AuroraMiddleware/zstd
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Merge pull request #2133 from felixhandte/single-size-calculation

Browse Source 
Consolidate CCtx Size Estimation Code
This commit is contained in:
Felix Handte 2020-05-28 13:07:18 -04:00 committed by GitHub
commit 2af4e07326
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 76 additions and 67 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

138
lib/compress/zstd_compress.c
View File

@ -1123,45 +1123,61 @@ ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
return tableSpace + optSpace;
}
static size_t ZSTD_estimateCCtxSize_usingCCtxParams_internal(
const ZSTD_compressionParameters* cParams,
const ldmParams_t* ldmParams,
const int isStatic,
const size_t buffInSize,
const size_t buffOutSize,
const U64 pledgedSrcSize)
{
size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << cParams->windowLog), pledgedSrcSize));
size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
U32 const divider = (cParams->minMatch==3) ? 3 : 4;
size_t const maxNbSeq = blockSize / divider;
size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+ ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))
+ 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
size_t const matchStateSize = ZSTD_sizeof_matchState(cParams, /* forCCtx */ 1);
size_t const ldmSpace = ZSTD_ldm_getTableSize(*ldmParams);
size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(*ldmParams, blockSize);
size_t const ldmSeqSpace = ldmParams->enableLdm ?
ZSTD_cwksp_alloc_size(maxNbLdmSeq * sizeof(rawSeq)) : 0;
size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize)
+ ZSTD_cwksp_alloc_size(buffOutSize);
size_t const cctxSpace = isStatic ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
size_t const neededSpace =
cctxSpace +
entropySpace +
blockStateSpace +
ldmSpace +
ldmSeqSpace +
matchStateSize +
tokenSpace +
bufferSpace;
DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);
return neededSpace;
}
size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
{
RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
{ ZSTD_compressionParameters const cParams =
ZSTD_compressionParameters const cParams =
ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0);
size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
U32 const divider = (cParams.minMatch==3) ? 3 : 4;
size_t const maxNbSeq = blockSize / divider;
size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+ ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))
+ 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 1);
size_t const ldmSpace = ZSTD_ldm_getTableSize(params->ldmParams);
size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq));
/* estimateCCtxSize is for one-shot compression. So no buffers should
* be needed. However, we still allocate two 0-sized buffers, which can
* take space under ASAN. */
size_t const bufferSpace = ZSTD_cwksp_alloc_size(0)
+ ZSTD_cwksp_alloc_size(0);
size_t const cctxSpace = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx));
size_t const neededSpace =
cctxSpace +
entropySpace +
blockStateSpace +
ldmSpace +
ldmSeqSpace +
matchStateSize +
tokenSpace +
bufferSpace;
DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);
return neededSpace;
}
RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
/* estimateCCtxSize is for one-shot compression. So no buffers should
* be needed. However, we still allocate two 0-sized buffers, which can
* take space under ASAN. */
return ZSTD_estimateCCtxSize_usingCCtxParams_internal(
&cParams, &params->ldmParams, 1, 0, 0, ZSTD_CONTENTSIZE_UNKNOWN);
}
size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams)
@ -1192,14 +1208,13 @@ size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params)
RETURN_ERROR_IF(params->nbWorkers > 0, GENERIC, "Estimate CCtx size is supported for single-threaded compression only.");
{ ZSTD_compressionParameters const cParams =
ZSTD_getCParamsFromCCtxParams(params, ZSTD_CONTENTSIZE_UNKNOWN, 0);
size_t const CCtxSize = ZSTD_estimateCCtxSize_usingCCtxParams(params);
