_sub(length1)
|
||||
.wrapping_sub(length2)
|
||||
.wrapping_sub(length3); /* check coherency !! */
|
||||
let start1 = c_src.add(6);
|
||||
let start2 = start1.wrapping_add(length1);
|
||||
let start3 = start2.wrapping_add(length2);
|
||||
let start4 = start3.wrapping_add(length3);
|
||||
|
||||
if length1 + length2 + length3 + 6 >= c_src_size {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
|
||||
let mut bit_d1 = FseDStream {
|
||||
bit_container: 0,
|
||||
bits_consumed: 0,
|
||||
ptr: ptr::null(),
|
||||
start: ptr::null(),
|
||||
};
|
||||
let mut bit_d2 = FseDStream {
|
||||
bit_container: 0,
|
||||
bits_consumed: 0,
|
||||
ptr: ptr::null(),
|
||||
start: ptr::null(),
|
||||
};
|
||||
let mut bit_d3 = FseDStream {
|
||||
bit_container: 0,
|
||||
bits_consumed: 0,
|
||||
ptr: ptr::null(),
|
||||
start: ptr::null(),
|
||||
};
|
||||
let mut bit_d4 = FseDStream {
|
||||
bit_container: 0,
|
||||
bits_consumed: 0,
|
||||
ptr: ptr::null(),
|
||||
start: ptr::null(),
|
||||
};
|
||||
|
||||
let error_code = fse_init_dstream(&mut bit_d1, start1, length1);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
let error_code = fse_init_dstream(&mut bit_d2, start2, length2);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
let error_code = fse_init_dstream(&mut bit_d3, start3, length3);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
let error_code = fse_init_dstream(&mut bit_d4, start4, length4);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
|
||||
let mut reload_status = fse_reload_dstream(&mut bit_d2);
|
||||
|
||||
/* 16 symbols per loop; D2-3-4 are supposed to be synchronized and finish together */
|
||||
while reload_status < FSE_DSTREAM_COMPLETED && (op as usize) < olimit_addr {
|
||||
huf_decode_symbol_1(op, &mut bit_d1, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(1), &mut bit_d2, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(2), &mut bit_d3, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(3), &mut bit_d4, dt, dt_log);
|
||||
huf_decode_symbol_2(op.add(4), &mut bit_d1, dt, dt_log);
|
||||
huf_decode_symbol_2(op.add(5), &mut bit_d2, dt, dt_log);
|
||||
huf_decode_symbol_2(op.add(6), &mut bit_d3, dt, dt_log);
|
||||
huf_decode_symbol_2(op.add(7), &mut bit_d4, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(8), &mut bit_d1, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(9), &mut bit_d2, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(10), &mut bit_d3, dt, dt_log);
|
||||
huf_decode_symbol_1(op.add(11), &mut bit_d4, dt, dt_log);
|
||||
*op.add(12) = huf_decode_symbol(&mut bit_d1, dt, dt_log);
|
||||
*op.add(13) = huf_decode_symbol(&mut bit_d2, dt, dt_log);
|
||||
*op.add(14) = huf_decode_symbol(&mut bit_d3, dt, dt_log);
|
||||
*op.add(15) = huf_decode_symbol(&mut bit_d4, dt, dt_log);
|
||||
|
||||
op = op.add(16);
|
||||
reload_status = fse_reload_dstream(&mut bit_d2)
|
||||
| fse_reload_dstream(&mut bit_d3)
|
||||
| fse_reload_dstream(&mut bit_d4);
|
||||
fse_reload_dstream(&mut bit_d1);
|
||||
}
|
||||
|
||||
if reload_status != FSE_DSTREAM_COMPLETED {
|
||||
/* not complete : some bitStream might be FSE_DStream_unfinished */
|
||||
return fse_error(FSE_ERROR_CORRUPTION_DETECTED);
|
||||
}
|
||||
|
||||
/* tail */
|
||||
{
|
||||
let mut bit_tail = FseDStream {
|
||||
bit_container: bit_d1.bit_container, /* required in case of FSE_DStream_endOfBuffer */
|
||||
bits_consumed: bit_d1.bits_consumed,
|
||||
ptr: bit_d1.ptr,
|
||||
start: start1,
|
||||
};
|
||||
while fse_reload_dstream(&mut bit_tail) < FSE_DSTREAM_COMPLETED && (op as usize) < omax_addr
|
||||
{
|
||||
*op = huf_decode_symbol(&mut bit_tail, dt, dt_log);
|
||||
op = op.add(1);
|
||||
}
|
||||
|
||||
if fse_end_of_dstream(&bit_tail) {
|
||||
return (op as usize) - (ostart as usize);
|
||||
}
|
||||
}
|
||||
|
||||
if (op as usize) == omax_addr {
|
||||
return fse_error(FSE_ERROR_DST_SIZE_TOO_SMALL); /* dst buffer is full, but cSrc unfinished */
|
||||
}
|
||||
|
||||
fse_error(FSE_ERROR_CORRUPTION_DETECTED)
|
||||
}
|
||||
|
||||
unsafe fn huf_decompress(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
c_src: *const u8,
|
||||
c_src_size: usize,
|
||||
) -> usize {
|
||||
/* HUF_CREATE_STATIC_DTABLE(DTable, HUF_MAX_TABLELOG) */
|
||||
let mut dtable = [0u16; 1 + (1 << HUF_MAX_TABLELOG)];
|
||||
dtable[0] = HUF_MAX_TABLELOG as u16;
|
||||
let mut ip = c_src;
|
||||
|
||||
let error_code = huf_read_dtable(dtable.as_mut_ptr(), c_src, c_src_size);
|
||||
if fse_is_error(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
if error_code >= c_src_size {
|
||||
return fse_error(FSE_ERROR_SRC_SIZE_WRONG);
|
||||
}
|
||||
ip = ip.add(error_code);
|
||||
|
||||
huf_decompress_using_dtable(
|
||||
dst,
|
||||
max_dst_size,
|
||||
ip,
|
||||
c_src_size - error_code,
|
||||
dtable.as_ptr(),
|
||||
)
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* zstd v0.1 frame decoding
|
||||
********************************************/
|
||||
const ZSTD_MAGIC_NUMBER: u32 = 0xFD2FB51E; /* 3rd version : seqNb header */
|
||||
|
||||
const KB: usize = 1 << 10;
|
||||
const BLOCKSIZE: usize = 128 * KB; /* define, for static allocation */
|
||||
|
||||
const MINMATCH: usize = 4;
|
||||
const MLBITS: u32 = 7;
|
||||
const LLBITS: u32 = 6;
|
||||
const OFFBITS: u32 = 5;
|
||||
const MAX_ML: u32 = (1 << MLBITS) - 1;
|
||||
const MAX_LL: u32 = (1 << LLBITS) - 1;
|
||||
const MAX_OFF: u32 = (1 << OFFBITS) - 1;
|
||||
#[allow(dead_code)] /* part of the frozen v0.1 constant set; used only by the compressor */
|
||||
const LIT_FSE_LOG: u32 = 11;
|
||||
const ML_FSE_LOG: u32 = 10;
|
||||
const LL_FSE_LOG: u32 = 10;
|
||||
const OFF_FSE_LOG: u32 = 9;
|
||||
|
||||
const ZSTD_CONTENTSIZE_ERROR: u64 = 0u64.wrapping_sub(2);
|
||||
|
||||
const ZSTD_BLOCK_HEADER_SIZE: usize = 3;
|
||||
const ZSTD_FRAME_HEADER_SIZE: usize = 4;
|
||||
|
||||
/* FSE_DTABLE_SIZE_U32(maxTableLog) == 1 + (1 << maxTableLog) */
|
||||
const LL_DTABLE_SIZE_U32: usize = 1 + (1 << LL_FSE_LOG as usize);
|
||||
const OFF_DTABLE_SIZE_U32: usize = 1 + (1 << OFF_FSE_LOG as usize);
|
||||
const ML_DTABLE_SIZE_U32: usize = 1 + (1 << ML_FSE_LOG as usize);
|
||||
|
||||
#[inline]
|
||||
unsafe fn zstd_copy4(dst: *mut u8, src: *const u8) {
|
||||
ptr::copy_nonoverlapping(src, dst, 4);
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn zstd_copy8(dst: *mut u8, src: *const u8) {
|
||||
ptr::copy_nonoverlapping(src, dst, 8);
|
||||
}
|
||||
|
||||
unsafe fn zstd_wildcopy(dst: *mut u8, src: *const u8, length: isize) {
|
||||
let mut ip = src;
|
||||
let mut op = dst;
|
||||
let oend_addr = (op as usize).wrapping_add(length as usize);
|
||||
while (op as usize) < oend_addr {
|
||||
zstd_copy8(op, ip);
|
||||
op = op.add(8);
|
||||
ip = ip.add(8);
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn zstd_read_le16(mem_ptr: *const u8) -> u16 {
|
||||
u16::from_le_bytes(ptr::read_unaligned(mem_ptr as *const [u8; 2]))
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn zstd_read_le24(mem_ptr: *const u8) -> u32 {
|
||||
zstd_read_le16(mem_ptr) as u32 + ((*mem_ptr.add(2) as u32) << 16)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
unsafe fn zstd_read_be32(mem_ptr: *const u8) -> u32 {
|
||||
u32::from_be_bytes(ptr::read_unaligned(mem_ptr as *const [u8; 4]))
|
||||
}
|
||||
|
||||
/* blockType_t */
|
||||
const BT_COMPRESSED: u32 = 0;
|
||||
const BT_RAW: u32 = 1;
|
||||
const BT_RLE: u32 = 2;
|
||||
const BT_END: u32 = 3;
|
||||
|
||||
struct BlockProperties {
|
||||
block_type: u32,
|
||||
orig_size: u32,
|
||||
}
|
||||
|
||||
/// The v0.1 streaming decompression context (`dctx_t` in C). The struct is
|
||||
/// opaque to C: `zstd_v01.h` only forward-declares it, so the definition now
|
||||
/// lives here.
|
||||
#[repr(C)]
|
||||
pub struct ZSTDv01_Dctx {
|
||||
ll_table: [u32; LL_DTABLE_SIZE_U32],
|
||||
off_table: [u32; OFF_DTABLE_SIZE_U32],
|
||||
ml_table: [u32; ML_DTABLE_SIZE_U32],
|
||||
previous_dst_end: *const u8,
|
||||
base: *const u8,
|
||||
expected: usize,
|
||||
b_type: u32,
|
||||
phase: u32,
|
||||
}
|
||||
|
||||
unsafe fn zstdv01_getc_block_size(
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
bp_ptr: &mut BlockProperties,
|
||||
) -> usize {
|
||||
if src_size < 3 {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
|
||||
let header_flags = *src;
|
||||
let c_size =
|
||||
(*src.add(2) as u32) + ((*src.add(1) as u32) << 8) + (((header_flags & 7) as u32) << 16);
|
||||
|
||||
bp_ptr.block_type = (header_flags >> 6) as u32;
|
||||
bp_ptr.orig_size = if bp_ptr.block_type == BT_RLE {
|
||||
c_size
|
||||
} else {
|
||||
0
|
||||
};
|
||||
|
||||
if bp_ptr.block_type == BT_END {
|
||||
return 0;
|
||||
}
|
||||
if bp_ptr.block_type == BT_RLE {
|
||||
return 1;
|
||||
}
|
||||
c_size as usize
|
||||
}
|
||||
|
||||
unsafe fn zstd_copy_uncompressed_block(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
if src_size > max_dst_size {
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
if src_size > 0 {
|
||||
ptr::copy_nonoverlapping(src, dst, src_size);
|
||||
}
|
||||
src_size
|
||||
}
|
||||
|
||||
unsafe fn zstd_decompress_literals(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
let oend = dst.add(max_dst_size);
|
||||
let ip = src;
|
||||
|
||||
/* check : minimum 2, for litSize, +1, for content */
|
||||
if src_size <= 3 {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
|
||||
let mut lit_size = (*ip.add(1) as usize) + ((*ip as usize) << 8);
|
||||
lit_size += (((*ip.offset(-3) as usize) >> 3) & 7) << 16; /* mmmmh.... */
|
||||
|
||||
if lit_size > max_dst_size {
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
let op = oend.sub(lit_size);
|
||||
let error_code = huf_decompress(op, lit_size, ip.add(2), src_size - 2);
|
||||
if fse_is_error(error_code) {
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
lit_size
|
||||
}
|
||||
|
||||
unsafe fn zstdv01_decode_literals_block(
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
lit_start: &mut *const u8,
|
||||
lit_size: &mut usize,
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
let istart = src;
|
||||
let mut ip = istart;
|
||||
let ostart = dst;
|
||||
let oend = ostart.add(max_dst_size);
|
||||
let mut litbp = BlockProperties {
|
||||
block_type: 0,
|
||||
orig_size: 0,
|
||||
};
|
||||
|
||||
let litc_size = zstdv01_getc_block_size(src, src_size, &mut litbp);
|
||||
if ERR_isError(litc_size) {
|
||||
return litc_size;
|
||||
}
|
||||
if litc_size > src_size - ZSTD_BLOCK_HEADER_SIZE {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
ip = ip.add(ZSTD_BLOCK_HEADER_SIZE);
|
||||
|
||||
match litbp.block_type {
|
||||
BT_RAW => {
|
||||
*lit_start = ip;
|
||||
ip = ip.add(litc_size);
|
||||
*lit_size = litc_size;
|
||||
}
|
||||
BT_RLE => {
|
||||
let rle_size = litbp.orig_size as usize;
|
||||
if rle_size > max_dst_size {
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
if src_size == 0 {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
if rle_size > 0 {
|
||||
ptr::write_bytes(oend.sub(rle_size), *ip, rle_size);
|
||||
}
|
||||
*lit_start = oend.sub(rle_size);
|
||||
*lit_size = rle_size;
|
||||
ip = ip.add(1);
|
||||
}
|
||||
BT_COMPRESSED => {
|
||||
let decoded_lit_size = zstd_decompress_literals(dst, max_dst_size, ip, litc_size);
|
||||
if ERR_isError(decoded_lit_size) {
|
||||
return decoded_lit_size;
|
||||
}
|
||||
*lit_start = oend.sub(decoded_lit_size);
|
||||
*lit_size = decoded_lit_size;
|
||||
ip = ip.add(litc_size);
|
||||
}
|
||||
_ => {
|
||||
/* bt_end and impossible values */
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
}
|
||||
|
||||
(ip as usize) - (istart as usize)
|
||||
}
|
||||
|
||||
#[allow(clippy::too_many_arguments)]
|
||||
unsafe fn zstdv01_decode_seq_headers(
|
||||
nb_seq: &mut i32,
|
||||
dumps_ptr: &mut *const u8,
|
||||
dumps_length_ptr: &mut usize,
|
||||
dtable_ll: *mut u32,
|
||||
dtable_ml: *mut u32,
|
||||
dtable_offb: *mut u32,
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
let istart = src;
|
||||
let mut ip = istart;
|
||||
let iend_addr = (istart as usize).wrapping_add(src_size);
|
||||
|
||||
/* check */
|
||||
if src_size < 5 {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
|
||||
/* SeqHead */
|
||||
*nb_seq = zstd_read_le16(ip) as i32;
|
||||
ip = ip.add(2);
|
||||
let ll_type = (*ip >> 6) as u32;
|
||||
let off_type = ((*ip >> 4) & 3) as u32;
|
||||
let ml_type = ((*ip >> 2) & 3) as u32;
|
||||
let dumps_length: usize;
|
||||
if (*ip & 2) != 0 {
|
||||
dumps_length = (*ip.add(2) as usize) + ((*ip.add(1) as usize) << 8);
|
||||
ip = ip.add(3);
|
||||
} else {
|
||||
dumps_length = (*ip.add(1) as usize) + (((*ip as usize) & 1) << 8);
|
||||
ip = ip.add(2);
|
||||
}
|
||||
*dumps_ptr = ip;
|
||||
ip = ip.wrapping_add(dumps_length);
|
||||
*dumps_length_ptr = dumps_length;
|
||||
|
||||
/* check */
|
||||
if (ip as usize) > iend_addr.wrapping_sub(3) {
|
||||
/* min : all 3 are "raw", hence no header, but at least xxLog bits per type */
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
|
||||
/* sequences */
|
||||
{
|
||||
let mut norm = [0i16; (MAX_ML + 1) as usize]; /* assumption : MaxML >= MaxLL and MaxOff */
|
||||
|
||||
/* Build DTables */
|
||||
match ll_type {
|
||||
BT_RLE => {
|
||||
fse_build_dtable_rle(dtable_ll, *ip);
|
||||
ip = ip.add(1);
|
||||
}
|
||||
BT_RAW => {
|
||||
fse_build_dtable_raw(dtable_ll, LLBITS);
