diff --git a/lib/fse.c b/lib/fse.c
index 38e40a57..e74c1e83 100644
--- a/lib/fse.c
+++ b/lib/fse.c
@@ -34,10 +34,10 @@
 
 #ifndef FSE_COMMONDEFS_ONLY
 
-/****************************************************************
+/* **************************************************************
 *  Tuning parameters
 ****************************************************************/
-/* MEMORY_USAGE :
+/*!MEMORY_USAGE :
 *  Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
 *  Increasing memory usage improves compression ratio
 *  Reduced memory usage can improve speed, due to cache effect
@@ -45,26 +45,23 @@
 #define FSE_MAX_MEMORY_USAGE 14
 #define FSE_DEFAULT_MEMORY_USAGE 13
 
-/* FSE_MAX_SYMBOL_VALUE :
+/*!FSE_MAX_SYMBOL_VALUE :
 *  Maximum symbol value authorized.
 *  Required for proper stack allocation */
 #define FSE_MAX_SYMBOL_VALUE 255
 
 
-/****************************************************************
+/* **************************************************************
 *  template functions type & suffix
 ****************************************************************/
 #define FSE_FUNCTION_TYPE BYTE
 #define FSE_FUNCTION_EXTENSION
+#define FSE_DECODE_TYPE FSE_decode_t
 
 
-/****************************************************************
-*  Byte symbol type
-****************************************************************/
 #endif   /* !FSE_COMMONDEFS_ONLY */
 
-
-/****************************************************************
+/* **************************************************************
 *  Compiler specifics
 ****************************************************************/
 #ifdef _MSC_VER    /* Visual Studio */
@@ -82,7 +79,7 @@
 #endif
 
 
-/****************************************************************
+/* **************************************************************
 *  Includes
 ****************************************************************/
 #include <stdlib.h>     /* malloc, free, qsort */
@@ -92,7 +89,7 @@
 #include "fse_static.h"
 
 
-/****************************************************************
+/* ***************************************************************
 *  Constants
 *****************************************************************/
 #define FSE_MAX_TABLELOG  (FSE_MAX_MEMORY_USAGE-2)
@@ -107,20 +104,20 @@
 #endif
 
 
-/****************************************************************
+/* **************************************************************
 *  Error Management
 ****************************************************************/
 #define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
 
 
-/****************************************************************
+/* **************************************************************
 *  Complex types
 ****************************************************************/
 typedef U32 CTable_max_t[FSE_CTABLE_SIZE_U32(FSE_MAX_TABLELOG, FSE_MAX_SYMBOL_VALUE)];
 typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
 
 
-/****************************************************************
+/* **************************************************************
 *  Templates
 ****************************************************************/
 /*
@@ -144,8 +141,7 @@ typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
 
 
 /* Function templates */
-size_t FSE_FUNCTION_NAME(FSE_count_generic, FSE_FUNCTION_EXTENSION)
-(unsigned* count, unsigned* maxSymbolValuePtr, const FSE_FUNCTION_TYPE* source, size_t sourceSize, unsigned safe)
+size_t FSE_count_generic(unsigned* count, unsigned* maxSymbolValuePtr, const FSE_FUNCTION_TYPE* source, size_t sourceSize, unsigned safe)
 {
     const FSE_FUNCTION_TYPE* ip = source;
     const FSE_FUNCTION_TYPE* const iend = ip+sourceSize;
@@ -226,7 +222,7 @@ size_t FSE_FUNCTION_NAME(FSE_count_generic, FSE_FUNCTION_EXTENSION)
 size_t FSE_FUNCTION_NAME(FSE_countFast, FSE_FUNCTION_EXTENSION)
 (unsigned* count, unsigned* maxSymbolValuePtr, const FSE_FUNCTION_TYPE* source, size_t sourceSize)
 {
-    return FSE_FUNCTION_NAME(FSE_count_generic, FSE_FUNCTION_EXTENSION) (count, maxSymbolValuePtr, source, sourceSize, 0);
+    return FSE_count_generic(count, maxSymbolValuePtr, source, sourceSize, 0);
 }
 
 size_t FSE_FUNCTION_NAME(FSE_count, FSE_FUNCTION_EXTENSION)
@@ -235,25 +231,26 @@ size_t FSE_FUNCTION_NAME(FSE_count, FSE_FUNCTION_EXTENSION)
     if ((sizeof(FSE_FUNCTION_TYPE)==1) && (*maxSymbolValuePtr >= 255))
     {
         *maxSymbolValuePtr = 255;
-        return FSE_FUNCTION_NAME(FSE_count_generic, FSE_FUNCTION_EXTENSION) (count, maxSymbolValuePtr, source, sourceSize, 0);
+        return FSE_count_generic(count, maxSymbolValuePtr, source, sourceSize, 0);
     }
-    return FSE_FUNCTION_NAME(FSE_count_generic, FSE_FUNCTION_EXTENSION) (count, maxSymbolValuePtr, source, sourceSize, 1);
+    return FSE_count_generic(count, maxSymbolValuePtr, source, sourceSize, 1);
 }
 
 
 static U32 FSE_tableStep(U32 tableSize) { return (tableSize>>1) + (tableSize>>3) + 3; }
 
