sharpetronics/zstd
1
0
Fork 0
mirror of https://github.com/facebook/zstd.git synced 2025-10-20 00:04:34 -04:00
Code Issues Projects Releases Wiki Activity

[fse] Reduce stack usage of FSE_decompress_wksp() by 512 bytes

Browse Source 
* Move `counting` into the workspace
* Inrease `HUF_DECOMPRESS_WORKSPACE_SIZE` by 512 bytes
This commit is contained in:
Nick Terrell 2021-03-05 13:14:06 -08:00
parent 0f18059a4e
commit 3b1aba42cc
3 changed files with 17 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/common/fse.h
View File

@ -352,7 +352,7 @@ size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits);
size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue); size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue);
/**< build a fake FSE_DTable, designed to always generate the same symbolValue */ /**< build a fake FSE_DTable, designed to always generate the same symbolValue */
#define FSE_DECOMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) (FSE_DTABLE_SIZE_U32(maxTableLog) + FSE_BUILD_DTABLE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue)) #define FSE_DECOMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) (FSE_DTABLE_SIZE_U32(maxTableLog) + FSE_BUILD_DTABLE_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) + (FSE_MAX_SYMBOL_VALUE + 1) / 2)
#define FSE_DECOMPRESS_WKSP_SIZE(maxTableLog, maxSymbolValue) (FSE_DECOMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) * sizeof(unsigned)) #define FSE_DECOMPRESS_WKSP_SIZE(maxTableLog, maxSymbolValue) (FSE_DECOMPRESS_WKSP_SIZE_U32(maxTableLog, maxSymbolValue) * sizeof(unsigned))
size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, unsigned maxLog, void* workSpace, size_t wkspSize); size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, unsigned maxLog, void* workSpace, size_t wkspSize);
/**< same as FSE_decompress(), using an externally allocated `workSpace` produced with `FSE_DECOMPRESS_WKSP_SIZE_U32(maxLog, maxSymbolValue)` */ /**< same as FSE_decompress(), using an externally allocated `workSpace` produced with `FSE_DECOMPRESS_WKSP_SIZE_U32(maxLog, maxSymbolValue)` */

11
lib/common/fse_decompress.c
View File

@ -318,22 +318,29 @@ FORCE_INLINE_TEMPLATE size_t FSE_decompress_wksp_body(
{ {
const BYTE* const istart = (const BYTE*)cSrc; const BYTE* const istart = (const BYTE*)cSrc;
const BYTE* ip = istart; const BYTE* ip = istart;
short counting[FSE_MAX_SYMBOL_VALUE+1];
unsigned tableLog; unsigned tableLog;
unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE; unsigned maxSymbolValue = FSE_MAX_SYMBOL_VALUE;
FSE_DTable* const dtable = (FSE_DTable*)workSpace; short* counting = (short*) workSpace;
size_t const countingSize = sizeof(short) * (FSE_MAX_SYMBOL_VALUE + 1);
FSE_DTable* const dtable = (FSE_DTable*)(void*)((BYTE*)workSpace + countingSize);
DEBUG_STATIC_ASSERT((FSE_MAX_SYMBOL_VALUE + 1) % 2 == 0);
if (wkspSize < countingSize) return ERROR(GENERIC);
/* normal FSE decoding mode */ /* normal FSE decoding mode */
{
size_t const NCountLength = FSE_readNCount_bmi2(counting, &maxSymbolValue, &tableLog, istart, cSrcSize, bmi2); size_t const NCountLength = FSE_readNCount_bmi2(counting, &maxSymbolValue, &tableLog, istart, cSrcSize, bmi2);
if (FSE_isError(NCountLength)) return NCountLength; if (FSE_isError(NCountLength)) return NCountLength;
if (tableLog > maxLog) return ERROR(tableLog_tooLarge); if (tableLog > maxLog) return ERROR(tableLog_tooLarge);
assert(NCountLength <= cSrcSize); assert(NCountLength <= cSrcSize);
ip += NCountLength; ip += NCountLength;
cSrcSize -= NCountLength; cSrcSize -= NCountLength;
}
if (FSE_DECOMPRESS_WKSP_SIZE(tableLog, maxSymbolValue) > wkspSize) return ERROR(tableLog_tooLarge); if (FSE_DECOMPRESS_WKSP_SIZE(tableLog, maxSymbolValue) > wkspSize) return ERROR(tableLog_tooLarge);
workSpace = dtable + FSE_DTABLE_SIZE_U32(tableLog); workSpace = dtable + FSE_DTABLE_SIZE_U32(tableLog);
wkspSize -= FSE_DTABLE_SIZE(tableLog); wkspSize -= FSE_DTABLE_SIZE(tableLog);
wkspSize -= countingSize;
CHECK_F( FSE_buildDTable_internal(dtable, counting, maxSymbolValue, tableLog, workSpace, wkspSize) ); CHECK_F( FSE_buildDTable_internal(dtable, counting, maxSymbolValue, tableLog, workSpace, wkspSize) );

2
lib/common/huf.h
View File

@ -279,7 +279,7 @@ U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
* a required workspace size greater than that specified in the following * a required workspace size greater than that specified in the following
* macro. * macro.
*/ */
#define HUF_DECOMPRESS_WORKSPACE_SIZE (2 << 10) #define HUF_DECOMPRESS_WORKSPACE_SIZE ((2 << 10) + (1 << 9))
#define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32)) #define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32))
#ifndef HUF_FORCE_DECOMPRESS_X2 #ifndef HUF_FORCE_DECOMPRESS_X2