|
||||
}
|
||||
_ => {
|
||||
let mut max = MAX_LL;
|
||||
let mut ll_log = 0u32;
|
||||
let header_size = fse_read_ncount(
|
||||
norm.as_mut_ptr(),
|
||||
&mut max,
|
||||
&mut ll_log,
|
||||
ip,
|
||||
iend_addr - ip as usize,
|
||||
);
|
||||
if fse_is_error(header_size) {
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
if ll_log > LL_FSE_LOG {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
ip = ip.add(header_size);
|
||||
fse_build_dtable(dtable_ll, norm.as_ptr(), max, ll_log);
|
||||
}
|
||||
}
|
||||
|
||||
match off_type {
|
||||
BT_RLE => {
|
||||
if (ip as usize) > iend_addr.wrapping_sub(2) {
|
||||
/* min : "raw", hence no header, but at least xxLog bits */
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
fse_build_dtable_rle(dtable_offb, *ip);
|
||||
ip = ip.add(1);
|
||||
}
|
||||
BT_RAW => {
|
||||
fse_build_dtable_raw(dtable_offb, OFFBITS);
|
||||
}
|
||||
_ => {
|
||||
let mut max = MAX_OFF;
|
||||
let mut off_log = 0u32;
|
||||
let header_size = fse_read_ncount(
|
||||
norm.as_mut_ptr(),
|
||||
&mut max,
|
||||
&mut off_log,
|
||||
ip,
|
||||
iend_addr - ip as usize,
|
||||
);
|
||||
if fse_is_error(header_size) {
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
if off_log > OFF_FSE_LOG {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
ip = ip.add(header_size);
|
||||
fse_build_dtable(dtable_offb, norm.as_ptr(), max, off_log);
|
||||
}
|
||||
}
|
||||
|
||||
match ml_type {
|
||||
BT_RLE => {
|
||||
if (ip as usize) > iend_addr.wrapping_sub(2) {
|
||||
/* min : "raw", hence no header, but at least xxLog bits */
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
fse_build_dtable_rle(dtable_ml, *ip);
|
||||
ip = ip.add(1);
|
||||
}
|
||||
BT_RAW => {
|
||||
fse_build_dtable_raw(dtable_ml, MLBITS);
|
||||
}
|
||||
_ => {
|
||||
let mut max = MAX_ML;
|
||||
let mut ml_log = 0u32;
|
||||
let header_size = fse_read_ncount(
|
||||
norm.as_mut_ptr(),
|
||||
&mut max,
|
||||
&mut ml_log,
|
||||
ip,
|
||||
iend_addr - ip as usize,
|
||||
);
|
||||
if fse_is_error(header_size) {
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
if ml_log > ML_FSE_LOG {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
ip = ip.add(header_size);
|
||||
fse_build_dtable(dtable_ml, norm.as_ptr(), max, ml_log);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
(ip as usize) - (istart as usize)
|
||||
}
|
||||
|
||||
#[derive(Clone, Copy)]
|
||||
struct Seq {
|
||||
lit_length: usize,
|
||||
offset: usize,
|
||||
match_length: usize,
|
||||
}
|
||||
|
||||
struct SeqState {
|
||||
d_stream: FseDStream,
|
||||
state_ll: FseDState,
|
||||
state_offb: FseDState,
|
||||
state_ml: FseDState,
|
||||
prev_offset: usize,
|
||||
dumps: *const u8,
|
||||
dumps_end: *const u8,
|
||||
}
|
||||
|
||||
unsafe fn zstd_decode_sequence(seq: &mut Seq, seq_state: &mut SeqState) {
|
||||
let mut dumps = seq_state.dumps;
|
||||
let de = seq_state.dumps_end;
|
||||
|
||||
/* Literal length */
|
||||
let mut lit_length =
|
||||
fse_decode_symbol(&mut seq_state.state_ll, &mut seq_state.d_stream) as usize;
|
||||
let prev_offset = if lit_length != 0 {
|
||||
seq.offset
|
||||
} else {
|
||||
seq_state.prev_offset
|
||||
};
|
||||
seq_state.prev_offset = seq.offset;
|
||||
if lit_length == MAX_LL as usize {
|
||||
let add = if (dumps as usize) < (de as usize) {
|
||||
let v = *dumps as u32;
|
||||
dumps = dumps.add(1);
|
||||
v
|
||||
} else {
|
||||
0
|
||||
};
|
||||
if add < 255 {
|
||||
lit_length += add as usize;
|
||||
} else if (dumps as usize) <= (de as usize).wrapping_sub(3) {
|
||||
lit_length = zstd_read_le24(dumps) as usize;
|
||||
dumps = dumps.add(3);
|
||||
}
|
||||
}
|
||||
|
||||
/* Offset */
|
||||
let mut offset: usize;
|
||||
{
|
||||
let offset_code =
|
||||
fse_decode_symbol(&mut seq_state.state_offb, &mut seq_state.d_stream) as u32;
|
||||
if IS_32BITS {
|
||||
fse_reload_dstream(&mut seq_state.d_stream);
|
||||
}
|
||||
let mut nb_bits = offset_code.wrapping_sub(1);
|
||||
if offset_code == 0 {
|
||||
nb_bits = 0; /* cmove */
|
||||
}
|
||||
offset = (1usize << (nb_bits & (USIZE_BITS - 1)))
|
||||
.wrapping_add(fse_read_bits(&mut seq_state.d_stream, nb_bits));
|
||||
if IS_32BITS {
|
||||
fse_reload_dstream(&mut seq_state.d_stream);
|
||||
}
|
||||
if offset_code == 0 {
|
||||
offset = prev_offset;
|
||||
}
|
||||
}
|
||||
|
||||
/* MatchLength */
|
||||
let mut match_length =
|
||||
fse_decode_symbol(&mut seq_state.state_ml, &mut seq_state.d_stream) as usize;
|
||||
if match_length == MAX_ML as usize {
|
||||
let add = if (dumps as usize) < (de as usize) {
|
||||
let v = *dumps as u32;
|
||||
dumps = dumps.add(1);
|
||||
v
|
||||
} else {
|
||||
0
|
||||
};
|
||||
if add < 255 {
|
||||
match_length += add as usize;
|
||||
} else if (dumps as usize) <= (de as usize).wrapping_sub(3) {
|
||||
match_length = zstd_read_le24(dumps) as usize;
|
||||
dumps = dumps.add(3);
|
||||
}
|
||||
}
|
||||
match_length += MINMATCH;
|
||||
|
||||
/* save result */
|
||||
seq.lit_length = lit_length;
|
||||
seq.offset = offset;
|
||||
seq.match_length = match_length;
|
||||
seq_state.dumps = dumps;
|
||||
}
|
||||
|
||||
unsafe fn zstd_exec_sequence(
|
||||
op: *mut u8,
|
||||
sequence: Seq,
|
||||
lit_ptr: &mut *const u8,
|
||||
lit_limit: *const u8,
|
||||
base: *const u8,
|
||||
oend: *mut u8,
|
||||
) -> usize {
|
||||
static DEC32TABLE: [usize; 8] = [0, 1, 2, 1, 4, 4, 4, 4]; /* added */
|
||||
static DEC64TABLE: [usize; 8] = [8, 8, 8, 7, 8, 9, 10, 11]; /* subtracted */
|
||||
let ostart = op;
|
||||
let mut op = op;
|
||||
let o_lit_end = op.wrapping_add(sequence.lit_length);
|
||||
let lit_length = sequence.lit_length;
|
||||
/* risk : address space overflow (32-bits) */
|
||||
let end_match = op
|
||||
.wrapping_add(lit_length)
|
||||
.wrapping_add(sequence.match_length);
|
||||
let lit_end = (*lit_ptr).wrapping_add(lit_length);
|
||||
|
||||
/* checks */
|
||||
let seq_length = sequence.lit_length.wrapping_add(sequence.match_length);
|
||||
|
||||
if seq_length > (oend as usize).wrapping_sub(op as usize) {
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
if sequence.lit_length > (lit_limit as usize).wrapping_sub(*lit_ptr as usize) {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
/* Now we know there are no overflow in literal nor match lengths, can use pointer checks */
|
||||
if sequence.offset > ((o_lit_end as usize).wrapping_sub(base as usize)) as u32 as usize {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
|
||||
if (end_match as usize) > (oend as usize) {
|
||||
/* overwrite beyond dst buffer */
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
if (lit_end as usize) > (lit_limit as usize) {
|
||||
/* overRead beyond lit buffer */
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
if sequence.match_length > (*lit_ptr as usize).wrapping_sub(op as usize) {
|
||||
/* overwrite literal segment */
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
|
||||
/* copy Literals */
|
||||
/* note : v0.1 seems to allow scenarios where output or input are close to end of buffer */
|
||||
ptr::copy(*lit_ptr, op, sequence.lit_length);
|
||||
|
||||
op = op.add(lit_length);
|
||||
*lit_ptr = lit_end; /* update for next sequence */
|
||||
|
||||
/* check : last match must be at a minimum distance of 8 from end of dest buffer */
|
||||
if (oend as usize) - (op as usize) < 8 {
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
|
||||
/* copy Match */
|
||||
{
|
||||
let overlap_risk = (lit_end as usize).wrapping_sub(end_match as usize) < 12;
|
||||
/* possible underflow at op - offset ? */
|
||||
let mut match_ptr = (op as usize).wrapping_sub(sequence.offset) as *const u8;
|
||||
let mut qutt: usize = 12;
|
||||
let mut saved = [0u8; 16]; /* U64 saved[2] */
|
||||
|
||||
/* check */
|
||||
if (match_ptr as usize) < (base as usize) {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
if sequence.offset > base as usize {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
|
||||
/* save beginning of literal sequence, in case of write overlap */
|
||||
if overlap_risk {
|
||||
if (end_match as usize).wrapping_add(qutt) > (oend as usize) {
|
||||
qutt = (oend as usize) - (end_match as usize);
|
||||
}
|
||||
ptr::copy_nonoverlapping(end_match as *const u8, saved.as_mut_ptr(), qutt);
|
||||
}
|
||||
|
||||
if sequence.offset < 8 {
|
||||
let dec64 = DEC64TABLE[sequence.offset];
|
||||
*op = *match_ptr;
|
||||
*op.add(1) = *match_ptr.add(1);
|
||||
*op.add(2) = *match_ptr.add(2);
|
||||
*op.add(3) = *match_ptr.add(3);
|
||||
match_ptr = match_ptr.add(DEC32TABLE[sequence.offset]);
|
||||
zstd_copy4(op.add(4), match_ptr);
|
||||
match_ptr = match_ptr.sub(dec64);
|
||||
} else {
|
||||
zstd_copy8(op, match_ptr);
|
||||
}
|
||||
op = op.add(8);
|
||||
match_ptr = match_ptr.add(8);
|
||||
|
||||
if (end_match as usize) > (oend as usize).wrapping_sub(16 - MINMATCH) {
|
||||
if (op as usize) < (oend as usize).wrapping_sub(8) {
|
||||
let dist = ((oend as usize) - 8) - (op as usize);
|
||||
zstd_wildcopy(op, match_ptr, dist as isize);
|
||||
match_ptr = match_ptr.add(dist);
|
||||
op = oend.sub(8);
|
||||
}
|
||||
while (op as usize) < (end_match as usize) {
|
||||
*op = *match_ptr;
|
||||
op = op.add(1);
|
||||
match_ptr = match_ptr.add(1);
|
||||
}
|
||||
} else {
|
||||
/* works even if matchLength < 8 */
|
||||
zstd_wildcopy(op, match_ptr, sequence.match_length as isize - 8);
|
||||
}
|
||||
|
||||
/* restore, in case of overlap */
|
||||
if overlap_risk {
|
||||
ptr::copy_nonoverlapping(saved.as_ptr(), end_match, qutt);
|
||||
}
|
||||
}
|
||||
|
||||
(end_match as usize) - (ostart as usize)
|
||||
}
|
||||
|
||||
unsafe fn zstd_decompress_sequences(
|
||||
ctx: *mut ZSTDv01_Dctx,
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
seq_start: *const u8,
|
||||
seq_size: usize,
|
||||
lit_start: *const u8,
|
||||
lit_size: usize,
|
||||
) -> usize {
|
||||
let dctx = ctx;
|
||||
let mut ip = seq_start;
|
||||
let iend_addr = (ip as usize).wrapping_add(seq_size);
|
||||
let ostart = dst;
|
||||
let mut op = ostart;
|
||||
let oend = ostart.add(max_dst_size);
|
||||
let mut lit_ptr = lit_start;
|
||||
let lit_end = lit_start.add(lit_size);
|
||||
let mut nb_seq: i32 = 0;
|
||||
let mut dumps: *const u8 = ptr::null();
|
||||
let mut dumps_length: usize = 0;
|
||||
let dtable_ll = (*dctx).ll_table.as_mut_ptr();
|
||||
let dtable_ml = (*dctx).ml_table.as_mut_ptr();
|
||||
let dtable_offb = (*dctx).off_table.as_mut_ptr();
|
||||
let base = (*dctx).base;
|
||||
|
||||
/* Build Decoding Tables */
|
||||
let error_code = zstdv01_decode_seq_headers(
|
||||
&mut nb_seq,
|
||||
&mut dumps,
|
||||
&mut dumps_length,
|
||||
dtable_ll,
|
||||
dtable_ml,
|
||||
dtable_offb,
|
||||
ip,
|
||||
iend_addr - ip as usize,
|
||||
);
|
||||
if ERR_isError(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
ip = ip.add(error_code);
|
||||
|
||||
/* Regen sequences */
|
||||
{
|
||||
let mut sequence = Seq {
|
||||
lit_length: 0,
|
||||
offset: 0,
|
||||
match_length: 0,
|
||||
};
|
||||
let mut seq_state = SeqState {
|
||||
d_stream: FseDStream {
|
||||
bit_container: 0,
|
||||
bits_consumed: 0,
|
||||
ptr: ptr::null(),
|
||||
start: ptr::null(),
|
||||
},
|
||||
state_ll: FseDState {
|
||||
state: 0,
|
||||
table: ptr::null(),
|
||||
},
|
||||
state_offb: FseDState {
|
||||
state: 0,
|
||||
table: ptr::null(),
|
||||
},
|
||||
state_ml: FseDState {
|
||||
state: 0,
|
||||
table: ptr::null(),
|
||||
},
|
||||
prev_offset: 1,
|
||||
dumps,
|
||||
dumps_end: dumps.wrapping_add(dumps_length),
|
||||
};
|
||||
let error_code = fse_init_dstream(&mut seq_state.d_stream, ip, iend_addr - ip as usize);
|
||||
if fse_is_error(error_code) {
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
fse_init_dstate(&mut seq_state.state_ll, &mut seq_state.d_stream, dtable_ll);
|
||||
fse_init_dstate(
|
||||
&mut seq_state.state_offb,
|
||||
&mut seq_state.d_stream,
|
||||
dtable_offb,
|
||||
);
|
||||
fse_init_dstate(&mut seq_state.state_ml, &mut seq_state.d_stream, dtable_ml);
|
||||
|
||||
while fse_reload_dstream(&mut seq_state.d_stream) <= FSE_DSTREAM_COMPLETED && nb_seq > 0 {
|
||||
nb_seq -= 1;
|
||||
zstd_decode_sequence(&mut sequence, &mut seq_state);
|
||||
let one_seq_size = zstd_exec_sequence(op, sequence, &mut lit_ptr, lit_end, base, oend);
|
||||
if ERR_isError(one_seq_size) {
|
||||
return one_seq_size;
|
||||
}
|
||||
op = op.add(one_seq_size);
|
||||
}
|
||||
|
||||
/* check if reached exact end */
|
||||
if !fse_end_of_dstream(&seq_state.d_stream) {
|
||||
/* requested too much : data is corrupted */
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
if nb_seq < 0 {
|
||||
/* requested too many sequences : data is corrupted */
|
||||
return ERROR(ZstdErrorCode::CorruptionDetected);
|
||||
}
|
||||
|
||||
/* last literal segment */
|
||||
{
|
||||
let last_ll_size = (lit_end as usize) - (lit_ptr as usize);
|
||||
if (op as usize).wrapping_add(last_ll_size) > (oend as usize) {
|
||||
return ERROR(ZstdErrorCode::DstSizeTooSmall);
|
||||
}
|
||||
if last_ll_size > 0 {
|
||||
if !std::ptr::eq(op as *const u8, lit_ptr) {
|
||||
ptr::copy(lit_ptr, op, last_ll_size);
|
||||
}
|
||||
op = op.add(last_ll_size);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
(op as usize) - (ostart as usize)
|
||||
}
|
||||
|
||||
unsafe fn zstd_decompress_block(
|
||||
ctx: *mut ZSTDv01_Dctx,