size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
size_t const inBuffSize = ((size_t)1 << cParams.windowLog) + blockSize;
size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1;
size_t const streamingSize = ZSTD_cwksp_alloc_size(inBuffSize)
+ ZSTD_cwksp_alloc_size(outBuffSize);
return CCtxSize + streamingSize;
return ZSTD_estimateCCtxSize_usingCCtxParams_internal(
&cParams, &params->ldmParams, 1, inBuffSize, outBuffSize,
ZSTD_CONTENTSIZE_UNKNOWN);
}
}
@ -1441,45 +1456,28 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
U32 const divider = (params.cParams.minMatch==3) ? 3 : 4;
size_t const maxNbSeq = blockSize / divider;
size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+ ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))
+ 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0;
size_t const buffInSize = (zbuff==ZSTDb_buffered) ? windowSize + blockSize : 0;
size_t const matchStateSize = ZSTD_sizeof_matchState(&params.cParams, /* forCCtx */ 1);
size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params.ldmParams, blockSize);
ZSTD_indexResetPolicy_e needsIndexReset = zc->initialized ? ZSTDirp_continue : ZSTDirp_reset;
int const indexTooClose = ZSTD_indexTooCloseToMax(zc->blockState.matchState.window);
ZSTD_indexResetPolicy_e needsIndexReset =
(!indexTooClose && zc->initialized) ? ZSTDirp_continue : ZSTDirp_reset;
if (ZSTD_indexTooCloseToMax(zc->blockState.matchState.window)) {
needsIndexReset = ZSTDirp_reset;
}
size_t const neededSpace =
ZSTD_estimateCCtxSize_usingCCtxParams_internal(
&params.cParams, &params.ldmParams, zc->staticSize != 0,
buffInSize, buffOutSize, pledgedSrcSize);
FORWARD_IF_ERROR(neededSpace, "cctx size estimate failed!");
if (!zc->staticSize) ZSTD_cwksp_bump_oversized_duration(ws, 0);
/* Check if workspace is large enough, alloc a new one if needed */
{ size_t const cctxSpace = zc->staticSize ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) + ZSTD_cwksp_alloc_size(buffOutSize);
size_t const ldmSpace = ZSTD_ldm_getTableSize(params.ldmParams);
size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(maxNbLdmSeq * sizeof(rawSeq));
size_t const neededSpace =
cctxSpace +
entropySpace +
blockStateSpace +
ldmSpace +
ldmSeqSpace +
matchStateSize +
tokenSpace +
bufferSpace;
{
int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace;
int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace);
DEBUGLOG(4, "Need %zuKB workspace, including %zuKB for match state, and %zuKB for buffers",
neededSpace>>10, matchStateSize>>10, bufferSpace>>10);
DEBUGLOG(4, "Need %zu B workspace", neededSpace);
DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize);
if (workspaceTooSmall || workspaceWasteful) {
@ -1579,6 +1577,12 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
zc->ldmState.loadedDictEnd = 0;
}
/* Due to alignment, when reusing a workspace, we can actually consume
* up to 3 extra bytes for alignment. See the comments in zstd_cwksp.h
*/
assert(ZSTD_cwksp_used(ws) >= neededSpace &&
ZSTD_cwksp_used(ws) <= neededSpace + 3);
DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
zc->initialized = 1;
@ -1638,7 +1642,7 @@ ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx,
assert(windowLog != 0);
/* Resize working context table params for input only, since the dict
* has its own tables. */
/* pledgeSrcSize == 0 means 0! */
/* pledgedSrcSize == 0 means 0! */
params.cParams = ZSTD_adjustCParams_internal(*cdict_cParams, pledgedSrcSize, 0);
params.cParams.windowLog = windowLog;
FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,

5
lib/compress/zstd_cwksp.h
View File

@ -483,6 +483,11 @@ MEM_STATIC size_t ZSTD_cwksp_sizeof(const ZSTD_cwksp* ws) {
return (size_t)((BYTE*)ws->workspaceEnd - (BYTE*)ws->workspace);
}
MEM_STATIC size_t ZSTD_cwksp_used(const ZSTD_cwksp* ws) {
return (size_t)((BYTE*)ws->tableEnd - (BYTE*)ws->workspace)
+ (size_t)((BYTE*)ws->workspaceEnd - (BYTE*)ws->allocStart);
}
MEM_STATIC int ZSTD_cwksp_reserve_failed(const ZSTD_cwksp* ws) {
return ws->allocFailed;
}