extlz4 0.2.5 → 0.3

data/contrib/lz4/lib/lz4hc.h

@@ -54,7 +54,7 @@ extern "C" {
  *  Block Compression
  **************************************/
 /*! LZ4_compress_HC() :
- *  Compress data from `src` into `dst`, using the more powerful but slower "HC" algorithm.
+ *  Compress data from `src` into `dst`, using the powerful but slower "HC" algorithm.
  * `dst` must be already allocated.
  *  Compression is guaranteed to succeed if `dstCapacity >= LZ4_compressBound(srcSize)` (see "lz4.h")
  *  Max supported `srcSize` value is LZ4_MAX_INPUT_SIZE (see "lz4.h")
@@ -77,7 +77,21 @@ LZ4LIB_API int LZ4_compress_HC (const char* src, char* dst, int srcSize, int dst
  *  Memory segment must be aligned on 8-bytes boundaries (which a normal malloc() should do properly).
  */
 LZ4LIB_API int LZ4_sizeofStateHC(void);
-LZ4LIB_API int LZ4_compress_HC_extStateHC(void* state, const char* src, char* dst, int srcSize, int maxDstSize, int compressionLevel);
+LZ4LIB_API int LZ4_compress_HC_extStateHC(void* stateHC, const char* src, char* dst, int srcSize, int maxDstSize, int compressionLevel);
+
+
+/*! LZ4_compress_HC_destSize() : v1.9.0+
+ *  Will compress as much data as possible from `src`
+ *  to fit into `targetDstSize` budget.
+ *  Result is provided in 2 parts :
+ * @return : the number of bytes written into 'dst' (necessarily <= targetDstSize)
+ *           or 0 if compression fails.
+ * `srcSizePtr` : on success, *srcSizePtr is updated to indicate how much bytes were read from `src`
+ */
+LZ4LIB_API int LZ4_compress_HC_destSize(void* stateHC,
+                                  const char* src, char* dst,
+                                        int* srcSizePtr, int targetDstSize,
+                                        int compressionLevel);
 
 
 /*-************************************
@@ -89,46 +103,92 @@ LZ4LIB_API int LZ4_compress_HC_extStateHC(void* state, const char* src, char* ds
 /*! LZ4_createStreamHC() and LZ4_freeStreamHC() :
  *  These functions create and release memory for LZ4 HC streaming state.
  *  Newly created states are automatically initialized.
- *  Existing states can be re-used several times, using LZ4_resetStreamHC().
- *  These methods are API and ABI stable, they can be used in combination with a DLL.
+ *  A same state can be used multiple times consecutively,
+ *  starting with LZ4_resetStreamHC_fast() to start a new stream of blocks.
  */
 LZ4LIB_API LZ4_streamHC_t* LZ4_createStreamHC(void);
 LZ4LIB_API int             LZ4_freeStreamHC (LZ4_streamHC_t* streamHCPtr);
 