-size_t FSE_FUNCTION_NAME(FSE_buildCTable, FSE_FUNCTION_EXTENSION)
-(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
 {
     const unsigned tableSize = 1 << tableLog;
     const unsigned tableMask = tableSize - 1;
-    U16* tableU16 = ( (U16*) ct) + 2;
-    FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) (((U32*)ct) + 1 + (tableLog ? tableSize>>1 : 1) );
+    void* const ptr = ct;
+    U16* const tableU16 = ( (U16*) ptr) + 2;
+    void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableLog ? tableSize>>1 : 1) ;
+    FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
     const unsigned step = FSE_tableStep(tableSize);
     unsigned cumul[FSE_MAX_SYMBOL_VALUE+2];
     U32 position = 0;
-    FSE_FUNCTION_TYPE tableSymbol[FSE_MAX_TABLESIZE]; /* init isn't necessary, even if static analyzer complain about it */
+    FSE_FUNCTION_TYPE tableSymbol[FSE_MAX_TABLESIZE]; /* memset() is not necessary, even if static analyzer complain about it */
     U32 highThreshold = tableSize-1;
     unsigned symbol;
     unsigned i;
@@ -269,7 +266,7 @@ size_t FSE_FUNCTION_NAME(FSE_buildCTable, FSE_FUNCTION_EXTENSION)
     cumul[0] = 0;
     for (i=1; i<=maxSymbolValue+1; i++)
     {
-        if (normalizedCounter[i-1]==-1)   /* Low prob symbol */
+        if (normalizedCounter[i-1]==-1)   /* Low proba symbol */
         {
             cumul[i] = cumul[i-1] + 1;
             tableSymbol[highThreshold--] = (FSE_FUNCTION_TYPE)(i-1);
@@ -287,7 +284,7 @@ size_t FSE_FUNCTION_NAME(FSE_buildCTable, FSE_FUNCTION_EXTENSION)
         {
             tableSymbol[position] = (FSE_FUNCTION_TYPE)symbol;
             position = (position + step) & tableMask;
-            while (position > highThreshold) position = (position + step) & tableMask;   /* Lowprob area */
+            while (position > highThreshold) position = (position + step) & tableMask;   /* Low proba area */
         }
     }
 
@@ -296,7 +293,7 @@ size_t FSE_FUNCTION_NAME(FSE_buildCTable, FSE_FUNCTION_EXTENSION)
     /* Build table */
     for (i=0; i<tableSize; i++)
     {
-        FSE_FUNCTION_TYPE s = tableSymbol[i];   /* static analyzer doesn't understand tableSymbol is properly initialized */
+        FSE_FUNCTION_TYPE s = tableSymbol[i];   /* note : static analyzer may not understand tableSymbol is properly initialized */
         tableU16[cumul[s]++] = (U16) (tableSize+i);   /* TableU16 : sorted by symbol order; gives next state value */
     }
 
@@ -332,24 +329,22 @@ size_t FSE_FUNCTION_NAME(FSE_buildCTable, FSE_FUNCTION_EXTENSION)
 }
 
 
-#define FSE_DECODE_TYPE FSE_TYPE_NAME(FSE_decode_t, FSE_FUNCTION_EXTENSION)
-
-FSE_DTable* FSE_FUNCTION_NAME(FSE_createDTable, FSE_FUNCTION_EXTENSION) (unsigned tableLog)
+FSE_DTable* FSE_createDTable (unsigned tableLog)
 {
     if (tableLog > FSE_TABLELOG_ABSOLUTE_MAX) tableLog = FSE_TABLELOG_ABSOLUTE_MAX;
     return (FSE_DTable*)malloc( FSE_DTABLE_SIZE_U32(tableLog) * sizeof (U32) );
 }
 
-void FSE_FUNCTION_NAME(FSE_freeDTable, FSE_FUNCTION_EXTENSION) (FSE_DTable* dt)
+void FSE_freeDTable (FSE_DTable* dt)
 {
     free(dt);
 }
 
-size_t FSE_FUNCTION_NAME(FSE_buildDTable, FSE_FUNCTION_EXTENSION)
-(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
+size_t FSE_buildDTable(FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog)
 {
-    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)dt;
-    FSE_DECODE_TYPE* const tableDecode = (FSE_DECODE_TYPE*) (dt+1);   /* because dt is unsigned, 32-bits aligned on 32-bits */
+    FSE_DTableHeader DTableH;
+    void* const tdPtr = dt+1;   /* because dt is unsigned, 32-bits aligned on 32-bits */
+    FSE_DECODE_TYPE* const tableDecode = (FSE_DECODE_TYPE*) (tdPtr);
     const U32 tableSize = 1 << tableLog;
     const U32 tableMask = tableSize-1;
     const U32 step = FSE_tableStep(tableSize);
@@ -365,7 +360,7 @@ size_t FSE_FUNCTION_NAME(FSE_buildDTable, FSE_FUNCTION_EXTENSION)
     if (tableLog > FSE_MAX_TABLELOG) return ERROR(tableLog_tooLarge);
 
     /* Init, lay down lowprob symbols */
-    DTableH[0].tableLog = (U16)tableLog;
+    DTableH.tableLog = (U16)tableLog;
     for (s=0; s<=maxSymbolValue; s++)
     {
         if (normalizedCounter[s]==-1)
@@ -406,7 +401,8 @@ size_t FSE_FUNCTION_NAME(FSE_buildDTable, FSE_FUNCTION_EXTENSION)
         }
     }
 
