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

* Move `counting` into the workspace
* Inrease `HUF_DECOMPRESS_WORKSPACE_SIZE` by 512 bytes

diff --git a/lib/common/fse.h b/lib/common/fse.h
index 52924469c9a94729abcd8415173001b509e9d3f3..1f922c35ac696679c6e2d8f2928e96c742d11bed 100644 (file)
--- a/lib/common/fse.h
+++ b/lib/common/fse.h
@@ -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);
 /**< 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))
 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)` */

diff --git a/lib/common/fse_decompress.c b/lib/common/fse_decompress.c
index c575db0de313056e72daa6338326fe25502b757c..cc45babd9f178d26d8f8218bfaca02a050eee739 100644 (file)
--- a/lib/common/fse_decompress.c
+++ b/lib/common/fse_decompress.c
@@ -318,22 +318,29 @@ FORCE_INLINE_TEMPLATE size_t FSE_decompress_wksp_body(
 {
     const BYTE* const istart = (const BYTE*)cSrc;
     const BYTE* ip = istart;
-    short counting[FSE_MAX_SYMBOL_VALUE+1];
     unsigned tableLog;
     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 */
-    size_t const NCountLength = FSE_readNCount_bmi2(counting, &maxSymbolValue, &tableLog, istart, cSrcSize, bmi2);
-    if (FSE_isError(NCountLength)) return NCountLength;
-    if (tableLog > maxLog) return ERROR(tableLog_tooLarge);
-    assert(NCountLength <= cSrcSize);
-    ip += NCountLength;
-    cSrcSize -= NCountLength;
+    {
+        size_t const NCountLength = FSE_readNCount_bmi2(counting, &maxSymbolValue, &tableLog, istart, cSrcSize, bmi2);
+        if (FSE_isError(NCountLength)) return NCountLength;
+        if (tableLog > maxLog) return ERROR(tableLog_tooLarge);
+        assert(NCountLength <= cSrcSize);
+        ip += NCountLength;
+        cSrcSize -= NCountLength;
+    }
 
     if (FSE_DECOMPRESS_WKSP_SIZE(tableLog, maxSymbolValue) > wkspSize) return ERROR(tableLog_tooLarge);
     workSpace = dtable + FSE_DTABLE_SIZE_U32(tableLog);
     wkspSize -= FSE_DTABLE_SIZE(tableLog);
+    wkspSize -= countingSize;
 
     CHECK_F( FSE_buildDTable_internal(dtable, counting, maxSymbolValue, tableLog, workSpace, wkspSize) );
 

diff --git a/lib/common/huf.h b/lib/common/huf.h
index a470157493752b6261e04c98becade506153bdcf..139793bfc500ed2f63b223fcb6336d41bfc184a2 100644 (file)
--- a/lib/common/huf.h
+++ b/lib/common/huf.h
@@ -279,7 +279,7 @@ U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
  *  a required workspace size greater than that specified in the following
  *  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))
 
 #ifndef HUF_FORCE_DECOMPRESS_X2