-LZ4LIB_API void LZ4_resetStreamHC (LZ4_streamHC_t* streamHCPtr, int compressionLevel);
-LZ4LIB_API int  LZ4_loadDictHC (LZ4_streamHC_t* streamHCPtr, const char* dictionary, int dictSize);
-
-LZ4LIB_API int LZ4_compress_HC_continue (LZ4_streamHC_t* streamHCPtr, const char* src, char* dst, int srcSize, int maxDstSize);
-
-LZ4LIB_API int LZ4_saveDictHC (LZ4_streamHC_t* streamHCPtr, char* safeBuffer, int maxDictSize);
-
 /*
-  These functions compress data in successive blocks of any size, using previous blocks as dictionary.
+  These functions compress data in successive blocks of any size,
+  using previous blocks as dictionary, to improve compression ratio.
   One key assumption is that previous blocks (up to 64 KB) remain read-accessible while compressing next blocks.
   There is an exception for ring buffers, which can be smaller than 64 KB.
-  Ring buffers scenario is automatically detected and handled by LZ4_compress_HC_continue().
+  Ring-buffer scenario is automatically detected and handled within LZ4_compress_HC_continue().
+
+  Before starting compression, state must be allocated and properly initialized.
+  LZ4_createStreamHC() does both, though compression level is set to LZ4HC_CLEVEL_DEFAULT.
+
+  Selecting the compression level can be done with LZ4_resetStreamHC_fast() (starts a new stream)
+  or LZ4_setCompressionLevel() (anytime, between blocks in the same stream) (experimental).
+  LZ4_resetStreamHC_fast() only works on states which have been properly initialized at least once,
+  which is automatically the case when state is created using LZ4_createStreamHC().
+
+  After reset, a first "fictional block" can be designated as initial dictionary,
+  using LZ4_loadDictHC() (Optional).
+
+  Invoke LZ4_compress_HC_continue() to compress each successive block.
+  The number of blocks is unlimited.
+  Previous input blocks, including initial dictionary when present,
+  must remain accessible and unmodified during compression.
+
+  It's allowed to update compression level anytime between blocks,
+  using LZ4_setCompressionLevel() (experimental).
+
+  'dst' buffer should be sized to handle worst case scenarios
+  (see LZ4_compressBound(), it ensures compression success).
+  In case of failure, the API does not guarantee recovery,
+  so the state _must_ be reset.
+  To ensure compression success
+  whenever `dst` buffer size cannot be made >= LZ4_compressBound(),
+  consider using LZ4_compress_HC_continue_destSize().
+
+  Whenever previous input blocks can't be preserved unmodified in-place during compression of next blocks,
+  it's possible to copy the last blocks into a more stable memory space, using LZ4_saveDictHC().
+  Return value of LZ4_saveDictHC() is the size of dictionary effectively saved into 'safeBuffer' (<= 64 KB)
+
+  After completing a streaming compression,
+  it's possible to start a new stream of blocks, using the same LZ4_streamHC_t state,
+  just by resetting it, using LZ4_resetStreamHC_fast().
+*/
 
-  Before starting compression, state must be properly initialized, using LZ4_resetStreamHC().
-  A first "fictional block" can then be designated as initial dictionary, using LZ4_loadDictHC() (Optional).
+LZ4LIB_API void LZ4_resetStreamHC_fast(LZ4_streamHC_t* streamHCPtr, int compressionLevel);   /* v1.9.0+ */
+LZ4LIB_API int  LZ4_loadDictHC (LZ4_streamHC_t* streamHCPtr, const char* dictionary, int dictSize);
 
-  Then, use LZ4_compress_HC_continue() to compress each successive block.
-  Previous memory blocks (including initial dictionary when present) must remain accessible and unmodified during compression.
-  'dst' buffer should be sized to handle worst case scenarios (see LZ4_compressBound()), to ensure operation success.
-  Because in case of failure, the API does not guarantee context recovery, and context will have to be reset.
-  If `dst` buffer budget cannot be >= LZ4_compressBound(), consider using LZ4_compress_HC_continue_destSize() instead.
+LZ4LIB_API int LZ4_compress_HC_continue (LZ4_streamHC_t* streamHCPtr,
+                                   const char* src, char* dst,
+                                         int srcSize, int maxDstSize);
 
-  If, for any reason, previous data block can't be preserved unmodified in memory for next compression block,
-  you can save it to a more stable memory space, using LZ4_saveDictHC().
-  Return value of LZ4_saveDictHC() is the size of dictionary effectively saved into 'safeBuffer'.
-*/
+/*! LZ4_compress_HC_continue_destSize() : v1.9.0+
+ *  Similar to LZ4_compress_HC_continue(),
+ *  but will read as much data as possible from `src`
+ *  to fit into `targetDstSize` budget.
+ *  Result is provided into 2 parts :
+ * @return : the number of bytes written into 'dst' (necessarily <= targetDstSize)
+ *           or 0 if compression fails.
+ * `srcSizePtr` : on success, *srcSizePtr will be updated to indicate how much bytes were read from `src`.
+ *           Note that this function may not consume the entire input.
+ */
+LZ4LIB_API int LZ4_compress_HC_continue_destSize(LZ4_streamHC_t* LZ4_streamHCPtr,
+                                           const char* src, char* dst,
+                                                 int* srcSizePtr, int targetDstSize);
+
+LZ4LIB_API int LZ4_saveDictHC (LZ4_streamHC_t* streamHCPtr, char* safeBuffer, int maxDictSize);
 