-    DTableH->fastMode = (U16)noLarge;
+    DTableH.fastMode = (U16)noLarge;
+    memcpy(dt, &DTableH, sizeof(DTableH));
     return 0;
 }
 
@@ -890,8 +886,10 @@ size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits)
     const unsigned tableSize = 1 << nbBits;
     const unsigned tableMask = tableSize - 1;
     const unsigned maxSymbolValue = tableMask;
-    U16* tableU16 = ( (U16*) ct) + 2;
-    FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) ((((U32*)ct)+1) + (tableSize>>1));
+    void* const ptr = ct;
+    U16* const tableU16 = ( (U16*) ptr) + 2;
+    void* const FSCT = ((U32*)ptr) + 1 /* header */ + (tableSize>>1);   /* assumption : tableLog >= 1 */
+    FSE_symbolCompressionTransform* const symbolTT = (FSE_symbolCompressionTransform*) (FSCT);
     unsigned s;
 
     /* Sanity checks */
@@ -918,8 +916,10 @@ size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits)
 /* fake FSE_CTable, for rle (100% always same symbol) input */
 size_t FSE_buildCTable_rle (FSE_CTable* ct, BYTE symbolValue)
 {
-    U16* tableU16 = ( (U16*) ct) + 2;
-    FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) ((U32*)ct + 2);
+    void* ptr = ct;
+    U16* tableU16 = ( (U16*) ptr) + 2;
+    void* FSCTptr = (U32*)ptr + 2;
+    FSE_symbolCompressionTransform* symbolTT = (FSE_symbolCompressionTransform*) FSCTptr;
 
     /* header */
     tableU16[-2] = (U16) 0;
@@ -1076,8 +1076,10 @@ size_t FSE_compress (void* dst, size_t dstSize, const void* src, size_t srcSize)
 *********************************************************/
 size_t FSE_buildDTable_rle (FSE_DTable* dt, BYTE symbolValue)
 {
-    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)dt;
-    FSE_decode_t* const cell = (FSE_decode_t*)(dt + 1);   /* because dt is unsigned */
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    void* dPtr = dt + 1;
+    FSE_decode_t* const cell = (FSE_decode_t*)dPtr;
 
     DTableH->tableLog = 0;
     DTableH->fastMode = 0;
@@ -1092,8 +1094,10 @@ size_t FSE_buildDTable_rle (FSE_DTable* dt, BYTE symbolValue)
 
 size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits)
 {
-    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)dt;
-    FSE_decode_t* const dinfo = (FSE_decode_t*)(dt + 1);   /* because dt is unsigned */
+    void* ptr = dt;
+    FSE_DTableHeader* const DTableH = (FSE_DTableHeader*)ptr;
+    void* dPtr = dt + 1;
+    FSE_decode_t* const dinfo = (FSE_decode_t*)dPtr;
     const unsigned tableSize = 1 << nbBits;
     const unsigned tableMask = tableSize - 1;
     const unsigned maxSymbolValue = tableMask;
@@ -1189,7 +1193,8 @@ size_t FSE_decompress_usingDTable(void* dst, size_t originalSize,
                             const void* cSrc, size_t cSrcSize,
                             const FSE_DTable* dt)
 {
-    const FSE_DTableHeader* DTableH = (const FSE_DTableHeader*)dt;
+    const void* ptr = dt;
+    const FSE_DTableHeader* DTableH = (const FSE_DTableHeader*)ptr;
     const U32 fastMode = DTableH->fastMode;
 
     /* select fast mode (static) */
diff --git a/lib/fse.h b/lib/fse.h
index 24d1d26d..dd1190f8 100644
--- a/lib/fse.h
+++ b/lib/fse.h
@@ -40,20 +40,20 @@ extern "C" {
 #endif
 
 
-/******************************************
+/* *****************************************
 *  Includes
 ******************************************/
 #include <stddef.h>    /* size_t, ptrdiff_t */
 
 
-/******************************************
+/* *****************************************
 *  FSE simple functions
 ******************************************/
 size_t FSE_compress(void* dst, size_t maxDstSize,
               const void* src, size_t srcSize);
 size_t FSE_decompress(void* dst,  size_t maxDstSize,
                 const void* cSrc, size_t cSrcSize);
-/*
+/*!
 FSE_compress():
     Compress content of buffer 'src', of size 'srcSize', into destination buffer 'dst'.
     'dst' buffer must be already allocated. Compression runs faster is maxDstSize >= FSE_compressBound(srcSize)
@@ -74,7 +74,7 @@ FSE_decompress():
 */
 
 
-/******************************************
+/* *****************************************
 *  Tool functions
 ******************************************/
 size_t FSE_compressBound(size_t size);       /* maximum compressed size */
@@ -84,10 +84,10 @@ unsigned    FSE_isError(size_t code);        /* tells if a return value is an er
 const char* FSE_getErrorName(size_t code);   /* provides error code string (useful for debugging) */
 
 
-/******************************************
+/* *****************************************
 *  FSE advanced functions
 ******************************************/
-/*
+/*!
 FSE_compress2():
     Same as FSE_compress(), but allows the selection of 'maxSymbolValue' and 'tableLog'
     Both parameters can be defined as '0' to mean : use default value
@@ -99,10 +99,10 @@ FSE_compress2():
 size_t FSE_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
 
 
-/******************************************
+/* *****************************************
 *  FSE detailed API
 ******************************************/
-/*
+/*!
 FSE_compress() does the following:
 1. count symbol occurrence from source[] into table count[]
 2. normalize counters so that sum(count[]) == Power_of_2 (2^tableLog)
@@ -122,7 +122,7 @@ or to save and provide normalized distribution using external method.
 