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
/* blockType == blockCompressed, srcSize is trusted */
|
||||
let mut ip = src;
|
||||
let mut lit_ptr: *const u8 = ptr::null();
|
||||
let mut lit_size: usize = 0;
|
||||
|
||||
/* Decode literals sub-block */
|
||||
let error_code = zstdv01_decode_literals_block(
|
||||
dst,
|
||||
max_dst_size,
|
||||
&mut lit_ptr,
|
||||
&mut lit_size,
|
||||
src,
|
||||
src_size,
|
||||
);
|
||||
if ERR_isError(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
ip = ip.add(error_code);
|
||||
let src_size = src_size - error_code;
|
||||
|
||||
zstd_decompress_sequences(ctx, dst, max_dst_size, ip, src_size, lit_ptr, lit_size)
|
||||
}
|
||||
|
||||
unsafe fn zstdv01_decompress_dctx(
|
||||
ctx: *mut ZSTDv01_Dctx,
|
||||
dst: *mut u8,
|
||||
max_dst_size: usize,
|
||||
src: *const u8,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
let mut ip = src;
|
||||
let iend_addr = (ip as usize).wrapping_add(src_size);
|
||||
let ostart = dst;
|
||||
let mut op = ostart;
|
||||
let oend_addr = (ostart as usize).wrapping_add(max_dst_size);
|
||||
let mut remaining_size = src_size;
|
||||
let mut error_code: usize = 0;
|
||||
|
||||
/* Frame Header */
|
||||
if src_size < ZSTD_FRAME_HEADER_SIZE + ZSTD_BLOCK_HEADER_SIZE {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
let magic_number = zstd_read_be32(src);
|
||||
if magic_number != ZSTD_MAGIC_NUMBER {
|
||||
return ERROR(ZstdErrorCode::PrefixUnknown);
|
||||
}
|
||||
ip = ip.add(ZSTD_FRAME_HEADER_SIZE);
|
||||
remaining_size -= ZSTD_FRAME_HEADER_SIZE;
|
||||
|
||||
/* Loop on each block */
|
||||
loop {
|
||||
let mut block_properties = BlockProperties {
|
||||
block_type: 0,
|
||||
orig_size: 0,
|
||||
};
|
||||
let block_size =
|
||||
zstdv01_getc_block_size(ip, iend_addr - ip as usize, &mut block_properties);
|
||||
if ERR_isError(block_size) {
|
||||
return block_size;
|
||||
}
|
||||
|
||||
ip = ip.add(ZSTD_BLOCK_HEADER_SIZE);
|
||||
remaining_size -= ZSTD_BLOCK_HEADER_SIZE;
|
||||
if block_size > remaining_size {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
|
||||
match block_properties.block_type {
|
||||
BT_COMPRESSED => {
|
||||
error_code =
|
||||
zstd_decompress_block(ctx, op, oend_addr - op as usize, ip, block_size);
|
||||
}
|
||||
BT_RAW => {
|
||||
error_code =
|
||||
zstd_copy_uncompressed_block(op, oend_addr - op as usize, ip, block_size);
|
||||
}
|
||||
BT_RLE => {
|
||||
return ERROR(ZstdErrorCode::Generic); /* not yet supported */
|
||||
}
|
||||
BT_END => {
|
||||
/* end of frame */
|
||||
if remaining_size != 0 {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
}
|
||||
_ => {
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
}
|
||||
if block_size == 0 {
|
||||
break; /* bt_end */
|
||||
}
|
||||
|
||||
if ERR_isError(error_code) {
|
||||
return error_code;
|
||||
}
|
||||
op = op.add(error_code);
|
||||
ip = ip.add(block_size);
|
||||
remaining_size -= block_size;
|
||||
}
|
||||
|
||||
(op as usize) - (ostart as usize)
|
||||
}
|
||||
|
||||
/* ZSTD_errorFrameSizeInfoLegacy() :
|
||||
assumes `cSize` and `dBound` are _not_ NULL */
|
||||
unsafe fn zstd_error_frame_size_info_legacy(c_size: *mut usize, d_bound: *mut u64, ret: usize) {
|
||||
*c_size = ret;
|
||||
*d_bound = ZSTD_CONTENTSIZE_ERROR;
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* Exported C ABI (zstd_v01.h)
|
||||
********************************************/
|
||||
|
||||
/// C ABI: `ZSTDv01_isError`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTDv01_isError(code: usize) -> c_uint {
|
||||
ERR_isError(code) as c_uint
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_decompressDCtx`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_decompressDCtx(
|
||||
ctx: *mut c_void,
|
||||
dst: *mut c_void,
|
||||
max_original_size: usize,
|
||||
src: *const c_void,
|
||||
compressed_size: usize,
|
||||
) -> usize {
|
||||
zstdv01_decompress_dctx(
|
||||
ctx as *mut ZSTDv01_Dctx,
|
||||
dst as *mut u8,
|
||||
max_original_size,
|
||||
src as *const u8,
|
||||
compressed_size,
|
||||
)
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_decompress`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_decompress(
|
||||
dst: *mut c_void,
|
||||
max_original_size: usize,
|
||||
src: *const c_void,
|
||||
compressed_size: usize,
|
||||
) -> usize {
|
||||
/* The C version uses an uninitialized on-stack dctx_t; only `base` is
|
||||
* read before being written, so a zeroed context is equivalent. */
|
||||
let mut ctx = std::mem::MaybeUninit::<ZSTDv01_Dctx>::zeroed();
|
||||
let ctx_ptr = ctx.as_mut_ptr();
|
||||
(*ctx_ptr).base = dst as *const u8;
|
||||
zstdv01_decompress_dctx(
|
||||
ctx_ptr,
|
||||
dst as *mut u8,
|
||||
max_original_size,
|
||||
src as *const u8,
|
||||
compressed_size,
|
||||
)
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_findFrameSizeInfoLegacy`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_findFrameSizeInfoLegacy(
|
||||
src: *const c_void,
|
||||
src_size: usize,
|
||||
c_size: *mut usize,
|
||||
d_bound: *mut u64,
|
||||
) {
|
||||
let mut ip = src as *const u8;
|
||||
let mut remaining_size = src_size;
|
||||
let mut nb_blocks: usize = 0;
|
||||
|
||||
/* Frame Header */
|
||||
if src_size < ZSTD_FRAME_HEADER_SIZE + ZSTD_BLOCK_HEADER_SIZE {
|
||||
zstd_error_frame_size_info_legacy(c_size, d_bound, ERROR(ZstdErrorCode::SrcSizeWrong));
|
||||
return;
|
||||
}
|
||||
let magic_number = zstd_read_be32(src as *const u8);
|
||||
if magic_number != ZSTD_MAGIC_NUMBER {
|
||||
zstd_error_frame_size_info_legacy(c_size, d_bound, ERROR(ZstdErrorCode::PrefixUnknown));
|
||||
return;
|
||||
}
|
||||
ip = ip.add(ZSTD_FRAME_HEADER_SIZE);
|
||||
remaining_size -= ZSTD_FRAME_HEADER_SIZE;
|
||||
|
||||
/* Loop on each block */
|
||||
loop {
|
||||
let mut block_properties = BlockProperties {
|
||||
block_type: 0,
|
||||
orig_size: 0,
|
||||
};
|
||||
let block_size = zstdv01_getc_block_size(ip, remaining_size, &mut block_properties);
|
||||
if ERR_isError(block_size) {
|
||||
zstd_error_frame_size_info_legacy(c_size, d_bound, block_size);
|
||||
return;
|
||||
}
|
||||
|
||||
ip = ip.add(ZSTD_BLOCK_HEADER_SIZE);
|
||||
remaining_size -= ZSTD_BLOCK_HEADER_SIZE;
|
||||
if block_size > remaining_size {
|
||||
zstd_error_frame_size_info_legacy(c_size, d_bound, ERROR(ZstdErrorCode::SrcSizeWrong));
|
||||
return;
|
||||
}
|
||||
|
||||
if block_size == 0 {
|
||||
break; /* bt_end */
|
||||
}
|
||||
|
||||
ip = ip.add(block_size);
|
||||
remaining_size -= block_size;
|
||||
nb_blocks += 1;
|
||||
}
|
||||
|
||||
*c_size = (ip as usize) - (src as usize);
|
||||
*d_bound = (nb_blocks * BLOCKSIZE) as u64;
|
||||
}
|
||||
|
||||
/* ******************************************
|
||||
* Streaming Decompression API
|
||||
********************************************/
|
||||
|
||||
/// C ABI: `ZSTDv01_resetDCtx`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_resetDCtx(dctx: *mut ZSTDv01_Dctx) -> usize {
|
||||
(*dctx).expected = ZSTD_FRAME_HEADER_SIZE;
|
||||
(*dctx).phase = 0;
|
||||
(*dctx).previous_dst_end = ptr::null();
|
||||
(*dctx).base = ptr::null();
|
||||
0
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_createDCtx`. Allocated with `malloc` exactly like the C
|
||||
/// implementation, so create/free may be paired across the C/Rust boundary.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_createDCtx() -> *mut ZSTDv01_Dctx {
|
||||
let dctx = libc::malloc(std::mem::size_of::<ZSTDv01_Dctx>()) as *mut ZSTDv01_Dctx;
|
||||
if dctx.is_null() {
|
||||
return ptr::null_mut();
|
||||
}
|
||||
ZSTDv01_resetDCtx(dctx);
|
||||
dctx
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_freeDCtx`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_freeDCtx(dctx: *mut ZSTDv01_Dctx) -> usize {
|
||||
libc::free(dctx as *mut c_void);
|
||||
0
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_nextSrcSizeToDecompress`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_nextSrcSizeToDecompress(dctx: *mut ZSTDv01_Dctx) -> usize {
|
||||
(*dctx).expected
|
||||
}
|
||||
|
||||
/// C ABI: `ZSTDv01_decompressContinue`.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTDv01_decompressContinue(
|
||||
dctx: *mut ZSTDv01_Dctx,
|
||||
dst: *mut c_void,
|
||||
max_dst_size: usize,
|
||||
src: *const c_void,
|
||||
src_size: usize,
|
||||
) -> usize {
|
||||
let ctx = dctx;
|
||||
let src = src as *const u8;
|
||||
let dst = dst as *mut u8;
|
||||
|
||||
/* Sanity check */
|
||||
if src_size != (*ctx).expected {
|
||||
return ERROR(ZstdErrorCode::SrcSizeWrong);
|
||||
}
|
||||
if !std::ptr::eq(dst as *const u8, (*ctx).previous_dst_end) {
|
||||
/* not contiguous */
|
||||
(*ctx).base = dst as *const u8;
|
||||
}
|
||||
|
||||
/* Decompress : frame header */
|
||||
if (*ctx).phase == 0 {
|
||||
/* Check frame magic header */
|
||||
let magic_number = zstd_read_be32(src);
|
||||
if magic_number != ZSTD_MAGIC_NUMBER {
|
||||
return ERROR(ZstdErrorCode::PrefixUnknown);
|
||||
}
|
||||
(*ctx).phase = 1;
|
||||
(*ctx).expected = ZSTD_BLOCK_HEADER_SIZE;
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* Decompress : block header */
|
||||
if (*ctx).phase == 1 {
|
||||
let mut bp = BlockProperties {
|
||||
block_type: 0,
|
||||
orig_size: 0,
|
||||
};
|
||||
let block_size = zstdv01_getc_block_size(src, ZSTD_BLOCK_HEADER_SIZE, &mut bp);
|
||||
if ERR_isError(block_size) {
|
||||
return block_size;
|
||||
}
|
||||
if bp.block_type == BT_END {
|
||||
(*ctx).expected = 0;
|
||||
(*ctx).phase = 0;
|
||||
} else {
|
||||
(*ctx).expected = block_size;
|
||||
(*ctx).b_type = bp.block_type;
|
||||
(*ctx).phase = 2;
|
||||
}
|
||||
|
||||
return 0;
|
||||
}
|
||||
|
||||
/* Decompress : block content */
|
||||
{
|
||||
let r_size = match (*ctx).b_type {
|
||||
BT_COMPRESSED => zstd_decompress_block(ctx, dst, max_dst_size, src, src_size),
|
||||
BT_RAW => zstd_copy_uncompressed_block(dst, max_dst_size, src, src_size),
|
||||
BT_RLE => {
|
||||
return ERROR(ZstdErrorCode::Generic); /* not yet handled */
|
||||
}
|
||||
BT_END => {
|
||||
/* should never happen (filtered at phase 1) */
|
||||
0
|
||||
}
|
||||
_ => {
|
||||
return ERROR(ZstdErrorCode::Generic);
|
||||
}
|
||||
};
|
||||
(*ctx).phase = 1;
|
||||
(*ctx).expected = ZSTD_BLOCK_HEADER_SIZE;
|
||||
if ERR_isError(r_size) {
|
||||
return r_size;
|
||||
}
|
||||
(*ctx).previous_dst_end = (dst as usize + r_size) as *const u8;
|
||||
r_size
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
/// 237-byte base text; the fixtures compress repetitions of it.
|
||||
const SAMPLE: &str = "snowden is snowed in / he's now then in his snow den / when does the snow end?\ngoodbye little dog / you dug some holes in your day / they'll be hard to fill.\nwhen life shuts a door, / just open it. it's a door. / that is how doors work.\n";
|
||||
|
||||
/// `SAMPLE` repeated 3 times (711 bytes), compressed by a zstd binary
|
||||
/// built from the v0.1.0 tag. Single compressed block (FSE + Huff0),
|
||||
/// verified byte-identical against the pristine C decoder.
|
||||
const FRAME_ENTROPY: &[u8] = &[
|
||||
0xFD, 0x2F, 0xB5, 0x1E, 0x00, 0x00, 0xB0, 0x00, 0x00, 0x93, 0x00, 0xD6, 0x19, 0x80, 0x87,
|
||||
0xB1, 0x00, 0xC0, 0x73, 0xC5, 0x8A, 0xA5, 0x6C, 0x6F, 0x4B, 0xF2, 0x9F, 0x94, 0xC4, 0x2B,
|
||||
0xE4, 0x6B, 0x2C, 0x96, 0xAE, 0x5F, 0xC8, 0x9C, 0x1F, 0x00, 0x1C, 0x00, 0x1B, 0x00, 0x43,
|
||||
0x31, 0x70, 0xE4, 0xC9, 0xA3, 0x46, 0xD9, 0xAD, 0xFD, 0x78, 0x37, 0x66, 0xA6, 0x4D, 0x15,
|
||||
0xCA, 0x33, 0xF5, 0xA7, 0x31, 0x83, 0x02, 0x96, 0x7E, 0xD0, 0x5E, 0xAC, 0xE9, 0x48, 0x03,
|
||||
0xC7, 0xEF, 0x44, 0x81, 0x0B, 0x00, 0x90, 0xEA, 0x55, 0x41, 0xD8, 0x4E, 0x36, 0x10, 0xAC,
|
||||
0x95, 0x92, 0x42, 0x5D, 0x92, 0xF4, 0x64, 0x7F, 0x96, 0x44, 0x59, 0x0B, 0x03, 0x4A, 0xF6,
|
||||
0x16, 0x9F, 0xC8, 0xB9, 0x11, 0x70, 0xB8, 0x35, 0x06, 0xC4, 0x26, 0xC7, 0x6A, 0x89, 0xE4,
|
||||
0x6F, 0x65, 0xC0, 0x9B, 0x55, 0x08, 0xEB, 0xE8, 0x37, 0x66, 0x67, 0x9E, 0x7D, 0x57, 0xA0,
|
||||
0x25, 0x0D, 0xA4, 0x3E, 0x26, 0x8F, 0x32, 0xF6, 0xC9, 0x49, 0x50, 0x40, 0x0A, 0xED, 0x34,
|
||||
0x94, 0x15, 0x5E, 0xE3, 0x1B, 0xD6, 0x27, 0x05, 0x00, 0x54, 0x05, 0x41, 0xFF, 0xD6, 0x01,
|
||||
0x00, 0x80, 0x36, 0x00, 0xFE, 0x8D, 0x38, 0x42, 0x51, 0x7E, 0x40, 0x38, 0xD5, 0x60, 0x46,
|
||||
0x03, 0x90, 0x25, 0xC0, 0x00, 0x00,
|
||||
];
|
||||
|
||||
/// 48 random bytes compressed by v0.1.0: a raw (uncompressed) block.
|
||||
const FRAME_RAW: &[u8] = &[
|
||||
0xFD, 0x2F, 0xB5, 0x1E, 0x40, 0x00, 0x30, 0x77, 0xE9, 0x9A, 0x70, 0x2F, 0x77, 0x16, 0x9A,
|
||||
0x92, 0xD4, 0xEE, 0x72, 0xFD, 0xDD, 0x0F, 0x39, 0x33, 0x8B, 0x8B, 0x3C, 0x6A, 0x91, 0xB2,
|
||||
0x2B, 0x3A, 0xF0, 0x0E, 0xA6, 0x4C, 0x51, 0xA0, 0xC9, 0x5F, 0x69, 0x1A, 0xCE, 0x93, 0x5C,
|
||||
0x7F, 0x43, 0x0C, 0xD9, 0x25, 0xBC, 0x91, 0xDD, 0x6E, 0xD7, 0xC0, 0x00, 0x00,
|
||||
];
|
||||
|
||||
const RAW_PAYLOAD: &[u8] = &[
|
||||
0x77, 0xE9, 0x9A, 0x70, 0x2F, 0x77, 0x16, 0x9A, 0x92, 0xD4, 0xEE, 0x72, 0xFD, 0xDD, 0x0F,
|
||||
0x39, 0x33, 0x8B, 0x8B, 0x3C, 0x6A, 0x91, 0xB2, 0x2B, 0x3A, 0xF0, 0x0E, 0xA6, 0x4C, 0x51,
|
||||
0xA0, 0xC9, 0x5F, 0x69, 0x1A, 0xCE, 0x93, 0x5C, 0x7F, 0x43, 0x0C, 0xD9, 0x25, 0xBC, 0x91,
|
||||
0xDD, 0x6E, 0xD7,
|
||||
];
|
||||
|
||||
/// `SAMPLE` repeated 1800 times (426600 bytes) compressed by v0.1.0:
|
||||
/// four data blocks, so it exercises multi-block frames and repeated
|
||||
/// offsets.