 
-/*-**************************************************************
+
+/*^**********************************************
+ * !!!!!!   STATIC LINKING ONLY   !!!!!!
+ ***********************************************/
+
+/*-******************************************************************
  * PRIVATE DEFINITIONS :
- * Do not use these definitions.
- * They are exposed to allow static allocation of `LZ4_streamHC_t`.
- * Using these definitions makes the code vulnerable to potential API break when upgrading LZ4
- ****************************************************************/
+ * Do not use these definitions directly.
+ * They are merely exposed to allow static allocation of `LZ4_streamHC_t`.
+ * Declare an `LZ4_streamHC_t` directly, rather than any type below.
+ * Even then, only do so in the context of static linking, as definitions may change between versions.
+ ********************************************************************/
+
 #define LZ4HC_DICTIONARY_LOGSIZE 16
 #define LZ4HC_MAXD (1<<LZ4HC_DICTIONARY_LOGSIZE)
 #define LZ4HC_MAXD_MASK (LZ4HC_MAXD - 1)
@@ -141,52 +201,75 @@ LZ4LIB_API int LZ4_saveDictHC (LZ4_streamHC_t* streamHCPtr, char* safeBuffer, in
 #if defined(__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
 #include <stdint.h>
 
-typedef struct
+typedef struct LZ4HC_CCtx_internal LZ4HC_CCtx_internal;
+struct LZ4HC_CCtx_internal
 {
     uint32_t   hashTable[LZ4HC_HASHTABLESIZE];
     uint16_t   chainTable[LZ4HC_MAXD];
     const uint8_t* end;         /* next block here to continue on current prefix */
     const uint8_t* base;        /* All index relative to this position */
     const uint8_t* dictBase;    /* alternate base for extDict */
-    uint8_t* inputBuffer;       /* deprecated */
     uint32_t   dictLimit;       /* below that point, need extDict */
     uint32_t   lowLimit;        /* below that point, no more dict */
     uint32_t   nextToUpdate;    /* index from which to continue dictionary update */
-    int        compressionLevel;
-} LZ4HC_CCtx_internal;
+    short      compressionLevel;
+    int8_t     favorDecSpeed;   /* favor decompression speed if this flag set,
+                                   otherwise, favor compression ratio */
+    int8_t     dirty;           /* stream has to be fully reset if this flag is set */
+    const LZ4HC_CCtx_internal* dictCtx;
+};
 
 #else
 
-typedef struct
+typedef struct LZ4HC_CCtx_internal LZ4HC_CCtx_internal;
+struct LZ4HC_CCtx_internal
 {
     unsigned int   hashTable[LZ4HC_HASHTABLESIZE];
     unsigned short chainTable[LZ4HC_MAXD];
     const unsigned char* end;        /* next block here to continue on current prefix */
     const unsigned char* base;       /* All index relative to this position */
     const unsigned char* dictBase;   /* alternate base for extDict */
-    unsigned char* inputBuffer;      /* deprecated */
     unsigned int   dictLimit;        /* below that point, need extDict */
     unsigned int   lowLimit;         /* below that point, no more dict */
     unsigned int   nextToUpdate;     /* index from which to continue dictionary update */
-    int            compressionLevel;
-} LZ4HC_CCtx_internal;
+    short          compressionLevel;
+    char           favorDecSpeed;    /* favor decompression speed if this flag set,
+                                        otherwise, favor compression ratio */
+    char           dirty;            /* stream has to be fully reset if this flag is set */
+    const LZ4HC_CCtx_internal* dictCtx;
+};
 