 /* *** COMPRESSION *** */
 
-/*
+/*!
 FSE_count():
    Provides the precise count of each symbol within a table 'count'
    'count' is a table of unsigned int, of minimum size (maxSymbolValuePtr[0]+1).
@@ -132,14 +132,14 @@ FSE_count():
             if FSE_isError(return), it's an error code. */
 size_t FSE_count(unsigned* count, unsigned* maxSymbolValuePtr, const unsigned char* src, size_t srcSize);
 
-/*
+/*!
 FSE_optimalTableLog():
    dynamically downsize 'tableLog' when conditions are met.
    It saves CPU time, by using smaller tables, while preserving or even improving compression ratio.
    return : recommended tableLog (necessarily <= initial 'tableLog') */
 unsigned FSE_optimalTableLog(unsigned tableLog, size_t srcSize, unsigned maxSymbolValue);
 
-/*
+/*!
 FSE_normalizeCount():
    normalize counters so that sum(count[]) == Power_of_2 (2^tableLog)
    'normalizedCounter' is a table of short, of minimum size (maxSymbolValue+1).
@@ -147,13 +147,13 @@ FSE_normalizeCount():
             or an errorCode, which can be tested using FSE_isError() */
 size_t FSE_normalizeCount(short* normalizedCounter, unsigned tableLog, const unsigned* count, size_t srcSize, unsigned maxSymbolValue);
 
-/*
+/*!
 FSE_NCountWriteBound():
    Provides the maximum possible size of an FSE normalized table, given 'maxSymbolValue' and 'tableLog'
    Typically useful for allocation purpose. */
 size_t FSE_NCountWriteBound(unsigned maxSymbolValue, unsigned tableLog);
 
-/*
+/*!
 FSE_writeNCount():
    Compactly save 'normalizedCounter' into 'buffer'.
    return : size of the compressed table
@@ -161,21 +161,21 @@ FSE_writeNCount():
 size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
 
 
-/*
+/*!
 Constructor and Destructor of type FSE_CTable
     Note that its size depends on 'tableLog' and 'maxSymbolValue' */
-typedef unsigned FSE_CTable;   /* don't allocate that. It's just a way to be more restrictive than void* */
+typedef unsigned FSE_CTable;   /* don't allocate that. It's only meant to be more restrictive than void* */
 FSE_CTable* FSE_createCTable (unsigned tableLog, unsigned maxSymbolValue);
 void        FSE_freeCTable (FSE_CTable* ct);
 
-/*
+/*!
 FSE_buildCTable():
    Builds 'ct', which must be already allocated, using FSE_createCTable()
    return : 0
             or an errorCode, which can be tested using FSE_isError() */
 size_t FSE_buildCTable(FSE_CTable* ct, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
 
-/*
+/*!
 FSE_compress_usingCTable():
    Compress 'src' using 'ct' into 'dst' which must be already allocated
    return : size of compressed data (<= maxDstSize)
@@ -183,7 +183,7 @@ FSE_compress_usingCTable():
             or an errorCode, which can be tested using FSE_isError() */
 size_t FSE_compress_usingCTable (void* dst, size_t maxDstSize, const void* src, size_t srcSize, const FSE_CTable* ct);
 
-/*
+/*!
 Tutorial :
 ----------
 The first step is to count all symbols. FSE_count() does this job very fast.
@@ -229,7 +229,7 @@ If there is an error, the function will return an ErrorCode (which can be tested
 
 /* *** DECOMPRESSION *** */
 
-/*
+/*!
 FSE_readNCount():
    Read compactly saved 'normalizedCounter' from 'rBuffer'.
    return : size read from 'rBuffer'
@@ -237,21 +237,21 @@ FSE_readNCount():
             maxSymbolValuePtr[0] and tableLogPtr[0] will also be updated with their respective values */
 size_t FSE_readNCount (short* normalizedCounter, unsigned* maxSymbolValuePtr, unsigned* tableLogPtr, const void* rBuffer, size_t rBuffSize);
 
-/*
+/*!
 Constructor and Destructor of type FSE_DTable
     Note that its size depends on 'tableLog' */
 typedef unsigned FSE_DTable;   /* don't allocate that. It's just a way to be more restrictive than void* */
 FSE_DTable* FSE_createDTable(unsigned tableLog);
 void        FSE_freeDTable(FSE_DTable* dt);
 
-/*
+/*!
 FSE_buildDTable():
    Builds 'dt', which must be already allocated, using FSE_createDTable()
    return : 0,
             or an errorCode, which can be tested using FSE_isError() */
 size_t FSE_buildDTable (FSE_DTable* dt, const short* normalizedCounter, unsigned maxSymbolValue, unsigned tableLog);
 
-/*
+/*!
 FSE_decompress_usingDTable():
    Decompress compressed source 'cSrc' of size 'cSrcSize' using 'dt'
    into 'dst' which must be already allocated.
@@ -259,7 +259,7 @@ FSE_decompress_usingDTable():
             or an errorCode, which can be tested using FSE_isError() */
 size_t FSE_decompress_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const FSE_DTable* dt);
 