|
||||
const FRAME_MULTI_BLOCK: &[u8] = &[
|
||||
0xFD, 0x2F, 0xB5, 0x1E, 0x00, 0x00, 0xB0, 0x00, 0x00, 0x93, 0x00, 0xD6, 0x19, 0x80, 0x87,
|
||||
0xB1, 0x00, 0xC0, 0x73, 0xC5, 0x8A, 0xA5, 0x6C, 0x6F, 0x4B, 0xF2, 0x9F, 0x94, 0xC4, 0x2B,
|
||||
0xE4, 0x6B, 0x2C, 0x96, 0xAE, 0x5F, 0xC8, 0x9C, 0x1F, 0x00, 0x1C, 0x00, 0x1B, 0x00, 0x43,
|
||||
0x31, 0x70, 0xE4, 0xC9, 0xA3, 0x46, 0xD9, 0xAD, 0xFD, 0x78, 0x37, 0x66, 0xA6, 0x4D, 0x15,
|
||||
0xCA, 0x33, 0xF5, 0xA7, 0x31, 0x83, 0x02, 0x96, 0x7E, 0xD0, 0x5E, 0xAC, 0xE9, 0x48, 0x03,
|
||||
0xC7, 0xEF, 0x44, 0x81, 0x0B, 0x00, 0x90, 0xEA, 0x55, 0x41, 0xD8, 0x4E, 0x36, 0x10, 0xAC,
|
||||
0x95, 0x92, 0x42, 0x5D, 0x92, 0xF4, 0x64, 0x7F, 0x96, 0x44, 0x59, 0x0B, 0x03, 0x4A, 0xF6,
|
||||
0x16, 0x9F, 0xC8, 0xB9, 0x11, 0x70, 0xB8, 0x35, 0x06, 0xC4, 0x26, 0xC7, 0x6A, 0x89, 0xE4,
|
||||
0x6F, 0x65, 0xC0, 0x9B, 0x55, 0x08, 0xEB, 0xE8, 0x37, 0x66, 0x67, 0x9E, 0x7D, 0x57, 0xA0,
|
||||
0x25, 0x0D, 0xA4, 0x3E, 0x26, 0x8F, 0x32, 0xF6, 0xC9, 0x49, 0x50, 0x40, 0x0A, 0xED, 0x34,
|
||||
0x94, 0x15, 0x5E, 0xE3, 0x1B, 0xD6, 0x27, 0x05, 0x00, 0x54, 0x05, 0x41, 0xFF, 0x0F, 0xFF,
|
||||
0x01, 0x80, 0x36, 0x00, 0xFE, 0x8D, 0x38, 0x42, 0x51, 0x7E, 0x40, 0x38, 0xD5, 0x60, 0x46,
|
||||
0x03, 0x90, 0x25, 0x00, 0x00, 0x12, 0x40, 0x00, 0x00, 0x01, 0x00, 0x54, 0x04, 0xFF, 0xFC,
|
||||
0xFF, 0x01, 0x80, 0xFA, 0x7F, 0x00, 0xFC, 0x23, 0x10, 0x00, 0x00, 0x12, 0x40, 0x00, 0x00,
|
||||
0x01, 0x00, 0x54, 0x04, 0xFF, 0xFC, 0xFF, 0x01, 0x00, 0xF5, 0xFF, 0x00, 0xF8, 0x4B, 0x20,
|
||||
0x00, 0x00, 0x12, 0x40, 0x00, 0x00, 0x01, 0x00, 0x54, 0x04, 0xFF, 0x64, 0x82, 0x00, 0x80,
|
||||
0xEF, 0xFF, 0x00, 0xF0, 0x9F, 0x40, 0xC0, 0x00, 0x00,
|
||||
];
|
||||
|
||||
fn decompress(frame: &[u8], capacity: usize) -> Result<Vec<u8>, usize> {
|
||||
let mut out = vec![0u8; capacity];
|
||||
let code = unsafe {
|
||||
ZSTDv01_decompress(
|
||||
out.as_mut_ptr() as *mut c_void,
|
||||
capacity,
|
||||
frame.as_ptr() as *const c_void,
|
||||
frame.len(),
|
||||
)
|
||||
};
|
||||
if ZSTDv01_isError(code) != 0 {
|
||||
return Err(code);
|
||||
}
|
||||
out.truncate(code);
|
||||
Ok(out)
|
||||
}
|
||||
|
||||
fn frame_size_info(frame: &[u8]) -> (usize, u64) {
|
||||
let mut c_size = 0usize;
|
||||
let mut d_bound = 0u64;
|
||||
unsafe {
|
||||
ZSTDv01_findFrameSizeInfoLegacy(
|
||||
frame.as_ptr() as *const c_void,
|
||||
frame.len(),
|
||||
&mut c_size,
|
||||
&mut d_bound,
|
||||
);
|
||||
}
|
||||
(c_size, d_bound)
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn decodes_entropy_frame_byte_identically() {
|
||||
let expected = SAMPLE.repeat(3).into_bytes();
|
||||
let out = decompress(FRAME_ENTROPY, expected.len()).unwrap();
|
||||
assert_eq!(out, expected);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn decodes_raw_block_frame_byte_identically() {
|
||||
let out = decompress(FRAME_RAW, RAW_PAYLOAD.len()).unwrap();
|
||||
assert_eq!(out, RAW_PAYLOAD);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn decodes_multi_block_frame_byte_identically() {
|
||||
let expected = SAMPLE.repeat(1800).into_bytes();
|
||||
assert_eq!(expected.len(), 426600);
|
||||
let out = decompress(FRAME_MULTI_BLOCK, expected.len()).unwrap();
|
||||
assert_eq!(out, expected);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn reports_frame_size_info_like_c() {
|
||||
/* Values verified against the pristine C implementation. */
|
||||
assert_eq!(
|
||||
frame_size_info(FRAME_ENTROPY),
|
||||
(FRAME_ENTROPY.len(), 131072)
|
||||
);
|
||||
assert_eq!(frame_size_info(FRAME_RAW), (FRAME_RAW.len(), 131072));
|
||||
assert_eq!(
|
||||
frame_size_info(FRAME_MULTI_BLOCK),
|
||||
(FRAME_MULTI_BLOCK.len(), 524288)
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_bad_magic_with_prefix_unknown() {
|
||||
let mut frame = FRAME_ENTROPY.to_vec();
|
||||
frame[0] ^= 0x55;
|
||||
assert_eq!(
|
||||
decompress(&frame, 1024).unwrap_err(),
|
||||
ERROR(ZstdErrorCode::PrefixUnknown)
|
||||
);
|
||||
let (c_size, d_bound) = frame_size_info(&frame);
|
||||
assert_eq!(c_size, ERROR(ZstdErrorCode::PrefixUnknown));
|
||||
assert_eq!(d_bound, ZSTD_CONTENTSIZE_ERROR);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_truncated_frames_with_src_size_wrong() {
|
||||
for frame in [FRAME_ENTROPY, FRAME_RAW, FRAME_MULTI_BLOCK] {
|
||||
let truncated = &frame[..frame.len() - 1];
|
||||
assert_eq!(
|
||||
decompress(truncated, 1 << 20).unwrap_err(),
|
||||
ERROR(ZstdErrorCode::SrcSizeWrong)
|
||||
);
|
||||
assert_eq!(
|
||||
decompress(&frame[..5], 1 << 20).unwrap_err(),
|
||||
ERROR(ZstdErrorCode::SrcSizeWrong)
|
||||
);
|
||||
}
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn rejects_small_destination_with_dst_size_too_small() {
|
||||
/* Error codes verified against the pristine C implementation. */
|
||||
let expected_len = SAMPLE.len() * 3;
|
||||
assert_eq!(
|
||||
decompress(FRAME_ENTROPY, expected_len - 1).unwrap_err(),
|
||||
ERROR(ZstdErrorCode::DstSizeTooSmall)
|
||||
);
|
||||
assert_eq!(
|
||||
decompress(FRAME_ENTROPY, 0).unwrap_err(),
|
||||
ERROR(ZstdErrorCode::DstSizeTooSmall)
|
||||
);
|
||||
assert_eq!(
|
||||
decompress(FRAME_RAW, RAW_PAYLOAD.len() - 1).unwrap_err(),
|
||||
ERROR(ZstdErrorCode::DstSizeTooSmall)
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn streaming_api_decodes_the_frame() {
|
||||
let expected = SAMPLE.repeat(3).into_bytes();
|
||||
let mut out = vec![0u8; expected.len()];
|
||||
let mut in_pos = 0usize;
|
||||
let mut out_pos = 0usize;
|
||||
unsafe {
|
||||
let dctx = ZSTDv01_createDCtx();
|
||||
assert!(!dctx.is_null());
|
||||
loop {
|
||||
let needed = ZSTDv01_nextSrcSizeToDecompress(dctx);
|
||||
if needed == 0 {
|
||||
break;
|
||||
}
|
||||
assert!(in_pos + needed <= FRAME_ENTROPY.len());
|
||||
let produced = ZSTDv01_decompressContinue(
|
||||
dctx,
|
||||
out.as_mut_ptr().add(out_pos) as *mut c_void,
|
||||
out.len() - out_pos,
|
||||
FRAME_ENTROPY.as_ptr().add(in_pos) as *const c_void,
|
||||
needed,
|
||||
);
|
||||
assert_eq!(ZSTDv01_isError(produced), 0);
|
||||
in_pos += needed;
|
||||
out_pos += produced;
|
||||
}
|
||||
assert_eq!(ZSTDv01_freeDCtx(dctx), 0);
|
||||
}
|
||||
assert_eq!(in_pos, FRAME_ENTROPY.len());
|
||||
assert_eq!(out_pos, expected.len());
|
||||
assert_eq!(out, expected);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn streaming_context_reset_and_null_free() {
|
||||
unsafe {
|
||||
let dctx = ZSTDv01_createDCtx();
|
||||
assert_eq!(ZSTDv01_nextSrcSizeToDecompress(dctx), 4);
|
||||
/* bad magic through the streaming entry point */
|
||||
let bad = [0u8; 4];
|
||||
assert_eq!(
|
||||
ZSTDv01_decompressContinue(
|
||||
dctx,
|
||||
ptr::null_mut(),
|
||||
0,
|
||||
bad.as_ptr() as *const c_void,
|
||||
4
|
||||
),
|
||||
ERROR(ZstdErrorCode::PrefixUnknown)
|
||||
);
|
||||
/* wrong srcSize */
|
||||
assert_eq!(
|
||||
ZSTDv01_decompressContinue(
|
||||
dctx,
|
||||
ptr::null_mut(),
|
||||
0,
|
||||
bad.as_ptr() as *const c_void,
|
||||
3
|
||||
),
|
||||
ERROR(ZstdErrorCode::SrcSizeWrong)
|
||||
);
|
||||
assert_eq!(ZSTDv01_resetDCtx(dctx), 0);
|
||||
assert_eq!(ZSTDv01_nextSrcSizeToDecompress(dctx), 4);
|
||||
assert_eq!(ZSTDv01_freeDCtx(dctx), 0);
|
||||
assert_eq!(ZSTDv01_freeDCtx(ptr::null_mut()), 0);
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -18,7 +18,6 @@ pub mod hist;
|
||||
pub mod huf_compress;
|
||||
#[cfg(feature = "decompression")]
|
||||
pub mod huf_decompress;
|
||||
pub mod legacy;
|
||||
pub mod mem;
|
||||
pub mod pool;
|
||||
pub mod threading;
|
||||
@@ -31,8 +30,6 @@ pub mod zstd_compress_frame;
|
||||
#[cfg(feature = "compression")]
|
||||
pub mod zstd_compress_literals;
|
||||
#[cfg(feature = "compression")]
|
||||
pub mod zstd_compress_params;
|
||||
#[cfg(feature = "compression")]
|
||||
pub mod zstd_compress_sequences;
|
||||
#[cfg(feature = "compression")]
|
||||
pub mod zstd_compress_superblock;
|
||||
|
||||
@@ -1,198 +0,0 @@
|
||||
#![allow(non_camel_case_types)]
|
||||
#![allow(non_snake_case)]
|
||||
|
||||
//! Precise monotonic time measurement for the command-line programs.
|
||||
//!
|
||||
//! Port of `programs/timefn.c`. `UTIL_time_t` is a plain nanosecond counter
|
||||
//! whose absolute value is meaningless; only spans between two measurements
|
||||
//! are valid. The struct crosses the C ABI by value, so it stays `repr(C)`
|
||||
//! with the exact `timefn.h` layout.
|
||||
//!
|
||||
//! Platform selection mirrors the C preprocessor structure: Windows uses the
|
||||
//! performance counter, Apple systems use the Mach absolute clock, and other
|
||||
//! POSIX systems use `clock_gettime(CLOCK_MONOTONIC)`. The C90 `clock()`
|
||||
//! fallback is never needed on targets Rust supports, so multi-threaded
|
||||
//! measurements are always supported.
|
||||
|
||||
use std::os::raw::c_int;
|
||||
|
||||
/// Precise Time (`PTime` in timefn.h): an unsigned 64-bit nanosecond count.
|
||||
pub type PTime = u64;
|
||||
|
||||
/// Nanosecond time counter with the `timefn.h` `UTIL_time_t` layout.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug)]
|
||||
pub struct UTIL_time_t {
|
||||
pub t: PTime,
|
||||
}
|
||||
|
||||
const _: () = assert!(std::mem::size_of::<PTime>() == 8);
|
||||
const _: () = assert!(std::mem::size_of::<UTIL_time_t>() == std::mem::size_of::<PTime>());
|
||||
|
||||
#[cfg(windows)]
|
||||
mod platform {
|
||||
use super::PTime;
|
||||
use std::sync::OnceLock;
|
||||
|
||||
#[link(name = "kernel32")]
|
||||
unsafe extern "system" {
|
||||
/// Takes a `LARGE_INTEGER*`; the union is ABI-identical to `i64*`.
|
||||
fn QueryPerformanceCounter(count: *mut i64) -> i32;
|
||||
fn QueryPerformanceFrequency(frequency: *mut i64) -> i32;
|
||||
}
|
||||
|
||||
pub fn monotonic_ns() -> PTime {
|
||||
static TICKS_PER_SECOND: OnceLock<i64> = OnceLock::new();
|
||||
let ticks_per_second = *TICKS_PER_SECOND.get_or_init(|| {
|
||||
let mut frequency = 0i64;
|
||||
if unsafe { QueryPerformanceFrequency(&mut frequency) } == 0 {
|
||||
eprintln!(
|
||||
"timefn::QueryPerformanceFrequency: {}",
|
||||
std::io::Error::last_os_error()
|
||||
);
|
||||
std::process::abort();
|
||||
}
|
||||
frequency
|
||||
});
|
||||
let mut counter = 0i64;
|
||||
unsafe { QueryPerformanceCounter(&mut counter) };
|
||||
(counter as PTime).wrapping_mul(1_000_000_000) / ticks_per_second as PTime
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(all(unix, target_vendor = "apple"))]
|
||||
mod platform {
|
||||
use super::PTime;
|
||||
use std::sync::OnceLock;
|
||||
|
||||
pub fn monotonic_ns() -> PTime {
|
||||
static RATE: OnceLock<(PTime, PTime)> = OnceLock::new();
|
||||
let (numer, denom) = *RATE.get_or_init(|| {
|
||||
let mut rate = libc::mach_timebase_info { numer: 0, denom: 0 };
|
||||
unsafe { libc::mach_timebase_info(&mut rate) };
|
||||
(PTime::from(rate.numer), PTime::from(rate.denom))
|
||||
});
|
||||
unsafe { libc::mach_absolute_time() }.wrapping_mul(numer) / denom
|
||||
}
|
||||
}
|
||||
|
||||
#[cfg(all(unix, not(target_vendor = "apple")))]
|
||||
mod platform {
|
||||
use super::PTime;
|
||||
|
||||
pub fn monotonic_ns() -> PTime {
|
||||
// Zero-initialized like the C source, which works around timespec_get
|
||||
// msan limitations on some targets.
|
||||
let mut time: libc::timespec = unsafe { std::mem::zeroed() };
|
||||
if unsafe { libc::clock_gettime(libc::CLOCK_MONOTONIC, &mut time) } != 0 {
|
||||
eprintln!(
|
||||
"timefn::clock_gettime(CLOCK_MONOTONIC): {}",
|
||||
std::io::Error::last_os_error()
|
||||
);
|
||||
std::process::abort();
|
||||
}
|
||||
(time.tv_sec as PTime)
|
||||
.wrapping_mul(1_000_000_000)
|
||||
.wrapping_add(time.tv_nsec as PTime)
|
||||
}
|
||||
}
|
||||
|
||||
/// Returns the current value of the platform's monotonic nanosecond clock.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_getTime() -> UTIL_time_t {
|
||||
UTIL_time_t {
|
||||
t: platform::monotonic_ns(),
|
||||
}
|
||||
}
|
||||
|
||||
/// Nanoseconds elapsed between two measurements, with C unsigned wrap-around.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_getSpanTimeNano(clockStart: UTIL_time_t, clockEnd: UTIL_time_t) -> PTime {
|
||||
clockEnd.t.wrapping_sub(clockStart.t)
|
||||
}
|
||||
|
||||
/// Microseconds elapsed between two measurements, truncated like C division.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_getSpanTimeMicro(begin: UTIL_time_t, end: UTIL_time_t) -> PTime {
|
||||
UTIL_getSpanTimeNano(begin, end) / 1000
|
||||
}
|
||||
|
||||
/// Microseconds elapsed since `clockStart`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_clockSpanMicro(clockStart: UTIL_time_t) -> PTime {
|
||||
UTIL_getSpanTimeMicro(clockStart, UTIL_getTime())
|
||||
}
|
||||
|
||||
/// Nanoseconds elapsed since `clockStart`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_clockSpanNano(clockStart: UTIL_time_t) -> PTime {
|
||||
UTIL_getSpanTimeNano(clockStart, UTIL_getTime())
|
||||
}
|
||||
|
||||
/// Busy-waits until the clock produces a new tick, improving measurement
|
||||
/// accuracy on platforms with a low timer resolution.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_waitForNextTick() {
|
||||
let clockStart = UTIL_getTime();
|
||||
loop {
|
||||
let clockEnd = UTIL_getTime();
|
||||
if UTIL_getSpanTimeNano(clockStart, clockEnd) != 0 {
|
||||
return;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// All clock sources used by the Rust port are valid under multi-threaded
|
||||
/// workloads; only the C90 `clock()` fallback of the C source was not.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn UTIL_support_MT_measurements() -> c_int {
|
||||
1
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
|
||||
#[test]
|
||||
fn nanosecond_spans_subtract_with_unsigned_wrap_around() {
|
||||
let start = UTIL_time_t { t: 100 };
|
||||
let end = UTIL_time_t { t: 350 };
|
||||
|
||||
assert_eq!(UTIL_getSpanTimeNano(start, end), 250);
|
||||
assert_eq!(UTIL_getSpanTimeNano(end, start), u64::MAX - 249);
|
||||
assert_eq!(UTIL_getSpanTimeNano(start, start), 0);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn microsecond_spans_truncate_sub_tick_remainders() {
|
||||
let start = UTIL_time_t { t: 0 };
|
||||
|
||||
assert_eq!(UTIL_getSpanTimeMicro(start, UTIL_time_t { t: 999 }), 0);
|
||||
assert_eq!(UTIL_getSpanTimeMicro(start, UTIL_time_t { t: 1_000 }), 1);
|
||||
assert_eq!(UTIL_getSpanTimeMicro(start, UTIL_time_t { t: 1_999 }), 1);
|
||||
assert_eq!(UTIL_getSpanTimeMicro(start, UTIL_time_t { t: 2_000 }), 2);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn clock_is_monotonic_across_measurements() {
|
||||
let first = UTIL_getTime();
|
||||
let second = UTIL_getTime();
|
||||
|
||||
assert!(second.t >= first.t);
|
||||
assert!(UTIL_clockSpanNano(first) >= UTIL_getSpanTimeNano(first, second));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn waiting_for_the_next_tick_advances_the_clock() {
|
||||
let before = UTIL_getTime();
|
||||
UTIL_waitForNextTick();
|
||||
let after = UTIL_getTime();
|
||||
|
||||
assert!(UTIL_getSpanTimeNano(before, after) > 0);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn multi_threaded_measurements_are_supported() {
|
||||
assert_eq!(UTIL_support_MT_measurements(), 1);
|
||||
}
|
||||
}
|
||||
+9
-194
@@ -10,15 +10,9 @@
|
||||
//! writes, dictionary loading, streaming, and metadata preservation remain in
|
||||
//! `programs/fileio.c` for this first migration step.
|
||||
//!
|
||||
//! Benchmark mode (`-b`) parses here and dispatches through the
|
||||
//! `ZSTD_rust_cli_bench` bridge in `programs/zstdcli.c`: the run/timing loop
|
||||
//! (benchfn, timefn) is Rust, while orchestration and result formatting
|
||||
//! (`benchzstd.c`) remain C behind the preprocessor-gated bridge, so builds
|
||||
//! with `ZSTD_NOBENCH` never reference benchmark symbols.
|
||||
//!
|
||||
//! Remaining C-only CLI boundaries are called out in `unsupported()` below:
|
||||
//! dictionary training, recursive/file-list expansion, tracing,
|
||||
//! alternate-format selection, and the advanced directory modes.
|
||||
//! benchmark execution, dictionary training, recursive/file-list expansion,
|
||||
//! tracing, alternate-format selection, and the advanced directory modes.
|
||||
|
||||
use std::env;
|
||||
use std::ffi::{CStr, CString, OsStr, OsString};
|
||||
@@ -36,7 +30,6 @@ use std::os::unix::fs::FileTypeExt;
|
||||
const DEFAULT_CLEVEL: i32 = 3;
|
||||
#[cfg(feature = "compression")]
|
||||
const DEFAULT_MAX_CLEVEL: i32 = 19;
|
||||
const DEFAULT_BENCH_NB_SECONDS: u32 = 3;
|
||||
const DEFAULT_MEM_LIMIT: u32 = 1 << 27;
|
||||
const DEFAULT_LONG_WINDOW_LOG: u32 = 27;
|
||||
const MAX_FAST_ACCELERATION: i32 = 128 << 10;
|
||||
@@ -180,22 +173,6 @@ unsafe extern "C" {
|
||||
output: *const c_char,
|
||||
dict: *const c_char,
|
||||
) -> c_int;
|
||||
|
||||
/// Benchmark bridge implemented by the `programs/zstdcli.c` shim, which
|
||||
/// owns the `ZSTD_NOBENCH` preprocessor decision. Returns the benchmark
|
||||
/// result (>= 0), or -1 when benchmarking is compiled out.