 #endif
 
-#define LZ4_STREAMHCSIZE       (4*LZ4HC_HASHTABLESIZE + 2*LZ4HC_MAXD + 56) /* 262200 */
+
+/* Do not use these definitions directly !
+ * Declare or allocate an LZ4_streamHC_t instead.
+ */
+#define LZ4_STREAMHCSIZE       (4*LZ4HC_HASHTABLESIZE + 2*LZ4HC_MAXD + 56 + ((sizeof(void*)==16) ? 56 : 0) /* AS400*/ ) /* 262200 or 262256*/
 #define LZ4_STREAMHCSIZE_SIZET (LZ4_STREAMHCSIZE / sizeof(size_t))
 union LZ4_streamHC_u {
     size_t table[LZ4_STREAMHCSIZE_SIZET];
     LZ4HC_CCtx_internal internal_donotuse;
-};   /* previously typedef'd to LZ4_streamHC_t */
-/*
-  LZ4_streamHC_t :
-  This structure allows static allocation of LZ4 HC streaming state.
-  State must be initialized using LZ4_resetStreamHC() before first use.
+}; /* previously typedef'd to LZ4_streamHC_t */
+
+/* LZ4_streamHC_t :
+ * This structure allows static allocation of LZ4 HC streaming state.
+ * This can be used to allocate statically, on state, or as part of a larger structure.
+ *
+ * Such state **must** be initialized using LZ4_initStreamHC() before first use.
+ *
+ * Note that invoking LZ4_initStreamHC() is not required when
+ * the state was created using LZ4_createStreamHC() (which is recommended).
+ * Using the normal builder, a newly created state is automatically initialized.
+ *
+ * Static allocation shall only be used in combination with static linking.
+ */
 
-  Static allocation shall only be used in combination with static linking.
-  When invoking LZ4 from a DLL, use create/free functions instead, which are API and ABI stable.
-*/
+/* LZ4_initStreamHC() : v1.9.0+
+ * Required before first use of a statically allocated LZ4_streamHC_t.
+ * Before v1.9.0 : use LZ4_resetStreamHC() instead
+ */
+LZ4LIB_API LZ4_streamHC_t* LZ4_initStreamHC (void* buffer, size_t size);
 
 
 /*-************************************
@@ -195,25 +278,43 @@ union LZ4_streamHC_u {
 /* see lz4.h LZ4_DISABLE_DEPRECATE_WARNINGS to turn off deprecation warnings */
 