-/*
+/*!
 Tutorial :
 ----------
 (Note : these functions only decompress FSE-compressed blocks.
diff --git a/lib/fse_static.h b/lib/fse_static.h
index 01e25669..c5c6f20c 100644
--- a/lib/fse_static.h
+++ b/lib/fse_static.h
@@ -40,31 +40,31 @@ extern "C" {
 #endif
 
 
-/******************************************
-*  FSE API compatible with DLL
-******************************************/
+/* *****************************************
+*  Dependencies
+*******************************************/
 #include "fse.h"
 #include "bitstream.h"
 
 
-/******************************************
+/* *****************************************
 *  Static allocation
-******************************************/
+*******************************************/
 /* FSE buffer bounds */
 #define FSE_NCOUNTBOUND 512
 #define FSE_BLOCKBOUND(size) (size + (size>>7))
 #define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
 
-/* You can statically allocate FSE CTable/DTable as a table of unsigned using below macro */
+/* It is possible to statically allocate FSE CTable/DTable as a table of unsigned using below macros */
 #define FSE_CTABLE_SIZE_U32(maxTableLog, maxSymbolValue)   (1 + (1<<(maxTableLog-1)) + ((maxSymbolValue+1)*2))
 #define FSE_DTABLE_SIZE_U32(maxTableLog)                   (1 + (1<<maxTableLog))
 
 
-/******************************************
+/* *****************************************
 *  FSE advanced API
-******************************************/
+*******************************************/
 size_t FSE_countFast(unsigned* count, unsigned* maxSymbolValuePtr, const unsigned char* src, size_t srcSize);
-/* same as FSE_count(), but blindly trust that all values within src are <= maxSymbolValuePtr[0] */
+/* same as FSE_count(), but blindly trust that all values within src are <= *maxSymbolValuePtr  */
 
 size_t FSE_buildCTable_raw (FSE_CTable* ct, unsigned nbBits);
 /* build a fake FSE_CTable, designed to not compress an input, where each symbol uses nbBits */
@@ -79,10 +79,10 @@ size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue);
 /* build a fake FSE_DTable, designed to always generate the same symbolValue */
 
 
-/******************************************
+/* *****************************************
 *  FSE symbol compression API
-******************************************/
-/*
+*******************************************/
+/*!
    This API consists of small unitary functions, which highly benefit from being inlined.
    You will want to enable link-time-optimization to ensure these functions are properly inlined in your binary.
    Visual seems to do it automatically.
@@ -103,7 +103,7 @@ static void FSE_encodeSymbol(BIT_CStream_t* bitC, FSE_CState_t* CStatePtr, unsig
 
 static void FSE_flushCState(BIT_CStream_t* bitC, const FSE_CState_t* CStatePtr);
 
-/*
+/*!
 These functions are inner components of FSE_compress_usingCTable().
 They allow the creation of custom streams, mixing multiple tables and bit sources.
 
@@ -147,9 +147,9 @@ If there is an error, it returns an errorCode (which can be tested using FSE_isE
 */
 
 
-/******************************************
+/* *****************************************
 *  FSE symbol decompression API
-******************************************/
+*******************************************/
 typedef struct
 {
     size_t      state;
@@ -163,7 +163,7 @@ static unsigned char FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bi
 
 static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr);
 
-/*
+/*!
 Let's now decompose FSE_decompress_usingDTable() into its unitary components.
 You will decode FSE-encoded symbols from the bitStream,
 and also any other bitFields you put in, **in reverse order**.
@@ -213,16 +213,16 @@ Check also the states. There might be some symbols left there, if some high prob
 */
 
 
-/******************************************
+/* *****************************************
 *  FSE unsafe API
-******************************************/
+*******************************************/
 static unsigned char FSE_decodeSymbolFast(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD);
 /* faster, but works only if nbBits is always >= 1 (otherwise, result will be corrupted) */
 
 
-/******************************************
-*  Implementation of inline functions
-******************************************/
+/* *****************************************
+*  Implementation of inlined functions
+*******************************************/
 typedef struct
 {
     int deltaFindState;
@@ -231,10 +231,12 @@ typedef struct
 
 MEM_STATIC void FSE_initCState(FSE_CState_t* statePtr, const FSE_CTable* ct)
 {
-    const U32 tableLog = ( (const U16*) ct) [0];
+    const void* ptr = ct;
+    const U16* u16ptr = (const U16*) ptr;
+    const U32 tableLog = *u16ptr;
     statePtr->value = (ptrdiff_t)1<<tableLog;
-    statePtr->stateTable = ((const U16*) ct) + 2;
-    statePtr->symbolTT = (const void*)((const U32*)ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1));
+    statePtr->stateTable = u16ptr+2;
+    statePtr->symbolTT = ((const U32*)ct + 1 + (tableLog ? (1<<(tableLog-1)) : 1));
     statePtr->stateLog = tableLog;
 }
 
@@ -269,7 +271,8 @@ typedef struct
 
 MEM_STATIC void FSE_initDState(FSE_DState_t* DStatePtr, BIT_DStream_t* bitD, const FSE_DTable* dt)
 {
-    const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)dt;
+    const void* ptr = dt;
+    const FSE_DTableHeader* const DTableH = (const FSE_DTableHeader*)ptr;
     DStatePtr->state = BIT_readBits(bitD, DTableH->tableLog);
     BIT_reloadDStream(bitD);
     DStatePtr->table = dt + 1;
diff --git a/lib/huff0.c b/lib/huff0.c
index 3ddb7e5e..26a7639f 100644
--- a/lib/huff0.c
+++ b/lib/huff0.c
@@ -32,7 +32,7 @@
     - Public forum : https://groups.google.com/forum/#!forum/lz4c
 ****************************************************************** */
 
