Increase maximum window size

* Maximum window size in 32-bit mode is 1GB, since allocations for 2GB fail
  on my Mac.
* Maximum window size in 64-bit mode is 2GB, since that is the largest
  power of 2 that works with the overflow prevention.
* Allow `--long=windowLog` to set the window log, along with
  `--zstd=wlog=#`. These options also set the window size during
  decompression, but don't override `--memory=#` if it is set.
* Present a helpful error message when the window size is too large during
  decompression.
* The long range matcher defaults to a hash log 7 less than the window log,
  which keeps it at 20 for window log 27.
* Keep the default long range matcher window size and the default maximum
  window size at 27 for the API and CLI.
* Add tests that use the maximum window size and hash size for compression
  and decompression.
This commit is contained in:
Nick Terrell
2017-09-26 14:00:01 -07:00
parent 8a4d0abc1c
commit c233bdbaee
12 changed files with 139 additions and 40 deletions
+1
View File
@@ -102,6 +102,7 @@ static const U32 repStartValue[ZSTD_REP_NUM] = { 1, 4, 8 };
#define BIT0 1
#define ZSTD_WINDOWLOG_ABSOLUTEMIN 10
#define ZSTD_WINDOWLOG_DEFAULTMAX 27 /* Default maximum allowed window log */
static const size_t ZSTD_fcs_fieldSize[4] = { 0, 2, 4, 8 };
static const size_t ZSTD_did_fieldSize[4] = { 0, 1, 2, 4 };
+7 -7
View File
@@ -429,7 +429,7 @@ size_t ZSTD_CCtxParam_setParameter(
case ZSTD_p_enableLongDistanceMatching :
if (value != 0) {
ZSTD_cLevelToCCtxParams(params);
params->cParams.windowLog = ZSTD_LDM_WINDOW_LOG;
params->cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG;
}
return ZSTD_ldm_initializeParameters(&params->ldmParams, value);
@@ -1599,7 +1599,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
const BYTE* ip = (const BYTE*)src;
BYTE* const ostart = (BYTE*)dst;
BYTE* op = ostart;
U32 const maxDist = 1 << cctx->appliedParams.cParams.windowLog;
U32 const maxDist = (U32)1 << cctx->appliedParams.cParams.windowLog;
if (cctx->appliedParams.fParams.checksumFlag && srcSize)
XXH64_update(&cctx->xxhState, src, srcSize);
@@ -1630,9 +1630,9 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
* windowLog <= 31 ==> 3<<29 + 1<<windowLog < 7<<29 < 1<<32.
*/
if (cctx->lowLimit > (3U<<29)) {
U32 const cycleMask = (1 << ZSTD_cycleLog(cctx->appliedParams.cParams.chainLog, cctx->appliedParams.cParams.strategy)) - 1;
U32 const cycleMask = ((U32)1 << ZSTD_cycleLog(cctx->appliedParams.cParams.chainLog, cctx->appliedParams.cParams.strategy)) - 1;
U32 const current = (U32)(ip - cctx->base);
U32 const newCurrent = (current & cycleMask) + (1 << cctx->appliedParams.cParams.windowLog);
U32 const newCurrent = (current & cycleMask) + ((U32)1 << cctx->appliedParams.cParams.windowLog);
U32 const correction = current - newCurrent;
ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30);
ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30);
@@ -1688,7 +1688,7 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
U32 const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536); /* 0-3 */
U32 const dictIDSizeCode = params.fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength; /* 0-3 */
U32 const checksumFlag = params.fParams.checksumFlag>0;
U32 const windowSize = 1U << params.cParams.windowLog;
U32 const windowSize = (U32)1 << params.cParams.windowLog;
U32 const singleSegment = params.fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);
BYTE const windowLogByte = (BYTE)((params.cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
U32 const fcsCode = params.fParams.contentSizeFlag ?