|
||||
fn ZSTD_rust_cli_bench(
|
||||
file_names: *const *const c_char,
|
||||
nb_files: c_uint,
|
||||
dict_file_name: *const c_char,
|
||||
start_level: c_int,
|
||||
end_level: c_int,
|
||||
compression_params: *const ZSTD_compressionParameters,
|
||||
display_level: c_int,
|
||||
nb_seconds: c_uint,
|
||||
block_size: usize,
|
||||
nb_workers: c_int,
|
||||
) -> c_int;
|
||||
}
|
||||
|
||||
#[derive(Clone, Copy, Debug, Eq, PartialEq)]
|
||||
@@ -203,7 +180,6 @@ enum Operation {
|
||||
Compress,
|
||||
Decompress,
|
||||
Test,
|
||||
Bench,
|
||||
}
|
||||
|
||||
#[derive(Debug)]
|
||||
@@ -234,7 +210,6 @@ struct Cli {
|
||||
mmap_dict: i32,
|
||||
progress: i32,
|
||||
workers: Option<i32>,
|
||||
single_thread: bool,
|
||||
block_size: Option<usize>,
|
||||
mem_limit: Option<u32>,
|
||||
ldm: bool,
|
||||
@@ -254,8 +229,6 @@ struct Cli {
|
||||
row_match_finder: i32,
|
||||
exclude_compressed: bool,
|
||||
compression_params: ZSTD_compressionParameters,
|
||||
bench_end_level: Option<i32>,
|
||||
bench_nb_seconds: Option<u32>,
|
||||
unsupported_program: Option<String>,
|
||||
}
|
||||
|
||||
@@ -281,7 +254,6 @@ impl Cli {
|
||||
mmap_dict: ZSTD_PS_AUTO,
|
||||
progress: FIO_PS_AUTO,
|
||||
workers: None,
|
||||
single_thread: false,
|
||||
block_size: None,
|
||||
mem_limit: None,
|
||||
ldm: false,
|
||||
@@ -301,8 +273,6 @@ impl Cli {
|
||||
row_match_finder: ZSTD_PS_AUTO,
|
||||
exclude_compressed: false,
|
||||
compression_params: ZSTD_compressionParameters::default(),
|
||||
bench_end_level: None,
|
||||
bench_nb_seconds: None,
|
||||
unsupported_program: None,
|
||||
};
|
||||
|
||||
@@ -362,11 +332,8 @@ unsafe fn default_worker_count() -> i32 {
|
||||
}
|
||||
|
||||
#[cfg(feature = "compression")]
|
||||
unsafe fn resolved_worker_count(workers: Option<i32>, single_thread: bool) -> i32 {
|
||||
unsafe fn resolved_worker_count(workers: Option<i32>) -> i32 {
|
||||
match workers {
|
||||
/* --single-thread pins zero workers; a bare zero (-T0 or the zstdmt
|
||||
* program name) auto-detects the core count as in the C CLI. */
|
||||
Some(0) if single_thread => 0,
|
||||
Some(0) => unsafe { UTIL_countPhysicalCores() }.max(1),
|
||||
Some(workers) => workers,
|
||||
None => unsafe { default_worker_count() },
|
||||
@@ -418,7 +385,7 @@ fn usage(advanced: bool) {
|
||||
let _ = writeln!(out, "\nImplemented advanced compression controls:");
|
||||
let _ = writeln!(
|
||||
out,
|
||||
" --fast[=#], --ultra, --long[=#], --threads=#, --single-thread, --block-size=#"
|
||||
" --fast[=#], --ultra, --long[=#], --threads=#, --block-size=#"
|
||||
);
|
||||
let _ = writeln!(
|
||||
out,
|
||||
@@ -432,22 +399,9 @@ fn usage(advanced: bool) {
|
||||
out,
|
||||
" --adapt[=min=#,max=#], --rsyncable, --[no-]row-match-finder"
|
||||
);
|
||||
let _ = writeln!(out, "\nBenchmark options:");
|
||||
let _ = writeln!(
|
||||
out,
|
||||
" -b# Benchmark file(s) at compression level #"
|
||||
);
|
||||
let _ = writeln!(
|
||||
out,
|
||||
" -e# Test all levels from -b# up to # included"
|
||||
);
|
||||
let _ = writeln!(
|
||||
out,
|
||||
" -i# Set the minimum evaluation time to # seconds"
|
||||
);
|
||||
let _ = writeln!(
|
||||
out,
|
||||
"\nNot yet migrated: dictionary training, recursive/file-list expansion,"
|
||||
"\nNot yet migrated: benchmark, dictionary training, recursive/file-list expansion,"
|
||||
);
|
||||
let _ = writeln!(out, "trace, alternate formats, and output-directory modes.");
|
||||
}
|
||||
@@ -684,7 +638,6 @@ fn parse_long_option(
|
||||
| "--no-row-match-finder"
|
||||
| "--row-match-finder"
|
||||
| "--rsyncable"
|
||||
| "--single-thread"
|
||||
| "--compress-literals"
|
||||
| "--no-compress-literals"
|
||||
| "--exclude-compressed"
|
||||
@@ -856,14 +809,6 @@ fn parse_long_option(
|
||||
cli.workers = Some(parse_worker_count(&value)?);
|
||||
Ok(None)
|
||||
}
|
||||
"--single-thread" => {
|
||||
/* As in the C CLI: zero workers plus a latch that suppresses the
|
||||
* automatic core-count resolution, so fileio runs its
|
||||
* single-thread streaming mode (slightly different from -T1). */
|
||||
cli.workers = Some(0);
|
||||
cli.single_thread = true;
|
||||
Ok(None)
|
||||
}
|
||||
"--memlimit" | "--memory" | "--memlimit-decompress" => {
|
||||
let value = next_value(attached, args, index, name)?;
|
||||
cli.mem_limit = Some(parse_u32(&value, "memory limit")?);
|
||||
@@ -909,6 +854,7 @@ fn parse_long_option(
|
||||
| "--trace"
|
||||
| "--format"
|
||||
| "--priority"
|
||||
| "--single-thread"
|
||||
| "--auto-threads"
|
||||
| "--fake-stdin-is-console"
|
||||
| "--fake-stdout-is-console"
|
||||
@@ -946,33 +892,6 @@ fn parse_short_options(
|
||||
'd' => cli.operation = Operation::Decompress,
|
||||
'z' => cli.operation = Operation::Compress,
|
||||
't' => cli.operation = Operation::Test,
|
||||
'b' => cli.operation = Operation::Bench,
|
||||
'e' | 'i' => {
|
||||
// Benchmark range end (-e#) and duration (-i#): like the C
|
||||
// parser, digits attach directly and default to 0.
|
||||
let mut digits_end = offset + 1;
|
||||
while digits_end < bytes.len() && bytes[digits_end].is_ascii_digit() {
|
||||
digits_end += 1;
|
||||
}
|
||||
let digits = &value[offset + 1..digits_end];
|
||||
if option == 'e' {
|
||||
cli.bench_end_level = Some(if digits.is_empty() {
|
||||
0
|
||||
} else {
|
||||
parse_i32(digits, "benchmark end level")?
|
||||
});
|
||||
} else {
|
||||
cli.bench_nb_seconds = Some(if digits.is_empty() {
|
||||
0
|
||||
} else {
|
||||
digits
|
||||
.parse::<u32>()
|
||||
.map_err(|_| format!("invalid benchmark duration: {digits:?}"))?
|
||||
});
|
||||
}
|
||||
offset = digits_end;
|
||||
continue;
|
||||
}
|
||||
'c' => {
|
||||
cli.output = Some(cstring(STDOUT_MARK)?);
|
||||
cli.force_stdout = true;
|
||||
@@ -1008,7 +927,7 @@ fn parse_short_options(
|
||||
}
|
||||
break;
|
||||
}
|
||||
'l' | 'p' | 'P' | 'r' | 's' | 'S' => {
|
||||
'b' | 'e' | 'i' | 'l' | 'p' | 'P' | 'r' | 's' | 'S' => {
|
||||
unsupported(&format!("-{option}"))?;
|
||||
}
|
||||
_ => return Err(format!("unknown option -{option}")),
|
||||
@@ -1093,7 +1012,7 @@ unsafe fn apply_preferences(cli: &Cli, prefs: *mut FIO_prefs_t, ctx: *mut FIO_ct
|
||||
}),
|
||||
);
|
||||
#[cfg(feature = "compression")]
|
||||
FIO_setNbWorkers(prefs, resolved_worker_count(cli.workers, cli.single_thread));
|
||||
FIO_setNbWorkers(prefs, resolved_worker_count(cli.workers));
|
||||
FIO_setLdmFlag(prefs, u32::from(cli.ldm));
|
||||
FIO_setAdaptiveMode(prefs, i32::from(cli.adapt));
|
||||
FIO_setRsyncable(prefs, i32::from(cli.rsyncable));
|
||||
@@ -1299,52 +1218,16 @@ unsafe fn run_decompress(
|
||||
dictionary,
|
||||
)
|
||||
},
|
||||
Operation::Compress | Operation::Bench => {
|
||||
unreachable!("compression and benchmark are dispatched separately")
|
||||
}
|
||||
Operation::Compress => unreachable!("compression is dispatched separately"),
|
||||
}
|
||||
}
|
||||
|
||||
/// Runs benchmark mode through the C bridge. Level clamping against
|
||||
/// `ZSTD_maxCLevel()` happens on the C side, where the symbol is always
|
||||
/// available when benchmarking is compiled in. No input file means a
|
||||
/// synthetic-sample benchmark, matching the C CLI.
|
||||
fn run_bench(cli: &Cli) -> Result<i32, String> {
|
||||
let inputs: Vec<*const c_char> = cli.inputs.iter().map(|value| value.as_ptr()).collect();
|
||||
let dictionary = cli
|
||||
.dictionary
|
||||
.as_ref()
|
||||
.map_or(ptr::null(), |value| value.as_ptr());
|
||||
let result = unsafe {
|
||||
ZSTD_rust_cli_bench(
|
||||
inputs.as_ptr(),
|
||||
inputs.len() as c_uint,
|
||||
dictionary,
|
||||
cli.level,
|
||||
cli.bench_end_level.unwrap_or(cli.level),
|
||||
&cli.compression_params,
|
||||
cli.display_level,
|
||||
cli.bench_nb_seconds.unwrap_or(DEFAULT_BENCH_NB_SECONDS),
|
||||
cli.block_size.unwrap_or(0),
|
||||
// The C CLI benchmarks single-threaded unless -T was given.
|
||||
cli.workers.unwrap_or(1),
|
||||
)
|
||||
};
|
||||
if result < 0 {
|
||||
return Err("benchmark mode is not available in this build".to_owned());
|
||||
}
|
||||
Ok(result)
|
||||
}
|
||||
|
||||
fn run_cli(mut cli: Cli) -> Result<i32, String> {
|
||||
if let Some(program_name) = &cli.unsupported_program {
|
||||
return Err(format!(
|
||||
"{program_name} compatibility mode is not yet implemented by the Rust CLI frontend"
|
||||
));
|
||||
}
|
||||
if cli.operation == Operation::Bench {
|
||||
return run_bench(&cli);
|
||||
}
|
||||
let explicit_input_count = cli.inputs.len();
|
||||
filter_symlink_inputs(&mut cli);
|
||||
if explicit_input_count > 0 && cli.inputs.is_empty() {
|
||||
@@ -1451,7 +1334,6 @@ fn run_cli(mut cli: Cli) -> Result<i32, String> {
|
||||
#[cfg(not(feature = "decompression"))]
|
||||
unreachable!("unsupported decompression was rejected above")
|
||||
}
|
||||
Operation::Bench => unreachable!("benchmark mode was dispatched earlier"),
|
||||
}
|
||||
};
|
||||
|
||||
@@ -1620,36 +1502,6 @@ mod tests {
|
||||
let cli = parse(&["zstdmt", "input"]);
|
||||
|
||||
assert_eq!(cli.workers, Some(0));
|
||||
assert!(!cli.single_thread);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn single_thread_pins_zero_workers() {
|
||||
let cli = parse(&["zstd", "--single-thread", "input"]);
|
||||
|
||||
assert_eq!(cli.workers, Some(0));
|
||||
assert!(cli.single_thread);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn a_later_thread_count_overrides_single_thread_workers() {
|
||||
/* Mirrors the C CLI: -T after --single-thread wins the worker count,
|
||||
* while the single-thread latch stays set. */
|
||||
let cli = parse(&["zstd", "--single-thread", "-T2", "input"]);
|
||||
|
||||
assert_eq!(cli.workers, Some(2));
|
||||
assert!(cli.single_thread);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn single_thread_rejects_attached_values() {
|
||||
let error = parse_args(vec![
|
||||
OsString::from("zstd"),
|
||||
OsString::from("--single-thread=1"),
|
||||
])
|
||||
.expect_err("an attached value must not activate --single-thread");
|
||||
|
||||
assert!(error.contains("does not take an argument"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
@@ -1686,41 +1538,4 @@ mod tests {
|
||||
|
||||
assert!(error.contains("not yet implemented"));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn bench_mode_parses_level_duration_and_defaults() {
|
||||
let cli = parse(&["zstd", "-b1", "-i0", "input"]);
|
||||
|
||||
assert_eq!(cli.operation, Operation::Bench);
|
||||
assert_eq!(cli.level, 1);
|
||||
assert_eq!(cli.bench_nb_seconds, Some(0));
|
||||
assert_eq!(cli.bench_end_level, None);
|
||||
assert_eq!(
|
||||
cli.inputs
|
||||
.iter()
|
||||
.map(|input| input.as_bytes())
|
||||
.collect::<Vec<_>>(),
|
||||
vec![&b"input"[..]]
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn bench_range_aggregates_within_a_single_argument() {
|
||||
let cli = parse(&["zstd", "-b5e6i2", "input"]);
|
||||
|
||||
assert_eq!(cli.operation, Operation::Bench);
|
||||
assert_eq!(cli.level, 5);
|
||||
assert_eq!(cli.bench_end_level, Some(6));
|
||||
assert_eq!(cli.bench_nb_seconds, Some(2));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn bench_duration_without_digits_defaults_to_zero() {
|
||||
let cli = parse(&["zstd", "-b", "-e", "-i"]);
|
||||
|
||||
assert_eq!(cli.operation, Operation::Bench);
|
||||
assert_eq!(cli.level, DEFAULT_CLEVEL);
|
||||
assert_eq!(cli.bench_end_level, Some(0));
|
||||
assert_eq!(cli.bench_nb_seconds, Some(0));
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,1048 +0,0 @@
|
||||
#![allow(non_camel_case_types)]
|
||||
#![allow(non_snake_case)]
|
||||
#![allow(clippy::missing_safety_doc)]
|
||||
#![allow(clippy::too_many_arguments)]
|
||||
|
||||
//! Context-free compression-parameter selection and sizing leaves.
|
||||
//!
|
||||
//! This module deliberately does **not** own the public `ZSTD_*` symbols yet.
|
||||
//! `zstd_compress.c` still owns configuration-sensitive policy: excluded block
|
||||
//! compressors, private `ZSTD_CCtx_params` layouts, LDM workspace sizing, and
|
||||
//! ASAN workspace policy. The C integration layer can select a raw table
|
||||
//! entry here, apply its configured strategy cascade, then use the adjustment
|
||||
//! and sizing leaves below. That keeps the Rust implementation independent of
|
||||
//! C preprocessor state while retaining byte-for-byte C policy for reduced
|
||||
//! builds.
|
||||
|
||||
use crate::errors::{ZstdErrorCode, ERROR};
|
||||
use std::mem::size_of;
|
||||
use std::os::raw::c_int;
|
||||
|
||||
pub const ZSTD_CONTENTSIZE_UNKNOWN: u64 = u64::MAX;
|
||||
|
||||
const ZSTD_CLEVEL_DEFAULT: c_int = 3;
|
||||
const ZSTD_MAX_CLEVEL: c_int = 22;
|
||||
const ZSTD_TARGETLENGTH_MAX: c_int = 1 << 17;
|
||||
const ZSTD_BLOCKSIZE_MAX: usize = 1 << 17;
|
||||
|
||||
const ZSTD_WINDOWLOG_MIN: c_int = 10;
|
||||
#[cfg(target_pointer_width = "32")]
|
||||
const ZSTD_WINDOWLOG_MAX: c_int = 30;
|
||||
#[cfg(not(target_pointer_width = "32"))]
|
||||
const ZSTD_WINDOWLOG_MAX: c_int = 31;
|
||||
const ZSTD_HASHLOG_MIN: c_int = 6;
|
||||
const ZSTD_HASHLOG_MAX: c_int = 30;
|
||||
const ZSTD_CHAINLOG_MIN: c_int = ZSTD_HASHLOG_MIN;
|
||||
#[cfg(target_pointer_width = "32")]
|
||||
const ZSTD_CHAINLOG_MAX: c_int = 29;
|
||||
#[cfg(not(target_pointer_width = "32"))]
|
||||
const ZSTD_CHAINLOG_MAX: c_int = 30;
|
||||
const ZSTD_SEARCHLOG_MIN: c_int = 1;
|
||||
const ZSTD_SEARCHLOG_MAX: c_int = ZSTD_WINDOWLOG_MAX - 1;
|
||||
const ZSTD_MINMATCH_MIN: c_int = 3;
|
||||
const ZSTD_MINMATCH_MAX: c_int = 7;
|
||||
const ZSTD_TARGETLENGTH_MIN: c_int = 0;
|
||||
|
||||
const ZSTD_FAST: c_int = 1;
|
||||
const ZSTD_DFAST: c_int = 2;
|
||||
const ZSTD_GREEDY: c_int = 3;
|
||||
const ZSTD_LAZY: c_int = 4;
|
||||
const ZSTD_LAZY2: c_int = 5;
|
||||
const ZSTD_BTLAZY2: c_int = 6;
|
||||
const ZSTD_BTOPT: c_int = 7;
|
||||
const ZSTD_BTULTRA: c_int = 8;
|
||||
const ZSTD_BTULTRA2: c_int = 9;
|
||||
|
||||
const ZSTD_C_COMPRESSION_LEVEL: c_int = 100;
|
||||
const ZSTD_C_WINDOW_LOG: c_int = 101;
|
||||
const ZSTD_C_HASH_LOG: c_int = 102;
|
||||
const ZSTD_C_CHAIN_LOG: c_int = 103;
|
||||
const ZSTD_C_SEARCH_LOG: c_int = 104;
|
||||
const ZSTD_C_MIN_MATCH: c_int = 105;
|
||||
const ZSTD_C_TARGET_LENGTH: c_int = 106;
|
||||
const ZSTD_C_STRATEGY: c_int = 107;
|
||||
|
||||
/// Private compression-parameter modes from `zstd_compress_internal.h`.
|
||||
///
|
||||
/// They are ABI-compatible with `ZSTD_CParamMode_e` and intentionally kept as
|
||||
/// integer constants because that enum remains private C API for now.