-/****************************************************************
+/* **************************************************************
 *  Compiler specifics
 ****************************************************************/
 #if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
@@ -57,7 +57,7 @@
 #endif
 
 
-/****************************************************************
+/* **************************************************************
 *  Includes
 ****************************************************************/
 #include <stdlib.h>     /* malloc, free, qsort */
@@ -68,7 +68,7 @@
 #include "fse.h"        /* header compression */
 
 
-/****************************************************************
+/* **************************************************************
 *  Constants
 ****************************************************************/
 #define HUF_ABSOLUTEMAX_TABLELOG  16   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
@@ -80,27 +80,21 @@
 #endif
 
 
-/****************************************************************
+/* **************************************************************
 *  Error Management
 ****************************************************************/
+unsigned HUF_isError(size_t code) { return ERR_isError(code); }
+const char* HUF_getErrorName(size_t code) { return ERR_getErrorName(code); }
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
 
 
-/******************************************
-*  Helper functions
-******************************************/
-unsigned HUF_isError(size_t code) { return ERR_isError(code); }
-
-const char* HUF_getErrorName(size_t code) { return ERR_getErrorName(code); }
-
-
-/*********************************************************
+/* *******************************************************
 *  Huff0 : Huffman block compression
 *********************************************************/
-typedef struct HUF_CElt_s {
+struct HUF_CElt_s {
   U16  val;
   BYTE nbBits;
-} HUF_CElt ;
+};   /* typedef'd to HUF_CElt within huff0_static.h */
 
 typedef struct nodeElt_s {
     U32 count;
@@ -320,7 +314,7 @@ size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U3
     /* sort, decreasing order */
     HUF_sort(huffNode, count, maxSymbolValue);
 
-    // init for parents
+    /* init for parents */
     nonNullRank = maxSymbolValue;
     while(huffNode[nonNullRank].count == 0) nonNullRank--;
     lowS = nonNullRank; nodeRoot = nodeNb + lowS - 1; lowN = nodeNb;
@@ -330,7 +324,7 @@ size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U3
     for (n=nodeNb; n<=nodeRoot; n++) huffNode[n].count = (U32)(1U<<30);
     huffNode0[0].count = (U32)(1U<<31);
 
-    // create parents
+    /* create parents */
     while (nodeNb <= nodeRoot)
     {
         U32 n1 = (huffNode[lowS].count < huffNode[lowN].count) ? lowS-- : lowN++;
@@ -340,7 +334,7 @@ size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U3
         nodeNb++;
     }
 
-    // distribute weights (unlimited tree height)
+    /* distribute weights (unlimited tree height) */
     huffNode[nodeRoot].nbBits = 0;
     for (n=nodeRoot-1; n>=STARTNODE; n--)
         huffNode[n].nbBits = huffNode[ huffNode[n].parent ].nbBits + 1;
@@ -368,9 +362,9 @@ size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U3
             }
         }
         for (n=0; n<=maxSymbolValue; n++)
-            tree[huffNode[n].byte].nbBits = huffNode[n].nbBits;   // push nbBits per symbol, symbol order
+            tree[huffNode[n].byte].nbBits = huffNode[n].nbBits;   /* push nbBits per symbol, symbol order */
         for (n=0; n<=maxSymbolValue; n++)
-            tree[n].val = valPerRank[tree[n].nbBits]++;   // assign value within rank, symbol order
+            tree[n].val = valPerRank[tree[n].nbBits]++;   /* assign value within rank, symbol order */
     }
 
     return maxNbBits;
@@ -636,12 +630,12 @@ size_t HUF_readDTableX2 (U16* DTable, const void* src, size_t srcSize)
     BYTE huffWeight[HUF_MAX_SYMBOL_VALUE + 1];
     U32 rankVal[HUF_ABSOLUTEMAX_TABLELOG + 1];   /* large enough for values from 0 to 16 */
     U32 tableLog = 0;
-    const BYTE* ip = (const BYTE*) src;
-    size_t iSize = ip[0];
+    size_t iSize;
     U32 nbSymbols = 0;
     U32 n;
     U32 nextRankStart;
-    HUF_DEltX2* const dt = (HUF_DEltX2*)(DTable + 1);
+    void* const dtPtr = DTable + 1;
+    HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
 
     HUF_STATIC_ASSERT(sizeof(HUF_DEltX2) == sizeof(U16));   /* if compilation fails here, assertion is false */
     //memset(huffWeight, 0, sizeof(huffWeight));   /* is not necessary, even though some analyzer complain ... */
@@ -730,7 +724,8 @@ size_t HUF_decompress1X2_usingDTable(
     BYTE* const oend = op + dstSize;
     size_t errorCode;
     const U32 dtLog = DTable[0];
-    const HUF_DEltX2* const dt = ((const HUF_DEltX2*)DTable) +1;
+    const void* dtPtr = DTable;
+    const HUF_DEltX2* const dt = ((const HUF_DEltX2*)dtPtr)+1;
     BIT_DStream_t bitD;
     errorCode = BIT_initDStream(&bitD, cSrc, cSrcSize);
     if (HUF_isError(errorCode)) return errorCode;
@@ -770,8 +765,8 @@ size_t HUF_decompress4X2_usingDTable(
         const BYTE* const istart = (const BYTE*) cSrc;
         BYTE* const ostart = (BYTE*) dst;
         BYTE* const oend = ostart + dstSize;
-
-        const HUF_DEltX2* const dt = ((const HUF_DEltX2*)DTable) +1;
+        const void* const dtPtr = DTable;
+        const HUF_DEltX2* const dt = ((const HUF_DEltX2*)dtPtr) +1;
         const U32 dtLog = DTable[0];
         size_t errorCode;
 