@@ -1788,7 +1788,7 @@ size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx)
{
ZSTD_compressionParameters const cParams =
ZSTD_getCParamsFromCCtxParams(cctx->appliedParams, 0, 0);
return MIN (ZSTD_BLOCKSIZE_MAX, 1 << cParams.windowLog);
return MIN (ZSTD_BLOCKSIZE_MAX, (U32)1 << cParams.windowLog);
}
size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
@@ -1837,7 +1837,7 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_CCtx* zc, const void* src, size_t
case ZSTD_btopt:
case ZSTD_btultra:
if (srcSize >= HASH_READ_SIZE)
ZSTD_updateTree(zc, iend-HASH_READ_SIZE, iend, 1 << zc->appliedParams.cParams.searchLog, zc->appliedParams.cParams.searchLength);
ZSTD_updateTree(zc, iend-HASH_READ_SIZE, iend, (U32)1 << zc->appliedParams.cParams.searchLog, zc->appliedParams.cParams.searchLength);
break;
default:
+6 -2
View File
@@ -14,14 +14,14 @@
#define LDM_BUCKET_SIZE_LOG 3
#define LDM_MIN_MATCH_LENGTH 64
#define LDM_HASH_LOG 20
#define LDM_HASH_RLOG 7
#define LDM_HASH_CHAR_OFFSET 10
size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm)
{
ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
params->enableLdm = enableLdm>0;
params->hashLog = LDM_HASH_LOG;
params->hashLog = 0;
params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
params->minMatchLength = LDM_MIN_MATCH_LENGTH;
params->hashEveryLog = ZSTD_LDM_HASHEVERYLOG_NOTSET;
@@ -30,6 +30,10 @@ size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm)
void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog)
{
if (params->hashLog == 0) {
params->hashLog = MAX(ZSTD_HASHLOG_MIN, windowLog - LDM_HASH_RLOG);
assert(params->hashLog <= ZSTD_HASHLOG_MAX);
}
if (params->hashEveryLog == ZSTD_LDM_HASHEVERYLOG_NOTSET) {
params->hashEveryLog =
windowLog < params->hashLog ? 0 : windowLog - params->hashLog;
+1 -1
View File
@@ -20,7 +20,7 @@ extern "C" {
* Long distance matching
***************************************/
#define ZSTD_LDM_WINDOW_LOG 27
#define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_DEFAULTMAX
#define ZSTD_LDM_HASHEVERYLOG_NOTSET 9999
/** ZSTD_compressBlock_ldm_generic() :
+3 -3
View File
@@ -35,7 +35,7 @@
* Frames requiring more memory will be rejected.
*/
#ifndef ZSTD_MAXWINDOWSIZE_DEFAULT
# define ZSTD_MAXWINDOWSIZE_DEFAULT (((U64)1 << ZSTD_WINDOWLOG_MAX) + 1) /* defined within zstd.h */
# define ZSTD_MAXWINDOWSIZE_DEFAULT (((U32)1 << ZSTD_WINDOWLOG_DEFAULTMAX) + 1)
#endif
@@ -913,7 +913,7 @@ static seq_t ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e l
offset = 0;
else {
ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 2);
ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
assert(ofBits <= MaxOff);
if (MEM_32bits() && longOffsets) {
U32 const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN_32-1);
@@ -1159,7 +1159,7 @@ seq_t ZSTD_decodeSequenceLong(seqState_t* seqState, ZSTD_longOffset_e const long
offset = 0;
else {
ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 2);
ZSTD_STATIC_ASSERT(LONG_OFFSETS_MAX_EXTRA_BITS_32 == 5);
assert(ofBits <= MaxOff);
if (MEM_32bits() && longOffsets) {
U32 const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN_32-1);
+6 -4
View File
@@ -376,8 +376,8 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void); /*!< recommended size for output
#define ZSTD_MAGIC_SKIPPABLE_START 0x184D2A50U
#define ZSTD_MAGIC_DICTIONARY 0xEC30A437 /* v0.7+ */
#define ZSTD_WINDOWLOG_MAX_32 27
#define ZSTD_WINDOWLOG_MAX_64 27
#define ZSTD_WINDOWLOG_MAX_32 30
#define ZSTD_WINDOWLOG_MAX_64 31
#define ZSTD_WINDOWLOG_MAX ((unsigned)(sizeof(size_t) == 4 ? ZSTD_WINDOWLOG_MAX_32 : ZSTD_WINDOWLOG_MAX_64))
#define ZSTD_WINDOWLOG_MIN 10
#define ZSTD_HASHLOG_MAX MIN(ZSTD_WINDOWLOG_MAX, 30)
@@ -941,7 +941,9 @@ typedef enum {
* Special: value 0 means "do not change cLevel". */
ZSTD_p_windowLog, /* Maximum allowed back-reference distance, expressed as power of 2.
* Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
* Special: value 0 means "do not change windowLog". */
* Special: value 0 means "do not change windowLog".
* Note: Using a window size greater than ZSTD_MAXWINDOWSIZE_DEFAULT (default: 2^27)
* requires setting the maximum window size at least as large during decompression. */
ZSTD_p_hashLog, /* Size of the probe table, as a power of 2.
* Resulting table size is (1 << (hashLog+2)).
* Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
@@ -1009,7 +1011,7 @@ typedef enum {
* Larger values increase memory usage and compression ratio, but decrease
* compression speed.
* Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
* (default: 20). */
* (default: windowlog - 7). */
ZSTD_p_ldmMinMatch, /* Minimum size of searched matches for long distance matcher.
* Larger/too small values usually decrease compression ratio.
* Must be clamped between ZSTD_LDM_MINMATCH_MIN