|
||||
pub const ZSTD_RUST_CPM_NO_ATTACH_DICT: c_int = 0;
|
||||
pub const ZSTD_RUST_CPM_ATTACH_DICT: c_int = 1;
|
||||
pub const ZSTD_RUST_CPM_CREATE_CDICT: c_int = 2;
|
||||
pub const ZSTD_RUST_CPM_UNKNOWN: c_int = 3;
|
||||
|
||||
/// `ZSTD_ParamSwitch_e` values used by the adjustment and sizing leaves.
|
||||
pub const ZSTD_RUST_PS_AUTO: c_int = 0;
|
||||
pub const ZSTD_RUST_PS_ENABLE: c_int = 1;
|
||||
pub const ZSTD_RUST_PS_DISABLE: c_int = 2;
|
||||
|
||||
/// ABI-compatible `ZSTD_compressionParameters` from `zstd.h`.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq)]
|
||||
pub struct ZSTD_compressionParameters {
|
||||
pub windowLog: u32,
|
||||
pub chainLog: u32,
|
||||
pub hashLog: u32,
|
||||
pub searchLog: u32,
|
||||
pub minMatch: u32,
|
||||
pub targetLength: u32,
|
||||
pub strategy: c_int,
|
||||
}
|
||||
|
||||
/// ABI-compatible `ZSTD_frameParameters` from `zstd.h`.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq)]
|
||||
pub struct ZSTD_frameParameters {
|
||||
pub contentSizeFlag: c_int,
|
||||
pub checksumFlag: c_int,
|
||||
pub noDictIDFlag: c_int,
|
||||
}
|
||||
|
||||
/// ABI-compatible `ZSTD_parameters` from `zstd.h`.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq)]
|
||||
pub struct ZSTD_parameters {
|
||||
pub cParams: ZSTD_compressionParameters,
|
||||
pub fParams: ZSTD_frameParameters,
|
||||
}
|
||||
|
||||
/// ABI-compatible `ZSTD_bounds` from `zstd.h`.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug, Default, PartialEq, Eq)]
|
||||
pub struct ZSTD_bounds {
|
||||
pub error: usize,
|
||||
pub lowerBound: c_int,
|
||||
pub upperBound: c_int,
|
||||
}
|
||||
|
||||
/* `clevels.h`, represented as raw fields so the table stays compact and easy
|
||||
* to compare mechanically against its C source. Field order is W, C, H, S,
|
||||
* L, TL, strategy. */
|
||||
const DEFAULT_CPARAMS: [[[u32; 7]; 23]; 4] = [
|
||||
[
|
||||
[19, 12, 13, 1, 6, 1, ZSTD_FAST as u32],
|
||||
[19, 13, 14, 1, 7, 0, ZSTD_FAST as u32],
|
||||
[20, 15, 16, 1, 6, 0, ZSTD_FAST as u32],
|
||||
[21, 16, 17, 1, 5, 0, ZSTD_DFAST as u32],
|
||||
[21, 18, 18, 1, 5, 0, ZSTD_DFAST as u32],
|
||||
[21, 18, 19, 3, 5, 2, ZSTD_GREEDY as u32],
|
||||
[21, 18, 19, 3, 5, 4, ZSTD_LAZY as u32],
|
||||
[21, 19, 20, 4, 5, 8, ZSTD_LAZY as u32],
|
||||
[21, 19, 20, 4, 5, 16, ZSTD_LAZY2 as u32],
|
||||
[22, 20, 21, 4, 5, 16, ZSTD_LAZY2 as u32],
|
||||
[22, 21, 22, 5, 5, 16, ZSTD_LAZY2 as u32],
|
||||
[22, 21, 22, 6, 5, 16, ZSTD_LAZY2 as u32],
|
||||
[22, 22, 23, 6, 5, 32, ZSTD_LAZY2 as u32],
|
||||
[22, 22, 22, 4, 5, 32, ZSTD_BTLAZY2 as u32],
|
||||
[22, 22, 23, 5, 5, 32, ZSTD_BTLAZY2 as u32],
|
||||
[22, 23, 23, 6, 5, 32, ZSTD_BTLAZY2 as u32],
|
||||
[22, 22, 22, 5, 5, 48, ZSTD_BTOPT as u32],
|
||||
[23, 23, 22, 5, 4, 64, ZSTD_BTOPT as u32],
|
||||
[23, 23, 22, 6, 3, 64, ZSTD_BTULTRA as u32],
|
||||
[23, 24, 22, 7, 3, 256, ZSTD_BTULTRA2 as u32],
|
||||
[25, 25, 23, 7, 3, 256, ZSTD_BTULTRA2 as u32],
|
||||
[26, 26, 24, 7, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[27, 27, 25, 9, 3, 999, ZSTD_BTULTRA2 as u32],
|
||||
],
|
||||
[
|
||||
[18, 12, 13, 1, 5, 1, ZSTD_FAST as u32],
|
||||
[18, 13, 14, 1, 6, 0, ZSTD_FAST as u32],
|
||||
[18, 14, 14, 1, 5, 0, ZSTD_DFAST as u32],
|
||||
[18, 16, 16, 1, 4, 0, ZSTD_DFAST as u32],
|
||||
[18, 16, 17, 3, 5, 2, ZSTD_GREEDY as u32],
|
||||
[18, 17, 18, 5, 5, 2, ZSTD_GREEDY as u32],
|
||||
[18, 18, 19, 3, 5, 4, ZSTD_LAZY as u32],
|
||||
[18, 18, 19, 4, 4, 4, ZSTD_LAZY as u32],
|
||||
[18, 18, 19, 4, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[18, 18, 19, 5, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[18, 18, 19, 6, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[18, 18, 19, 5, 4, 12, ZSTD_BTLAZY2 as u32],
|
||||
[18, 19, 19, 7, 4, 12, ZSTD_BTLAZY2 as u32],
|
||||
[18, 18, 19, 4, 4, 16, ZSTD_BTOPT as u32],
|
||||
[18, 18, 19, 4, 3, 32, ZSTD_BTOPT as u32],
|
||||
[18, 18, 19, 6, 3, 128, ZSTD_BTOPT as u32],
|
||||
[18, 19, 19, 6, 3, 128, ZSTD_BTULTRA as u32],
|
||||
[18, 19, 19, 8, 3, 256, ZSTD_BTULTRA as u32],
|
||||
[18, 19, 19, 6, 3, 128, ZSTD_BTULTRA2 as u32],
|
||||
[18, 19, 19, 8, 3, 256, ZSTD_BTULTRA2 as u32],
|
||||
[18, 19, 19, 10, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[18, 19, 19, 12, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[18, 19, 19, 13, 3, 999, ZSTD_BTULTRA2 as u32],
|
||||
],
|
||||
[
|
||||
[17, 12, 12, 1, 5, 1, ZSTD_FAST as u32],
|
||||
[17, 12, 13, 1, 6, 0, ZSTD_FAST as u32],
|
||||
[17, 13, 15, 1, 5, 0, ZSTD_FAST as u32],
|
||||
[17, 15, 16, 2, 5, 0, ZSTD_DFAST as u32],
|
||||
[17, 17, 17, 2, 4, 0, ZSTD_DFAST as u32],
|
||||
[17, 16, 17, 3, 4, 2, ZSTD_GREEDY as u32],
|
||||
[17, 16, 17, 3, 4, 4, ZSTD_LAZY as u32],
|
||||
[17, 16, 17, 3, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[17, 16, 17, 4, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[17, 16, 17, 5, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[17, 16, 17, 6, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[17, 17, 17, 5, 4, 8, ZSTD_BTLAZY2 as u32],
|
||||
[17, 18, 17, 7, 4, 12, ZSTD_BTLAZY2 as u32],
|
||||
[17, 18, 17, 3, 4, 12, ZSTD_BTOPT as u32],
|
||||
[17, 18, 17, 4, 3, 32, ZSTD_BTOPT as u32],
|
||||
[17, 18, 17, 6, 3, 256, ZSTD_BTOPT as u32],
|
||||
[17, 18, 17, 6, 3, 128, ZSTD_BTULTRA as u32],
|
||||
[17, 18, 17, 8, 3, 256, ZSTD_BTULTRA as u32],
|
||||
[17, 18, 17, 10, 3, 512, ZSTD_BTULTRA as u32],
|
||||
[17, 18, 17, 5, 3, 256, ZSTD_BTULTRA2 as u32],
|
||||
[17, 18, 17, 7, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[17, 18, 17, 9, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[17, 18, 17, 11, 3, 999, ZSTD_BTULTRA2 as u32],
|
||||
],
|
||||
[
|
||||
[14, 12, 13, 1, 5, 1, ZSTD_FAST as u32],
|
||||
[14, 14, 15, 1, 5, 0, ZSTD_FAST as u32],
|
||||
[14, 14, 15, 1, 4, 0, ZSTD_FAST as u32],
|
||||
[14, 14, 15, 2, 4, 0, ZSTD_DFAST as u32],
|
||||
[14, 14, 14, 4, 4, 2, ZSTD_GREEDY as u32],
|
||||
[14, 14, 14, 3, 4, 4, ZSTD_LAZY as u32],
|
||||
[14, 14, 14, 4, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[14, 14, 14, 6, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[14, 14, 14, 8, 4, 8, ZSTD_LAZY2 as u32],
|
||||
[14, 15, 14, 5, 4, 8, ZSTD_BTLAZY2 as u32],
|
||||
[14, 15, 14, 9, 4, 8, ZSTD_BTLAZY2 as u32],
|
||||
[14, 15, 14, 3, 4, 12, ZSTD_BTOPT as u32],
|
||||
[14, 15, 14, 4, 3, 24, ZSTD_BTOPT as u32],
|
||||
[14, 15, 14, 5, 3, 32, ZSTD_BTULTRA as u32],
|
||||
[14, 15, 15, 6, 3, 64, ZSTD_BTULTRA as u32],
|
||||
[14, 15, 15, 7, 3, 256, ZSTD_BTULTRA as u32],
|
||||
[14, 15, 15, 5, 3, 48, ZSTD_BTULTRA2 as u32],
|
||||
[14, 15, 15, 6, 3, 128, ZSTD_BTULTRA2 as u32],
|
||||
[14, 15, 15, 7, 3, 256, ZSTD_BTULTRA2 as u32],
|
||||
[14, 15, 15, 8, 3, 256, ZSTD_BTULTRA2 as u32],
|
||||
[14, 15, 15, 8, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[14, 15, 15, 9, 3, 512, ZSTD_BTULTRA2 as u32],
|
||||
[14, 15, 15, 10, 3, 999, ZSTD_BTULTRA2 as u32],
|
||||
],
|
||||
];
|
||||
|
||||
#[inline]
|
||||
fn cparams_from_row(row: [u32; 7]) -> ZSTD_compressionParameters {
|
||||
ZSTD_compressionParameters {
|
||||
windowLog: row[0],
|
||||
chainLog: row[1],
|
||||
hashLog: row[2],
|
||||
searchLog: row[3],
|
||||
minMatch: row[4],
|
||||
targetLength: row[5],
|
||||
strategy: row[6] as c_int,
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn bounds(param: c_int) -> ZSTD_bounds {
|
||||
let (lowerBound, upperBound) = match param {
|
||||
ZSTD_C_COMPRESSION_LEVEL => (ZSTD_rust_params_minCLevel(), ZSTD_rust_params_maxCLevel()),
|
||||
ZSTD_C_WINDOW_LOG => (ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX),
|
||||
ZSTD_C_HASH_LOG => (ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX),
|
||||
ZSTD_C_CHAIN_LOG => (ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX),
|
||||
ZSTD_C_SEARCH_LOG => (ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX),
|
||||
ZSTD_C_MIN_MATCH => (ZSTD_MINMATCH_MIN, ZSTD_MINMATCH_MAX),
|
||||
ZSTD_C_TARGET_LENGTH => (ZSTD_TARGETLENGTH_MIN, ZSTD_TARGETLENGTH_MAX),
|
||||
ZSTD_C_STRATEGY => (ZSTD_FAST, ZSTD_BTULTRA2),
|
||||
_ => {
|
||||
return ZSTD_bounds {
|
||||
error: ERROR(ZstdErrorCode::ParameterUnsupported),
|
||||
lowerBound: 0,
|
||||
upperBound: 0,
|
||||
}
|
||||
}
|
||||
};
|
||||
ZSTD_bounds {
|
||||
error: 0,
|
||||
lowerBound,
|
||||
upperBound,
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn within_bounds(param: c_int, value: c_int) -> bool {
|
||||
let bounds = bounds(param);
|
||||
bounds.error == 0 && value >= bounds.lowerBound && value <= bounds.upperBound
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn clamp_unsigned(value: u32, param: c_int) -> u32 {
|
||||
let bounds = bounds(param);
|
||||
debug_assert_eq!(bounds.error, 0);
|
||||
let signed = value as c_int;
|
||||
if signed < bounds.lowerBound {
|
||||
bounds.lowerBound as u32
|
||||
} else if signed > bounds.upperBound {
|
||||
bounds.upperBound as u32
|
||||
} else {
|
||||
value
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn clamp_cparams(mut cparams: ZSTD_compressionParameters) -> ZSTD_compressionParameters {
|
||||
cparams.windowLog = clamp_unsigned(cparams.windowLog, ZSTD_C_WINDOW_LOG);
|
||||
cparams.chainLog = clamp_unsigned(cparams.chainLog, ZSTD_C_CHAIN_LOG);
|
||||
cparams.hashLog = clamp_unsigned(cparams.hashLog, ZSTD_C_HASH_LOG);
|
||||
cparams.searchLog = clamp_unsigned(cparams.searchLog, ZSTD_C_SEARCH_LOG);
|
||||
cparams.minMatch = clamp_unsigned(cparams.minMatch, ZSTD_C_MIN_MATCH);
|
||||
cparams.targetLength = clamp_unsigned(cparams.targetLength, ZSTD_C_TARGET_LENGTH);
|
||||
let strategy_bounds = bounds(ZSTD_C_STRATEGY);
|
||||
if cparams.strategy < strategy_bounds.lowerBound {
|
||||
cparams.strategy = strategy_bounds.lowerBound;
|
||||
} else if cparams.strategy > strategy_bounds.upperBound {
|
||||
cparams.strategy = strategy_bounds.upperBound;
|
||||
}
|
||||
cparams
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn highbit32(value: u32) -> u32 {
|
||||
debug_assert_ne!(value, 0);
|
||||
u32::BITS - 1 - value.leading_zeros()
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn cycle_log(hash_log: u32, strategy: c_int) -> u32 {
|
||||
hash_log.wrapping_sub((strategy >= ZSTD_BTLAZY2) as u32)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn strategy_supports_row_match_finder(strategy: c_int) -> bool {
|
||||
(ZSTD_GREEDY..=ZSTD_LAZY2).contains(&strategy)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn row_match_finder_used(strategy: c_int, mode: c_int) -> bool {
|
||||
debug_assert_ne!(mode, ZSTD_RUST_PS_AUTO);
|
||||
strategy_supports_row_match_finder(strategy) && mode == ZSTD_RUST_PS_ENABLE
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn resolve_row_match_finder(mode: c_int, cparams: ZSTD_compressionParameters) -> c_int {
|
||||
if mode != ZSTD_RUST_PS_AUTO {
|
||||
return mode;
|
||||
}
|
||||
if strategy_supports_row_match_finder(cparams.strategy) && cparams.windowLog > 14 {
|
||||
ZSTD_RUST_PS_ENABLE
|
||||
} else {
|
||||
ZSTD_RUST_PS_DISABLE
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn dict_and_window_log(window_log: u32, src_size: u64, dict_size: u64) -> u32 {
|
||||
/* 1ULL << ZSTD_WINDOWLOG_MAX, which is smaller for 32-bit builds. */
|
||||
const MAX_WINDOW_SIZE: u64 = 1u64 << ZSTD_WINDOWLOG_MAX;
|
||||
if dict_size == 0 {
|
||||
return window_log;
|
||||
}
|
||||
debug_assert!(window_log <= ZSTD_WINDOWLOG_MAX as u32);
|
||||
debug_assert_ne!(src_size, ZSTD_CONTENTSIZE_UNKNOWN);
|
||||
let window_size = 1u64 << window_log;
|
||||
if window_size >= dict_size.wrapping_add(src_size) {
|
||||
window_log
|
||||
} else {
|
||||
let dict_and_window_size = dict_size.wrapping_add(window_size);
|
||||
if dict_and_window_size >= MAX_WINDOW_SIZE {
|
||||
ZSTD_WINDOWLOG_MAX as u32
|
||||
} else {
|
||||
highbit32((dict_and_window_size as u32).wrapping_sub(1)) + 1
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// Adjusts a *validated, configuration-resolved* parameter set.
|
||||
///
|
||||
/// This is the direct Rust leaf for `ZSTD_adjustCParams_internal()` after the
|
||||
/// C wrapper has applied its `ZSTD_EXCLUDE_*_BLOCK_COMPRESSOR` cascade. In
|
||||
/// particular, it does not select a fallback strategy itself. `srcSize == 0`
|
||||
/// means a known empty input here, just as it does in the C internal helper;
|
||||
/// the public C wrapper must translate zero to `ZSTD_CONTENTSIZE_UNKNOWN`.