@@ -978,9 +973,9 @@ size_t HUF_readDTableX4 (U32* DTable, const void* src, size_t srcSize)
     rankVal_t rankVal;
     U32 tableLog, maxW, sizeOfSort, nbSymbols;
     const U32 memLog = DTable[0];
-    const BYTE* ip = (const BYTE*) src;
-    size_t iSize = ip[0];
-    HUF_DEltX4* const dt = ((HUF_DEltX4*)DTable) + 1;
+    size_t iSize;
+    void* dtPtr = DTable;
+    HUF_DEltX4* const dt = ((HUF_DEltX4*)dtPtr) + 1;
 
     HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(U32));   /* if compilation fails here, assertion is false */
     if (memLog > HUF_ABSOLUTEMAX_TABLELOG) return ERROR(tableLog_tooLarge);
@@ -1127,7 +1122,8 @@ size_t HUF_decompress1X4_usingDTable(
     BYTE* const oend = ostart + dstSize;
 
     const U32 dtLog = DTable[0];
-    const HUF_DEltX4* const dt = ((const HUF_DEltX4*)DTable) +1;
+    const void* const dtPtr = DTable;
+    const HUF_DEltX4* const dt = ((const HUF_DEltX4*)dtPtr) +1;
     size_t errorCode;
 
     /* Init */
@@ -1170,8 +1166,8 @@ size_t HUF_decompress4X4_usingDTable(
         const BYTE* const istart = (const BYTE*) cSrc;
         BYTE* const ostart = (BYTE*) dst;
         BYTE* const oend = ostart + dstSize;
-
-        const HUF_DEltX4* const dt = ((const HUF_DEltX4*)DTable) +1;
+        const void* const dtPtr = DTable;
+        const HUF_DEltX4* const dt = ((const HUF_DEltX4*)dtPtr) +1;
         const U32 dtLog = DTable[0];
         size_t errorCode;
 
@@ -1352,8 +1348,7 @@ size_t HUF_readDTableX6 (U32* DTable, const void* src, size_t srcSize)
     U32 tableLog, maxW, sizeOfSort, nbSymbols;
     rankVal_t rankVal;
     const U32 memLog = DTable[0];
-    const BYTE* ip = (const BYTE*) src;
-    size_t iSize = ip[0];
+    size_t iSize;
 
     if (memLog > HUF_ABSOLUTEMAX_TABLELOG) return ERROR(tableLog_tooLarge);
     //memset(weightList, 0, sizeof(weightList));   /* is not necessary, even though some analyzer complain ... */
@@ -1418,8 +1413,10 @@ size_t HUF_readDTableX6 (U32* DTable, const void* src, size_t srcSize)
 
     /* fill tables */
     {
-        HUF_DDescX6* DDescription = (HUF_DDescX6*)(DTable+1);
-        HUF_DSeqX6* DSequence = (HUF_DSeqX6*)(DTable + 1 + ((size_t)1<<(memLog-1)));
+        void* ddPtr = DTable+1;
+        HUF_DDescX6* DDescription = (HUF_DDescX6*)ddPtr;
+        void* dsPtr = DTable + 1 + ((size_t)1<<(memLog-1));
+        HUF_DSeqX6* DSequence = (HUF_DSeqX6*)dsPtr;
         HUF_DSeqX6 DSeq;
         HUF_DDescX6 DDesc;
         DSeq.sequence = 0;
@@ -1478,8 +1475,10 @@ static U32 HUF_decodeLastSymbolsX6(void* op, const U32 maxL, BIT_DStream_t* DStr
 
 static inline size_t HUF_decodeStreamX6(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const U32* DTable, const U32 dtLog)
 {
-    const HUF_DDescX6* dd = (const HUF_DDescX6*)(DTable+1);
-    const HUF_DSeqX6* ds = (const HUF_DSeqX6*)(DTable + 1 + ((size_t)1<<(dtLog-1)));
+    const void* const ddPtr = DTable+1;
+    const HUF_DDescX6* dd = (const HUF_DDescX6*)ddPtr;
+    const void* const dsPtr = DTable + 1 + ((size_t)1<<(dtLog-1));
+    const HUF_DSeqX6* ds = (const HUF_DSeqX6*)dsPtr;
     BYTE* const pStart = p;
 
     /* up to 16 symbols at a time */
@@ -1557,8 +1556,10 @@ size_t HUF_decompress4X6_usingDTable(
         BYTE* const oend = ostart + dstSize;
 
         const U32 dtLog = DTable[0];
-        const HUF_DDescX6* dd = (const HUF_DDescX6*)(DTable+1);
-        const HUF_DSeqX6* ds = (const HUF_DSeqX6*)(DTable + 1 + ((size_t)1<<(dtLog-1)));
+        const void* const ddPtr = DTable+1;
+        const HUF_DDescX6* dd = (const HUF_DDescX6*)ddPtr;
+        const void* const dsPtr = DTable + 1 + ((size_t)1<<(dtLog-1));
+        const HUF_DSeqX6* ds = (const HUF_DSeqX6*)dsPtr;
         size_t errorCode;
 