 /* deprecated compression functions */
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC() instead") int LZ4_compressHC               (const char* source, char* dest, int inputSize);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC() instead") int LZ4_compressHC_limitedOutput (const char* source, char* dest, int inputSize, int maxOutputSize);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC() instead") int LZ4_compressHC2 (const char* source, char* dest, int inputSize, int compressionLevel);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC() instead") int LZ4_compressHC2_limitedOutput (const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") int LZ4_compressHC_withStateHC               (void* state, const char* source, char* dest, int inputSize);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") int LZ4_compressHC_limitedOutput_withStateHC (void* state, const char* source, char* dest, int inputSize, int maxOutputSize);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") int LZ4_compressHC2_withStateHC (void* state, const char* source, char* dest, int inputSize, int compressionLevel);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") int LZ4_compressHC2_limitedOutput_withStateHC(void* state, const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") int LZ4_compressHC_continue               (LZ4_streamHC_t* LZ4_streamHCPtr, const char* source, char* dest, int inputSize);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") int LZ4_compressHC_limitedOutput_continue (LZ4_streamHC_t* LZ4_streamHCPtr, const char* source, char* dest, int inputSize, int maxOutputSize);
-
-/* Deprecated Streaming functions using older model; should no longer be used */
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_createStreamHC() instead") void* LZ4_createHC (char* inputBuffer);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_saveDictHC() instead")     char* LZ4_slideInputBufferHC (void* LZ4HC_Data);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_freeStreamHC() instead")   int   LZ4_freeHC (void* LZ4HC_Data);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") int LZ4_compressHC2_continue (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int compressionLevel);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") int LZ4_compressHC2_limitedOutput_continue (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_createStreamHC() instead") int   LZ4_sizeofStreamStateHC(void);
-LZ4LIB_API LZ4_DEPRECATED("use LZ4_resetStreamHC() instead")  int   LZ4_resetStreamStateHC(void* state, char* inputBuffer);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC               (const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC_limitedOutput (const char* source, char* dest, int inputSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC2              (const char* source, char* dest, int inputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC() instead") LZ4LIB_API int LZ4_compressHC2_limitedOutput(const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC_withStateHC               (void* state, const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC_limitedOutput_withStateHC (void* state, const char* source, char* dest, int inputSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC2_withStateHC              (void* state, const char* source, char* dest, int inputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_extStateHC() instead") LZ4LIB_API int LZ4_compressHC2_limitedOutput_withStateHC(void* state, const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC_continue               (LZ4_streamHC_t* LZ4_streamHCPtr, const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC_limitedOutput_continue (LZ4_streamHC_t* LZ4_streamHCPtr, const char* source, char* dest, int inputSize, int maxOutputSize);
+
+/* Obsolete streaming functions; degraded functionality; do not use!
+ *
+ * In order to perform streaming compression, these functions depended on data
+ * that is no longer tracked in the state. They have been preserved as well as
+ * possible: using them will still produce a correct output. However, use of
+ * LZ4_slideInputBufferHC() will truncate the history of the stream, rather
+ * than preserve a window-sized chunk of history.
+ */
+LZ4_DEPRECATED("use LZ4_createStreamHC() instead") LZ4LIB_API void* LZ4_createHC (const char* inputBuffer);
+LZ4_DEPRECATED("use LZ4_saveDictHC() instead") LZ4LIB_API     char* LZ4_slideInputBufferHC (void* LZ4HC_Data);
+LZ4_DEPRECATED("use LZ4_freeStreamHC() instead") LZ4LIB_API   int   LZ4_freeHC (void* LZ4HC_Data);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC2_continue               (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_compress_HC_continue() instead") LZ4LIB_API int LZ4_compressHC2_limitedOutput_continue (void* LZ4HC_Data, const char* source, char* dest, int inputSize, int maxOutputSize, int compressionLevel);
+LZ4_DEPRECATED("use LZ4_createStreamHC() instead") LZ4LIB_API int   LZ4_sizeofStreamStateHC(void);
+LZ4_DEPRECATED("use LZ4_initStreamHC() instead") LZ4LIB_API  int   LZ4_resetStreamStateHC(void* state, char* inputBuffer);
+
+
+/* LZ4_resetStreamHC() is now replaced by LZ4_initStreamHC().
+ * The intention is to emphasize the difference with LZ4_resetStreamHC_fast(),
+ * which is now the recommended function to start a new stream of blocks,
+ * but cannot be used to initialize a memory segment containing arbitrary garbage data.
+ *
+ * It is recommended to switch to LZ4_initStreamHC().
+ * LZ4_resetStreamHC() will generate deprecation warnings in a future version.
+ */
+LZ4LIB_API void LZ4_resetStreamHC (LZ4_streamHC_t* streamHCPtr, int compressionLevel);
 
 
 #if defined (__cplusplus)
@@ -235,38 +336,100 @@ LZ4LIB_API LZ4_DEPRECATED("use LZ4_resetStreamHC() instead")  int   LZ4_resetStr
 #ifndef LZ4_HC_SLO_098092834
 #define LZ4_HC_SLO_098092834
 
