ddidderr/zstd-rs
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Add init once memory (#3528) (#3529)

Browse Source 
- Adds memory type that is guaranteed to have been initialized at least once in the workspace's lifetime.
- Changes tag space in row hash to be based on init once memory.
This commit is contained in:
Yonatan Komornik
2023-03-13 13:20:49 -07:00
committed by GitHub
parent e2965edd10
commit 9420bce8a4
4 changed files with 180 additions and 111 deletions
Download Patch File Download Diff File Expand all files Collapse all files
lib/compress/zstd_compress.c
+52 -52
View File
@@ -1955,6 +1955,17 @@ ZSTD_reset_matchState(ZSTD_matchState_t* ms,
ZSTD_cwksp_clean_tables(ws);
}
if (ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)) {
/* Row match finder needs an additional table of hashes ("tags") */
size_t const tagTableSize = hSize;
ms->tagTable = (BYTE*) ZSTD_cwksp_reserve_aligned_init_once(ws, tagTableSize);
{ /* Switch to 32-entry rows if searchLog is 5 (or more) */
U32 const rowLog = BOUNDED(4, cParams->searchLog, 6);
assert(cParams->hashLog >= rowLog);
ms->rowHashLog = cParams->hashLog - rowLog;
}
}
/* opt parser space */
if ((forWho == ZSTD_resetTarget_CCtx) && (cParams->strategy >= ZSTD_btopt)) {
DEBUGLOG(4, "reserving optimal parser space");
@@ -1966,19 +1977,6 @@ ZSTD_reset_matchState(ZSTD_matchState_t* ms,
ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
}
if (ZSTD_rowMatchFinderUsed(cParams->strategy, useRowMatchFinder)) {
{ /* Row match finder needs an additional table of hashes ("tags") */
size_t const tagTableSize = hSize;
ms->tagTable = (BYTE*)ZSTD_cwksp_reserve_aligned(ws, tagTableSize);
if (ms->tagTable) ZSTD_memset(ms->tagTable, 0, tagTableSize);
}
{ /* Switch to 32-entry rows if searchLog is 5 (or more) */
U32 const rowLog = BOUNDED(4, cParams->searchLog, 6);
assert(cParams->hashLog >= rowLog);
ms->rowHashLog = cParams->hashLog - rowLog;
}
}
ms->cParams = *cParams;
RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
@@ -2125,45 +2123,16 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock);
/* ZSTD_wildcopy() is used to copy into the literals buffer,
* so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.
*/
zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH);
zc->seqStore.maxNbLit = blockSize;
/* buffers */
zc->bufferedPolicy = zbuff;
zc->inBuffSize = buffInSize;
zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize);
zc->outBuffSize = buffOutSize;
zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize);
/* ldm bucketOffsets table */
if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
/* TODO: avoid memset? */
size_t const numBuckets =
((size_t)1) << (params->ldmParams.hashLog -
params->ldmParams.bucketSizeLog);
zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, numBuckets);
ZSTD_memset(zc->ldmState.bucketOffsets, 0, numBuckets);
}
/* sequences storage */
ZSTD_referenceExternalSequences(zc, NULL, 0);
zc->seqStore.maxNbSeq = maxNbSeq;
zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef));
FORWARD_IF_ERROR(ZSTD_reset_matchState(
&zc->blockState.matchState,
ws,
&params->cParams,
params->useRowMatchFinder,
crp,
needsIndexReset,
ZSTD_resetTarget_CCtx), "");
&zc->blockState.matchState,
ws,
&params->cParams,
params->useRowMatchFinder,
crp,
needsIndexReset,
ZSTD_resetTarget_CCtx), "");
zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef));
/* ldm hash table */
if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
@@ -2186,8 +2155,39 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
(ZSTD_Sequence*)ZSTD_cwksp_reserve_aligned(ws, maxNbExternalSeq * sizeof(ZSTD_Sequence));
}
/* buffers */
/* ZSTD_wildcopy() is used to copy into the literals buffer,
* so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.
*/
zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH);
zc->seqStore.maxNbLit = blockSize;
zc->bufferedPolicy = zbuff;
zc->inBuffSize = buffInSize;
zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize);
zc->outBuffSize = buffOutSize;
zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize);
/* ldm bucketOffsets table */
if (params->ldmParams.enableLdm == ZSTD_ps_enable) {
/* TODO: avoid memset? */
size_t const numBuckets =
((size_t)1) << (params->ldmParams.hashLog -
params->ldmParams.bucketSizeLog);
zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, numBuckets);
ZSTD_memset(zc->ldmState.bucketOffsets, 0, numBuckets);
}
/* sequences storage */
ZSTD_referenceExternalSequences(zc, NULL, 0);
zc->seqStore.maxNbSeq = maxNbSeq;
zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
assert(ZSTD_cwksp_estimated_space_within_bounds(ws, neededSpace, resizeWorkspace));
assert(ZSTD_cwksp_estimated_space_within_bounds(ws, neededSpace));
zc->initialized = 1;