         /* Init */
diff --git a/lib/huff0.h b/lib/huff0.h
index 613d9b9c..2ebd5cfb 100644
--- a/lib/huff0.h
+++ b/lib/huff0.h
@@ -40,25 +40,25 @@ extern "C" {
 #endif
 
 
-/******************************************
+/* ****************************************
 *  Dependency
 ******************************************/
 #include <stddef.h>    /* size_t */
 
 
-/******************************************
+/* ****************************************
 *  Huff0 simple functions
 ******************************************/
 size_t HUF_compress(void* dst, size_t maxDstSize,
               const void* src, size_t srcSize);
 size_t HUF_decompress(void* dst,  size_t dstSize,
                 const void* cSrc, size_t cSrcSize);
-/*
+/*!
 HUF_compress():
     Compress content of buffer 'src', of size 'srcSize', into destination buffer 'dst'.
     'dst' buffer must be already allocated. Compression runs faster if maxDstSize >= HUF_compressBound(srcSize).
     Note : srcSize must be <= 128 KB
-    return : size of compressed data (<= maxDstSize)
+    @return : size of compressed data (<= maxDstSize)
     Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!!
                      if return == 1, srcData is a single repeated byte symbol (RLE compression)
                      if HUF_isError(return), compression failed (more details using HUF_getErrorName())
@@ -68,12 +68,12 @@ HUF_decompress():
     into already allocated destination buffer 'dst', of size 'dstSize'.
     'dstSize' must be the exact size of original (uncompressed) data.
     Note : in contrast with FSE, HUF_decompress can regenerate RLE (cSrcSize==1) and uncompressed (cSrcSize==dstSize) data, because it knows size to regenerate.
-    return : size of regenerated data (== dstSize)
-             or an error code, which can be tested using HUF_isError()
+    @return : size of regenerated data (== dstSize)
+              or an error code, which can be tested using HUF_isError()
 */
 
 
-/******************************************
+/* ****************************************
 *  Tool functions
 ******************************************/
 size_t HUF_compressBound(size_t size);       /* maximum compressed size */
@@ -83,7 +83,7 @@ unsigned    HUF_isError(size_t code);        /* tells if a return value is an er
 const char* HUF_getErrorName(size_t code);   /* provides error code string (useful for debugging) */
 
 
-/******************************************
+/* ****************************************
 *  Advanced functions
 ******************************************/
 size_t HUF_compress2 (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
diff --git a/lib/huff0_static.h b/lib/huff0_static.h
index e8af19ed..5df0727d 100644
--- a/lib/huff0_static.h
+++ b/lib/huff0_static.h
@@ -40,13 +40,13 @@ extern "C" {
 #endif
 
 
-/******************************************
+/* ****************************************
 *  Dependency
 ******************************************/
 #include "huff0.h"
 
 
-/******************************************
+/* ****************************************
 *  Static allocation macros
 ******************************************/
 /* Huff0 buffer bounds */
@@ -64,14 +64,57 @@ extern "C" {
         unsigned int DTable[HUF_DTABLE_SIZE(maxTableLog) * 3 / 2] = { maxTableLog }
 
 
-/******************************************
-*  Advanced functions
+/* ****************************************
+*  Advanced decompression functions
 ******************************************/
 size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* single-symbol decoder */
 size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbols decoder */
 size_t HUF_decompress4X6 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* quad-symbols decoder */
 
 
+/* ****************************************
+*  Huff0 detailed API
+******************************************/
+/*!
+HUF_compress() does the following:
+1. count symbol occurrence from source[] into table count[] using FSE_count()
+2. build Huffman table from count using HUF_buildCTable()
+3. save Huffman table to memory buffer using HUF_writeCTable()
+4. encode the data stream using HUF_compress_usingCTable()
+
+The following API allows targeting specific sub-functions for advanced tasks.
+For example, it's possible to compress several blocks using the same 'CTable',
+or to save and regenerate 'CTable' using external methods.
+*/
+
+/* FSE_count() : find it within "fse.h" */
+
+typedef struct HUF_CElt_s HUF_CElt;   /* incomplete type */
+size_t HUF_buildCTable (HUF_CElt* tree, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits);
+size_t HUF_writeCTable (void* dst, size_t maxDstSize, const HUF_CElt* tree, unsigned maxSymbolValue, unsigned huffLog);
+size_t HUF_compress_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
+
+
+/*!
+HUF_decompress() does the following:
+1. select the decompression algorithm (X2, X4, X6) based on pre-computed heuristics
+2. build Huffman table from save, using HUF_readDTableXn()
+3. decode 1 or 4 segments in parallel using HUF_decompressSXn_usingDTable
+
+*/
+size_t HUF_readDTableX2 (unsigned short* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX4 (unsigned* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX6 (unsigned* DTable, const void* src, size_t srcSize);
+
+size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* single-symbol decoder */
+size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbol decoder */
+size_t HUF_decompress1X6 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* quad-symbol decoder */
+
+size_t HUF_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned short* DTable);
+size_t HUF_decompress4X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned* DTable);
+size_t HUF_decompress4X6_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const unsigned* DTable);
+
+
 #if defined (__cplusplus)
 }
 #endif