|
||||
fn adjust_cparams(
|
||||
mut cparams: ZSTD_compressionParameters,
|
||||
mut src_size: u64,
|
||||
mut dict_size: usize,
|
||||
mode: c_int,
|
||||
mut use_row_match_finder: c_int,
|
||||
) -> ZSTD_compressionParameters {
|
||||
debug_assert_eq!(check_cparams(cparams), 0);
|
||||
|
||||
const MIN_SRC_SIZE: u64 = 513;
|
||||
/* 1ULL << (ZSTD_WINDOWLOG_MAX - 1), which is smaller for 32-bit builds. */
|
||||
const MAX_WINDOW_RESIZE: u64 = 1u64 << (ZSTD_WINDOWLOG_MAX - 1);
|
||||
|
||||
match mode {
|
||||
ZSTD_RUST_CPM_UNKNOWN | ZSTD_RUST_CPM_NO_ATTACH_DICT => {}
|
||||
ZSTD_RUST_CPM_CREATE_CDICT => {
|
||||
if dict_size != 0 && src_size == ZSTD_CONTENTSIZE_UNKNOWN {
|
||||
src_size = MIN_SRC_SIZE;
|
||||
}
|
||||
}
|
||||
ZSTD_RUST_CPM_ATTACH_DICT => dict_size = 0,
|
||||
_ => debug_assert!(false, "invalid ZSTD_CParamMode_e"),
|
||||
}
|
||||
|
||||
if src_size <= MAX_WINDOW_RESIZE && (dict_size as u64) <= MAX_WINDOW_RESIZE {
|
||||
let total_size = src_size.wrapping_add(dict_size as u64) as u32;
|
||||
let hash_size_min = 1u32 << ZSTD_HASHLOG_MIN;
|
||||
let src_log = if total_size < hash_size_min {
|
||||
ZSTD_HASHLOG_MIN as u32
|
||||
} else {
|
||||
highbit32(total_size.wrapping_sub(1)) + 1
|
||||
};
|
||||
cparams.windowLog = cparams.windowLog.min(src_log);
|
||||
}
|
||||
|
||||
if src_size != ZSTD_CONTENTSIZE_UNKNOWN {
|
||||
let dict_and_window_log =
|
||||
dict_and_window_log(cparams.windowLog, src_size, dict_size as u64);
|
||||
let cycle_log = cycle_log(cparams.chainLog, cparams.strategy);
|
||||
cparams.hashLog = cparams.hashLog.min(dict_and_window_log + 1);
|
||||
if cycle_log > dict_and_window_log {
|
||||
cparams.chainLog = cparams
|
||||
.chainLog
|
||||
.wrapping_sub(cycle_log - dict_and_window_log);
|
||||
}
|
||||
}
|
||||
|
||||
cparams.windowLog = cparams.windowLog.max(ZSTD_WINDOWLOG_MIN as u32);
|
||||
|
||||
/* The short-cache tags used for fast and dfast CDicts consume eight bits
|
||||
* of each 32-bit index. This is a fixed private-header constant today;
|
||||
* if it becomes configurable, the C shim must pass it as an explicit
|
||||
* adjustment input rather than silently changing this leaf. */
|
||||
if mode == ZSTD_RUST_CPM_CREATE_CDICT
|
||||
&& (cparams.strategy == ZSTD_FAST || cparams.strategy == ZSTD_DFAST)
|
||||
{
|
||||
const SHORT_CACHE_TAG_BITS: u32 = 8;
|
||||
let max_short_cache_hash_log = 32 - SHORT_CACHE_TAG_BITS;
|
||||
cparams.hashLog = cparams.hashLog.min(max_short_cache_hash_log);
|
||||
cparams.chainLog = cparams.chainLog.min(max_short_cache_hash_log);
|
||||
}
|
||||
|
||||
if use_row_match_finder == ZSTD_RUST_PS_AUTO {
|
||||
use_row_match_finder = ZSTD_RUST_PS_ENABLE;
|
||||
}
|
||||
if row_match_finder_used(cparams.strategy, use_row_match_finder) {
|
||||
const ROW_HASH_TAG_BITS: u32 = 8;
|
||||
let row_log = cparams.searchLog.clamp(4, 6);
|
||||
let max_hash_log = (32 - ROW_HASH_TAG_BITS) + row_log;
|
||||
debug_assert!(cparams.hashLog >= row_log);
|
||||
cparams.hashLog = cparams.hashLog.min(max_hash_log);
|
||||
}
|
||||
|
||||
cparams
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn get_cparam_row_size(src_size_hint: u64, dict_size: usize, mode: c_int) -> u64 {
|
||||
let mut dict_size = dict_size as u64;
|
||||
match mode {
|
||||
ZSTD_RUST_CPM_UNKNOWN | ZSTD_RUST_CPM_NO_ATTACH_DICT | ZSTD_RUST_CPM_CREATE_CDICT => {}
|
||||
ZSTD_RUST_CPM_ATTACH_DICT => dict_size = 0,
|
||||
_ => debug_assert!(false, "invalid ZSTD_CParamMode_e"),
|
||||
}
|
||||
let unknown = src_size_hint == ZSTD_CONTENTSIZE_UNKNOWN;
|
||||
let added_size = if unknown && dict_size > 0 { 500 } else { 0 };
|
||||
if unknown && dict_size == 0 {
|
||||
ZSTD_CONTENTSIZE_UNKNOWN
|
||||
} else {
|
||||
src_size_hint
|
||||
.wrapping_add(dict_size)
|
||||
.wrapping_add(added_size)
|
||||
}
|
||||
}
|
||||
|
||||
/// Selects a raw compression-level-table entry, without strategy cascading or
|
||||
/// source/dictionary adjustment.
|
||||
///
|
||||
/// A C wrapper must apply its active excluded-compressor cascade to the
|
||||
/// returned strategy before calling [`ZSTD_rust_params_adjustCParams`].
|
||||
fn select_cparams(
|
||||
compression_level: c_int,
|
||||
src_size_hint: u64,
|
||||
dict_size: usize,
|
||||
mode: c_int,
|
||||
) -> ZSTD_compressionParameters {
|
||||
let row_size = get_cparam_row_size(src_size_hint, dict_size, mode);
|
||||
let table_id = usize::from(row_size <= 256 * 1024)
|
||||
+ usize::from(row_size <= 128 * 1024)
|
||||
+ usize::from(row_size <= 16 * 1024);
|
||||
let row = if compression_level == 0 {
|
||||
ZSTD_CLEVEL_DEFAULT
|
||||
} else if compression_level < 0 {
|
||||
0
|
||||
} else {
|
||||
compression_level.min(ZSTD_MAX_CLEVEL)
|
||||
} as usize;
|
||||
|
||||
let mut cparams = cparams_from_row(DEFAULT_CPARAMS[table_id][row]);
|
||||
if compression_level < 0 {
|
||||
let clamped = compression_level.max(ZSTD_rust_params_minCLevel());
|
||||
cparams.targetLength = (-clamped) as u32;
|
||||
}
|
||||
cparams
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn make_params(cparams: ZSTD_compressionParameters) -> ZSTD_parameters {
|
||||
ZSTD_parameters {
|
||||
cParams: cparams,
|
||||
fParams: ZSTD_frameParameters {
|
||||
contentSizeFlag: 1,
|
||||
checksumFlag: 0,
|
||||
noDictIDFlag: 0,
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn check_cparams(cparams: ZSTD_compressionParameters) -> usize {
|
||||
if !within_bounds(ZSTD_C_WINDOW_LOG, cparams.windowLog as c_int)
|
||||
|| !within_bounds(ZSTD_C_CHAIN_LOG, cparams.chainLog as c_int)
|
||||
|| !within_bounds(ZSTD_C_HASH_LOG, cparams.hashLog as c_int)
|
||||
|| !within_bounds(ZSTD_C_SEARCH_LOG, cparams.searchLog as c_int)
|
||||
|| !within_bounds(ZSTD_C_MIN_MATCH, cparams.minMatch as c_int)
|
||||
|| !within_bounds(ZSTD_C_TARGET_LENGTH, cparams.targetLength as c_int)
|
||||
|| !within_bounds(ZSTD_C_STRATEGY, cparams.strategy)
|
||||
{
|
||||
ERROR(ZstdErrorCode::ParameterOutOfBound)
|
||||
} else {
|
||||
0
|
||||
}
|
||||
}
|
||||
|
||||
/// Returns the highest table-backed compression level (`ZSTD_MAX_CLEVEL`).
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_maxCLevel() -> c_int {
|
||||
ZSTD_MAX_CLEVEL
|
||||
}
|
||||
|
||||
/// Returns the lowest public fast level (`-ZSTD_TARGETLENGTH_MAX`).
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_minCLevel() -> c_int {
|
||||
-ZSTD_TARGETLENGTH_MAX
|
||||
}
|
||||
|
||||
/// Returns the default compression level from `zstd.h`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_defaultCLevel() -> c_int {
|
||||
ZSTD_CLEVEL_DEFAULT
|
||||
}
|
||||
|
||||
/// Returns bounds for the seven core compression parameters and compression
|
||||
/// level. Other `ZSTD_cParameter` cases remain C-owned.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_getBounds(param: c_int) -> ZSTD_bounds {
|
||||
bounds(param)
|
||||
}
|
||||
|
||||
/// Checks the seven fields of `ZSTD_compressionParameters`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_checkCParams(cparams: ZSTD_compressionParameters) -> usize {
|
||||
check_cparams(cparams)
|
||||
}
|
||||
|
||||
/// Clamps the seven fields of `ZSTD_compressionParameters` to public bounds.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_clampCParams(
|
||||
cparams: ZSTD_compressionParameters,
|
||||
) -> ZSTD_compressionParameters {
|
||||
clamp_cparams(cparams)
|
||||
}
|
||||
|
||||
/// C ABI for `ZSTD_cycleLog()`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_cycleLog(hashLog: u32, strategy: c_int) -> u32 {
|
||||
cycle_log(hashLog, strategy)
|
||||
}
|
||||
|
||||
/// C ABI for private `ZSTD_getCParamRowSize()`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_getCParamRowSize(
|
||||
srcSizeHint: u64,
|
||||
dictSize: usize,
|
||||
mode: c_int,
|
||||
) -> u64 {
|
||||
get_cparam_row_size(srcSizeHint, dictSize, mode)
|
||||
}
|
||||
|
||||
/// Returns the unadjusted compression-level-table entry.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_selectCParams(
|
||||
compressionLevel: c_int,
|
||||
srcSizeHint: u64,
|
||||
dictSize: usize,
|
||||
mode: c_int,
|
||||
) -> ZSTD_compressionParameters {
|
||||
select_cparams(compressionLevel, srcSizeHint, dictSize, mode)
|
||||
}
|
||||
|
||||
/// Adjusts a validated, strategy-resolved C parameter set.
|
||||
///
|
||||
/// See [`adjust_cparams`] for the required C-side policy step.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_adjustCParams(
|
||||
cparams: ZSTD_compressionParameters,
|
||||
srcSize: u64,
|
||||
dictSize: usize,
|
||||
mode: c_int,
|
||||
useRowMatchFinder: c_int,
|
||||
) -> ZSTD_compressionParameters {
|
||||
adjust_cparams(cparams, srcSize, dictSize, mode, useRowMatchFinder)
|
||||
}
|
||||
|
||||
/// Builds `ZSTD_parameters` with the public default frame parameters.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_makeParams(
|
||||
cparams: ZSTD_compressionParameters,
|
||||
) -> ZSTD_parameters {
|
||||
make_params(cparams)
|
||||
}
|
||||
|
||||
/// Runtime sizes needed to reproduce `ZSTD_sizeof_matchState()` without
|
||||
/// exposing private C structures to Rust.
|
||||
///
|
||||
/// `asanRedzoneSize` is `ZSTD_CWKSP_ASAN_REDZONE_SIZE` when workspace
|
||||
/// poisoning is enabled and zero otherwise. `matchTSize` and `optimalTSize`
|
||||
/// must be `sizeof(ZSTD_match_t)` and `sizeof(ZSTD_optimal_t)` respectively.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug, Default)]
|
||||
pub struct ZSTD_rustMatchStateSizing {
|
||||
pub hashLog3Max: u32,
|
||||
pub matchTSize: usize,
|
||||
pub optimalTSize: usize,
|
||||
pub asanRedzoneSize: usize,
|
||||
}
|
||||
|
||||
/// C-only layout inputs for `ZSTD_estimateCDictSize_advanced()`.
|
||||
#[repr(C)]
|
||||
#[derive(Clone, Copy, Debug, Default)]
|
||||
pub struct ZSTD_rustCDictSizing {
|
||||
pub cdictSize: usize,
|
||||
pub hufWorkspaceSize: usize,
|
||||
pub hashLog3Max: u32,
|
||||
pub matchTSize: usize,
|
||||
pub optimalTSize: usize,
|
||||
pub asanRedzoneSize: usize,
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn cwksp_alloc_size(size: usize, asan_redzone_size: usize) -> usize {
|
||||
if size == 0 {
|
||||
0
|
||||
} else {
|
||||
size.wrapping_add(asan_redzone_size.wrapping_mul(2))
|
||||
}
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn cwksp_align(size: usize, alignment: usize) -> usize {
|
||||
debug_assert!(alignment.is_power_of_two());
|
||||
size.wrapping_add(alignment - 1) & !(alignment - 1)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn cwksp_aligned64_alloc_size(size: usize, asan_redzone_size: usize) -> usize {
|
||||
cwksp_alloc_size(cwksp_align(size, 64), asan_redzone_size)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn shift_size(log: u32) -> usize {
|
||||
debug_assert!(log < usize::BITS);
|
||||
1usize << log
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn allocate_chain_table(strategy: c_int, use_row_match_finder: c_int, for_dds_dict: bool) -> bool {
|
||||
for_dds_dict
|
||||
|| (strategy != ZSTD_FAST && !row_match_finder_used(strategy, use_row_match_finder))
|
||||
}
|
||||
|
||||
fn estimate_match_state_size(
|
||||
cparams: ZSTD_compressionParameters,
|
||||
use_row_match_finder: c_int,
|
||||
enable_dedicated_dict_search: bool,
|
||||
for_cctx: bool,
|
||||
sizing: ZSTD_rustMatchStateSizing,
|
||||
) -> usize {
|
||||
if check_cparams(cparams) != 0 || use_row_match_finder == ZSTD_RUST_PS_AUTO {
|
||||
return 0;
|
||||
}
|
||||
|
||||
let chain_size = if allocate_chain_table(
|
||||
cparams.strategy,
|
||||
use_row_match_finder,
|
||||
enable_dedicated_dict_search && !for_cctx,
|
||||
) {
|
||||
shift_size(cparams.chainLog)
|
||||
} else {
|
||||
0
|
||||
};
|
||||
let hash_size = shift_size(cparams.hashLog);
|
||||
let hash_log_3 = if for_cctx && cparams.minMatch == 3 {
|
||||
cparams.windowLog.min(sizing.hashLog3Max)
|
||||
} else {
|
||||
0
|
||||
};
|
||||
let hash3_size = if hash_log_3 == 0 {
|
||||
0
|
||||
} else {
|
||||
shift_size(hash_log_3)
|
||||
};
|
||||
let table_space = chain_size
|
||||
.wrapping_mul(size_of::<u32>())
|
||||
.wrapping_add(hash_size.wrapping_mul(size_of::<u32>()))
|
||||
.wrapping_add(hash3_size.wrapping_mul(size_of::<u32>()));
|
||||
|
||||
let redzone = sizing.asanRedzoneSize;
|
||||
let opt_potential_space = cwksp_aligned64_alloc_size((52 + 1) * size_of::<u32>(), redzone)
|
||||
.wrapping_add(cwksp_aligned64_alloc_size(
|
||||
(35 + 1) * size_of::<u32>(),
|
||||
redzone,
|
||||
))
|
||||
.wrapping_add(cwksp_aligned64_alloc_size(
|
||||
(31 + 1) * size_of::<u32>(),
|
||||
redzone,
|
||||
))
|
||||
.wrapping_add(cwksp_aligned64_alloc_size(
|
||||
(1 << 8) * size_of::<u32>(),
|
||||
redzone,
|
||||
))
|
||||
.wrapping_add(cwksp_aligned64_alloc_size(
|
||||
4099usize.wrapping_mul(sizing.matchTSize),
|
||||
redzone,
|
||||
))
|
||||
.wrapping_add(cwksp_aligned64_alloc_size(
|
||||
4099usize.wrapping_mul(sizing.optimalTSize),
|
||||
redzone,
|
||||
));
|
||||
let lazy_additional_space = if row_match_finder_used(cparams.strategy, use_row_match_finder) {
|
||||
cwksp_aligned64_alloc_size(hash_size, redzone)
|
||||
} else {
|
||||
0
|
||||
};
|
||||
let opt_space = if for_cctx && cparams.strategy >= ZSTD_BTOPT {
|
||||
opt_potential_space
|
||||
} else {
|
||||
0
|
||||
};
|
||||
let slack_space = 2 * 64;
|
||||
|
||||
table_space
|
||||
.wrapping_add(opt_space)
|
||||
.wrapping_add(slack_space)
|
||||
.wrapping_add(lazy_additional_space)
|
||||
}
|
||||
|
||||
/// Pure leaf for private `ZSTD_sizeof_matchState()`.
|
||||
///
|
||||
/// `useRowMatchFinder` must already be resolved to enable or disable. A NULL
|
||||
/// `sizing` pointer or invalid C parameters returns zero; C never supplies
|
||||
/// either in a valid estimator call.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTD_rust_params_estimateMatchStateSize(
|
||||
cparams: ZSTD_compressionParameters,
|
||||
useRowMatchFinder: c_int,
|
||||
enableDedicatedDictSearch: c_int,
|
||||
forCCtx: u32,
|
||||
sizing: *const ZSTD_rustMatchStateSizing,
|
||||
) -> usize {
|
||||
if sizing.is_null() {
|
||||
return 0;
|
||||
}
|
||||
let sizing = unsafe { *sizing };
|
||||
estimate_match_state_size(
|
||||
cparams,
|
||||
useRowMatchFinder,
|
||||
enableDedicatedDictSearch != 0,
|
||||
forCCtx != 0,
|
||||
sizing,
|
||||
)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
fn max_nb_seq(block_size: usize, min_match: u32, use_sequence_producer: bool) -> usize {
|
||||
let divider = if min_match == 3 || use_sequence_producer {
|
||||
3
|
||||
} else {
|
||||
4
|
||||
};
|
||||
block_size / divider
|
||||
}
|
||||
|
||||
/// Pure leaf for private `ZSTD_maxNbSeq()`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_maxNbSeq(
|
||||
blockSize: usize,
|
||||
minMatch: u32,
|
||||
useSequenceProducer: c_int,
|
||||
) -> usize {
|
||||
max_nb_seq(blockSize, minMatch, useSequenceProducer != 0)
|
||||
}
|
||||
|
||||
/// Pure leaf for private `ZSTD_resolveMaxBlockSize()`.
|
||||
#[no_mangle]
|
||||
pub extern "C" fn ZSTD_rust_params_resolveMaxBlockSize(maxBlockSize: usize) -> usize {
|
||||
if maxBlockSize == 0 {
|
||||
ZSTD_BLOCKSIZE_MAX
|
||||
} else {
|
||||
maxBlockSize
|
||||
}
|
||||
}
|
||||
|
||||
fn estimate_cdict_size_from_cparams(
|
||||
dict_size: usize,
|
||||
cparams: ZSTD_compressionParameters,
|
||||
dict_load_method: c_int,
|
||||
sizing: ZSTD_rustCDictSizing,
|
||||
) -> usize {
|
||||
if check_cparams(cparams) != 0 {
|
||||
return 0;
|
||||
}
|
||||
let match_state_sizing = ZSTD_rustMatchStateSizing {
|
||||
hashLog3Max: sizing.hashLog3Max,
|
||||
matchTSize: sizing.matchTSize,
|
||||
optimalTSize: sizing.optimalTSize,
|
||||
asanRedzoneSize: sizing.asanRedzoneSize,
|
||||
};
|
||||
let row_match_finder = resolve_row_match_finder(ZSTD_RUST_PS_AUTO, cparams);
|
||||
let copied_dict_space = if dict_load_method == 1 {
|
||||
0
|
||||
} else {
|
||||
cwksp_alloc_size(
|
||||
cwksp_align(dict_size, size_of::<usize>()),
|
||||
sizing.asanRedzoneSize,
|
||||
)
|
||||
};
|
||||
cwksp_alloc_size(sizing.cdictSize, sizing.asanRedzoneSize)
|
||||
.wrapping_add(cwksp_alloc_size(
|
||||
sizing.hufWorkspaceSize,
|
||||
sizing.asanRedzoneSize,
|
||||
))
|
||||
.wrapping_add(estimate_match_state_size(
|
||||
cparams,
|
||||
row_match_finder,
|
||||
true,
|
||||
false,
|
||||
match_state_sizing,
|
||||
))
|
||||
.wrapping_add(copied_dict_space)
|
||||
}
|
||||
|
||||
/// Parameter-only leaf for `ZSTD_estimateCDictSize_advanced()`.
|
||||
///
|
||||
/// C retains ownership of `sizeof(ZSTD_CDict)`, HUF workspace configuration,
|
||||
/// and sanitizer workspace policy, then passes them in `sizing`. A NULL sizing
|
||||
/// pointer returns zero.
|
||||
#[no_mangle]
|
||||
pub unsafe extern "C" fn ZSTD_rust_params_estimateCDictSizeFromCParams(
|
||||
dictSize: usize,
|
||||
cparams: ZSTD_compressionParameters,
|
||||
dictLoadMethod: c_int,
|
||||
sizing: *const ZSTD_rustCDictSizing,
|
||||
) -> usize {
|
||||
if sizing.is_null() {
|
||||
return 0;
|
||||
}
|
||||
estimate_cdict_size_from_cparams(dictSize, cparams, dictLoadMethod, unsafe { *sizing })
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
use super::*;
|
||||
use crate::errors::ERR_isError;
|
||||
use std::mem::{align_of, size_of};
|
||||
|
||||
fn adjust_public(
|
||||
cparams: ZSTD_compressionParameters,
|
||||
src_size: u64,
|
||||
dict_size: usize,
|
||||
) -> ZSTD_compressionParameters {
|
||||
let src_size = if src_size == 0 {
|
||||
ZSTD_CONTENTSIZE_UNKNOWN
|
||||
} else {
|
||||
src_size
|
||||
};
|
||||
adjust_cparams(
|
||||
clamp_cparams(cparams),
|
||||
src_size,
|
||||
dict_size,
|
||||
ZSTD_RUST_CPM_UNKNOWN,
|
||||
ZSTD_RUST_PS_AUTO,
|
||||
)
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn abi_parameter_layouts_match_zstd_h() {
|
||||
assert_eq!(
|
||||
size_of::<ZSTD_compressionParameters>(),
|
||||
7 * size_of::<u32>()
|
||||
);
|
||||
assert_eq!(align_of::<ZSTD_compressionParameters>(), align_of::<u32>());
|
||||
assert_eq!(size_of::<ZSTD_frameParameters>(), 3 * size_of::<c_int>());
|
||||
assert_eq!(size_of::<ZSTD_parameters>(), 10 * size_of::<u32>());
|
||||
assert_eq!(
|
||||
size_of::<ZSTD_bounds>(),
|
||||
size_of::<usize>() + 2 * size_of::<c_int>()
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn level_tables_match_representative_clevels_entries() {
|
||||
let large = select_cparams(3, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN);
|
||||
assert_eq!(
|
||||
large,
|
||||
ZSTD_compressionParameters {
|
||||
windowLog: 21,
|
||||
chainLog: 16,
|
||||
hashLog: 17,
|
||||
searchLog: 1,
|
||||
minMatch: 5,
|
||||
targetLength: 0,
|
||||
strategy: ZSTD_DFAST,
|
||||
}
|
||||
);
|
||||
|
||||
let small = select_cparams(3, 16 * 1024, 0, ZSTD_RUST_CPM_UNKNOWN);
|
||||
assert_eq!(
|
||||
small,
|
||||
ZSTD_compressionParameters {
|
||||
windowLog: 14,
|
||||
chainLog: 14,
|
||||
hashLog: 15,
|
||||
searchLog: 2,
|
||||
minMatch: 4,
|
||||
targetLength: 0,
|
||||
strategy: ZSTD_DFAST,
|
||||
}
|
||||
);
|
||||
|
||||
let fast = select_cparams(-5, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN);
|
||||
assert_eq!(fast.strategy, ZSTD_FAST);
|
||||
assert_eq!(fast.targetLength, 5);
|
||||
assert_eq!(
|
||||
select_cparams(999, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN),
|
||||
select_cparams(22, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN)
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn adjustment_clamps_and_downsizes_like_the_c_leaf() {
|
||||
let input = ZSTD_compressionParameters {
|
||||
windowLog: 31,
|
||||
chainLog: 30,
|
||||
hashLog: 30,
|
||||
searchLog: 1,
|
||||
minMatch: 3,
|
||||
targetLength: 0,
|
||||
strategy: ZSTD_BTOPT,
|
||||
};
|
||||
let adjusted = adjust_public(input, 1, 0);
|
||||
assert_eq!(adjusted.windowLog, 10);
|
||||
/* C applies WINDOWLOG_ABSOLUTEMIN only after it has used the smaller
|
||||
* temporary window to downsize the hash and chain logs. */
|
||||
assert_eq!(adjusted.hashLog, 7);
|
||||
assert_eq!(adjusted.chainLog, 7);
|
||||
assert_eq!(ZSTD_rust_params_checkCParams(adjusted), 0);
|
||||
|
||||
let clamped = ZSTD_rust_params_clampCParams(ZSTD_compressionParameters {
|
||||
windowLog: u32::MAX,
|
||||
chainLog: 0,
|
||||
hashLog: 0,
|
||||
searchLog: 0,
|
||||
minMatch: 0,
|
||||
targetLength: u32::MAX,
|
||||
strategy: 99,
|
||||
});
|
||||
assert_eq!(clamped.windowLog, ZSTD_WINDOWLOG_MIN as u32);
|
||||
assert_eq!(clamped.chainLog, ZSTD_CHAINLOG_MIN as u32);
|
||||
assert_eq!(clamped.hashLog, ZSTD_HASHLOG_MIN as u32);
|
||||
assert_eq!(clamped.searchLog, ZSTD_SEARCHLOG_MIN as u32);
|
||||
assert_eq!(clamped.minMatch, ZSTD_MINMATCH_MIN as u32);
|
||||
assert_eq!(clamped.targetLength, ZSTD_TARGETLENGTH_MIN as u32);
|
||||
assert_eq!(clamped.strategy, ZSTD_BTULTRA2);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn checking_and_bounds_preserve_the_public_error_contract() {
|
||||
let valid = select_cparams(1, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN);
|
||||
assert_eq!(check_cparams(valid), 0);
|
||||
let mut invalid = valid;
|
||||
invalid.minMatch = 2;
|
||||
assert!(ERR_isError(check_cparams(invalid)));
|
||||
let bounds = ZSTD_rust_params_getBounds(ZSTD_C_HASH_LOG);
|
||||
assert_eq!(bounds.error, 0);
|
||||
assert_eq!(bounds.lowerBound, 6);
|
||||
assert_eq!(bounds.upperBound, 30);
|
||||
assert!(ERR_isError(ZSTD_rust_params_getBounds(-1).error));
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn params_default_frame_flags_match_c() {
|
||||
let cparams = select_cparams(5, ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN);
|
||||
let params = ZSTD_rust_params_makeParams(cparams);
|
||||
assert_eq!(params.cParams, cparams);
|
||||
assert_eq!(params.fParams.contentSizeFlag, 1);
|
||||
assert_eq!(params.fParams.checksumFlag, 0);
|
||||
assert_eq!(params.fParams.noDictIDFlag, 0);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn row_size_preserves_unknown_dictionary_overflow_semantics() {
|
||||
assert_eq!(
|
||||
get_cparam_row_size(ZSTD_CONTENTSIZE_UNKNOWN, 0, ZSTD_RUST_CPM_UNKNOWN),
|
||||
ZSTD_CONTENTSIZE_UNKNOWN
|
||||
);
|
||||
assert_eq!(
|
||||
get_cparam_row_size(ZSTD_CONTENTSIZE_UNKNOWN, 1, ZSTD_RUST_CPM_UNKNOWN),
|
||||
500
|
||||
);
|
||||
assert_eq!(
|
||||
get_cparam_row_size(128 * 1024, 1, ZSTD_RUST_CPM_ATTACH_DICT),
|
||||
128 * 1024
|
||||
);
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn match_state_sizing_reproduces_table_and_row_rules() {
|
||||
let sizing = ZSTD_rustMatchStateSizing {
|
||||
hashLog3Max: 17,
|
||||
matchTSize: 16,
|
||||
optimalTSize: 32,
|
||||
asanRedzoneSize: 0,
|
||||
};
|
||||
let fast = ZSTD_compressionParameters {
|
||||
windowLog: 10,
|
||||
chainLog: 10,
|
||||
hashLog: 10,
|
||||
searchLog: 1,
|
||||
minMatch: 4,
|
||||
targetLength: 0,
|
||||
strategy: ZSTD_FAST,
|
||||
};
|
||||
assert_eq!(
|
||||
estimate_match_state_size(fast, ZSTD_RUST_PS_DISABLE, false, true, sizing),
|
||||
4096 + 128
|
||||
);
|
||||
|
||||
let row = ZSTD_compressionParameters {
|
||||
windowLog: 15,
|
||||
chainLog: 10,
|
||||
hashLog: 10,
|
||||
searchLog: 4,
|
||||
minMatch: 4,
|
||||
targetLength: 0,
|
||||
strategy: ZSTD_GREEDY,
|
||||
};
|
||||
assert_eq!(
|
||||
estimate_match_state_size(row, ZSTD_RUST_PS_ENABLE, false, true, sizing),
|
||||
4096 + 1024 + 128
|
||||
);
|
||||
assert_eq!(ZSTD_rust_params_maxNbSeq(100, 3, 0), 33);
|
||||
assert_eq!(ZSTD_rust_params_maxNbSeq(100, 4, 0), 25);
|
||||
assert_eq!(ZSTD_rust_params_resolveMaxBlockSize(0), 128 * 1024);
|
||||
}
|
||||
}
|
||||
+7
-49
@@ -65,24 +65,16 @@ endif
|
||||
endif
|
||||
|
||||
RUST_HUF_FEATURE :=
|
||||
RUST_HUF_MODE := default
|
||||
RUST_BUILD_CONFIG := default
|
||||
ifneq ($(RUST_HUF_FORCE_X1),0)
|
||||
RUST_HUF_FEATURE := huf-force-decompress-x1
|
||||
RUST_HUF_MODE := huf-force-decompress-x1
|
||||
RUST_BUILD_CONFIG := huf-force-decompress-x1
|
||||
endif
|
||||
ifneq ($(RUST_HUF_FORCE_X2),0)
|
||||
RUST_HUF_FEATURE := huf-force-decompress-x2
|
||||
RUST_HUF_MODE := huf-force-decompress-x2
|
||||
RUST_BUILD_CONFIG := huf-force-decompress-x2
|
||||
endif
|
||||
|
||||
# Test binaries compile every lib/legacy/*.c file regardless of the dispatch
|
||||
# level (ZSTDLEGACY_FILES is a plain wildcard below), so the Rust archive must
|
||||
# always carry all ported legacy decoders too. The build configuration still
|
||||
# encodes the level because the flat C objects bake -DZSTD_LEGACY_SUPPORT into
|
||||
# the dispatch code and must never outlive a level change.
|
||||
RUST_LEGACY_FEATURES := legacy-v01,legacy-v02,legacy-v03,legacy-v04,legacy-v05,legacy-v06,legacy-v07
|
||||
RUST_BUILD_CONFIG := $(RUST_HUF_MODE)-legacy$(ZSTD_LEGACY_SUPPORT)
|
||||
|
||||
RUST_TARGET_DIR := $(RUST_DIR)/target/$(RUST_BUILD_CONFIG)
|
||||
RUST_STATICLIB := $(RUST_TARGET_DIR)/release/libzstd_rs.a
|
||||
RUST_TARGET_32 ?= i686-unknown-linux-gnu
|
||||
@@ -94,7 +86,6 @@ RUST_CARGO_FLAGS := --manifest-path $(RUST_MANIFEST) --release \
|
||||
ifneq ($(RUST_HUF_FEATURE),)
|
||||
RUST_CARGO_FLAGS += --features $(RUST_HUF_FEATURE)
|
||||
endif
|
||||
RUST_CARGO_FLAGS += --features $(RUST_LEGACY_FEATURES)
|
||||
|
||||
$(RUST_STATICLIB): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_CARGO_FLAGS)
|
||||
@@ -102,28 +93,6 @@ $(RUST_STATICLIB): $(RUST_SOURCES)
|
||||
$(RUST_STATICLIB_32): $(RUST_SOURCES)
|
||||
$(CARGO) build $(RUST_CARGO_FLAGS) --target $(RUST_TARGET_32)
|
||||
|
||||
# Program-only helpers that were C sources shared with the tests (timefn,
|
||||
# benchfn) now live in the Rust CLI package. The tests link a helpers-only
|
||||
# archive, built without the `cli` feature: the parser/dispatch layer needs
|
||||
# the C fileio backend, which test binaries do not provide.
|
||||
RUST_CLI_DIR := $(RUST_DIR)/cli
|
||||
RUST_CLI_MANIFEST := $(RUST_CLI_DIR)/Cargo.toml
|
||||
RUST_CLI_HELPER_SOURCES := $(RUST_CLI_MANIFEST) $(RUST_CLI_DIR)/Cargo.lock \
|
||||
$(RUST_CLI_DIR)/src/lib.rs \
|
||||
$(RUST_DIR)/src/timefn.rs $(RUST_DIR)/src/benchfn.rs
|
||||
RUST_CLI_HELPERS_TARGET_DIR := $(RUST_DIR)/target/cli-helpers
|
||||
RUST_CLI_HELPERS_STATICLIB := $(RUST_CLI_HELPERS_TARGET_DIR)/release/libzstd_cli_rs.a
|
||||
RUST_CLI_HELPERS_STATICLIB_32 := $(RUST_CLI_HELPERS_TARGET_DIR)/$(RUST_TARGET_32)/release/libzstd_cli_rs.a
|
||||
RUST_CLI_HELPERS_CARGO_FLAGS := --manifest-path $(RUST_CLI_MANIFEST) --release \
|
||||
--target-dir $(RUST_CLI_HELPERS_TARGET_DIR) \
|
||||
--no-default-features
|
||||
|
||||
$(RUST_CLI_HELPERS_STATICLIB): $(RUST_CLI_HELPER_SOURCES)
|
||||
$(CARGO) build $(RUST_CLI_HELPERS_CARGO_FLAGS)
|
||||
|
||||
$(RUST_CLI_HELPERS_STATICLIB_32): $(RUST_CLI_HELPER_SOURCES)
|
||||
$(CARGO) build $(RUST_CLI_HELPERS_CARGO_FLAGS) --target $(RUST_TARGET_32)
|
||||
|
||||
# These test objects have flat filenames, unlike the configuration-hashed
|
||||
# program objects. Track the HUF mode separately so a C object set compiled
|
||||
# for one decoder is never relinked with a Rust archive for another decoder.
|
||||
@@ -283,8 +252,8 @@ fuzzer32 : $(ZSTD_FILES)
|
||||
$(LINK.c) $^ -o $@$(EXT)
|
||||
|
||||
# note : broken : requires symbols unavailable from dynamic library
|
||||
fuzzer-dll : $(LIB_SRCDIR)/common/xxhash.c $(PRGDIR)/util.c $(PRGDIR)/timefn.c $(PRGDIR)/datagen.c fuzzer.c $(RUST_CLI_HELPERS_STATICLIB)
|
||||
$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(RUST_CLI_HELPERS_STATICLIB) $(LDFLAGS) -o $@$(EXT)
|
||||
fuzzer-dll : $(LIB_SRCDIR)/common/xxhash.c $(PRGDIR)/util.c $(PRGDIR)/timefn.c $(PRGDIR)/datagen.c fuzzer.c
|
||||
$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(LDFLAGS) -o $@$(EXT)
|
||||
|
||||
CLEAN += zstreamtest zstreamtest32
|
||||
ZSTREAM_LOCAL_FILES := $(PRGDIR)/datagen.c $(PRGDIR)/util.c $(PRGDIR)/timefn.c seqgen.c zstreamtest.c external_matchfinder.c
|
||||
@@ -315,8 +284,8 @@ zstreamtest_ubsan : $(ZSTREAMFILES)
|
||||
|
||||
# note : broken : requires symbols unavailable from dynamic library
|
||||
zstreamtest-dll : $(LIB_SRCDIR)/common/xxhash.c # xxh symbols not exposed from dll
|
||||
zstreamtest-dll : $(ZSTREAM_LOCAL_FILES) $(RUST_CLI_HELPERS_STATICLIB)
|
||||
$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(RUST_CLI_HELPERS_STATICLIB) $(LDFLAGS) -o $@$(EXT)
|
||||
zstreamtest-dll : $(ZSTREAM_LOCAL_FILES)
|
||||
$(CC) $(CPPFLAGS) $(CFLAGS) $(filter %.c,$^) $(LDFLAGS) -o $@$(EXT)
|
||||
|
||||
CLEAN += paramgrill
|
||||
paramgrill : DEBUGFLAGS = # turn off debug for speed measurements
|
||||
@@ -375,17 +344,6 @@ $(RUST_LINK_TARGETS_32): $(RUST_STATICLIB_32)
|
||||
|
||||
$(RUST_LINK_TARGETS) $(RUST_LINK_TARGETS_32): $(RUST_HUF_C_MODE_STAMP)
|
||||
|
||||
# Tests that compile the timefn/benchfn C shims also link the Rust CLI
|
||||
# helpers archive, which owns those implementations. The archive is a
|
||||
# prerequisite so `$^` places it after every C object referencing its symbols.
|
||||
RUST_CLI_LINK_TARGETS := fullbench fullbench-lib fullbench-dll fuzzer \
|
||||
zstreamtest zstreamtest_asan zstreamtest_tsan \
|
||||
zstreamtest_ubsan paramgrill decodecorpus poolTests
|
||||
$(RUST_CLI_LINK_TARGETS): $(RUST_CLI_HELPERS_STATICLIB)
|
||||
|
||||
RUST_CLI_LINK_TARGETS_32 := fullbench32 fuzzer32 zstreamtest32
|
||||
$(RUST_CLI_LINK_TARGETS_32): $(RUST_CLI_HELPERS_STATICLIB_32)
|
||||
|
||||
.PHONY: versionsTest
|
||||
versionsTest: clean
|
||||
$(PYTHON) test-zstd-versions.py
|
||||
|
||||
Reference in New Issue
Block a user