-/*! LZ4_compress_HC_destSize() : v1.8.0 (experimental)
- *  Will try to compress as much data from `src` as possible
- *  that can fit into `targetDstSize` budget.
- *  Result is provided in 2 parts :
- * @return : the number of bytes written into 'dst'
- *           or 0 if compression fails.
- * `srcSizePtr` : value will be updated to indicate how much bytes were read from `src`
- */
-int LZ4_compress_HC_destSize(void* LZ4HC_Data,
-                            const char* src, char* dst,
-                            int* srcSizePtr, int targetDstSize,
-                            int compressionLevel);
-
-/*! LZ4_compress_HC_continue_destSize() : v1.8.0 (experimental)
- *  Similar as LZ4_compress_HC_continue(),
- *  but will read a variable nb of bytes from `src`
- *  to fit into `targetDstSize` budget.
- *  Result is provided in 2 parts :
- * @return : the number of bytes written into 'dst'
- *           or 0 if compression fails.
- * `srcSizePtr` : value will be updated to indicate how much bytes were read from `src`.
- */
-int LZ4_compress_HC_continue_destSize(LZ4_streamHC_t* LZ4_streamHCPtr,
-                            const char* src, char* dst,
-                            int* srcSizePtr, int targetDstSize);
+#if defined (__cplusplus)
+extern "C" {
+#endif
 
-/*! LZ4_setCompressionLevel() : v1.8.0 (experimental)
- *  It's possible to change compression level between 2 invocations of LZ4_compress_HC_continue*()
+/*! LZ4_setCompressionLevel() : v1.8.0+ (experimental)
+ *  It's possible to change compression level
+ *  between successive invocations of LZ4_compress_HC_continue*()
+ *  for dynamic adaptation.
  */
-void LZ4_setCompressionLevel(LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel);
+LZ4LIB_STATIC_API void LZ4_setCompressionLevel(
+    LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel);
 
+/*! LZ4_favorDecompressionSpeed() : v1.8.2+ (experimental)
+ *  Opt. Parser will favor decompression speed over compression ratio.
+ *  Only applicable to levels >= LZ4HC_CLEVEL_OPT_MIN.
+ */
+LZ4LIB_STATIC_API void LZ4_favorDecompressionSpeed(
+    LZ4_streamHC_t* LZ4_streamHCPtr, int favor);
+
+/*! LZ4_resetStreamHC_fast() : v1.9.0+
+ *  When an LZ4_streamHC_t is known to be in a internally coherent state,
+ *  it can often be prepared for a new compression with almost no work, only
+ *  sometimes falling back to the full, expensive reset that is always required
+ *  when the stream is in an indeterminate state (i.e., the reset performed by
+ *  LZ4_resetStreamHC()).
+ *
+ *  LZ4_streamHCs are guaranteed to be in a valid state when:
+ *  - returned from LZ4_createStreamHC()
+ *  - reset by LZ4_resetStreamHC()
+ *  - memset(stream, 0, sizeof(LZ4_streamHC_t))
+ *  - the stream was in a valid state and was reset by LZ4_resetStreamHC_fast()
+ *  - the stream was in a valid state and was then used in any compression call
+ *    that returned success
+ *  - the stream was in an indeterminate state and was used in a compression
+ *    call that fully reset the state (LZ4_compress_HC_extStateHC()) and that
+ *    returned success
+ *
+ *  Note:
+ *  A stream that was last used in a compression call that returned an error
+ *  may be passed to this function. However, it will be fully reset, which will
+ *  clear any existing history and settings from the context.
+ */
+LZ4LIB_STATIC_API void LZ4_resetStreamHC_fast(
+    LZ4_streamHC_t* LZ4_streamHCPtr, int compressionLevel);
+
+/*! LZ4_compress_HC_extStateHC_fastReset() :
+ *  A variant of LZ4_compress_HC_extStateHC().
+ *
+ *  Using this variant avoids an expensive initialization step. It is only safe
+ *  to call if the state buffer is known to be correctly initialized already
+ *  (see above comment on LZ4_resetStreamHC_fast() for a definition of
+ *  "correctly initialized"). From a high level, the difference is that this
+ *  function initializes the provided state with a call to
+ *  LZ4_resetStreamHC_fast() while LZ4_compress_HC_extStateHC() starts with a
+ *  call to LZ4_resetStreamHC().
+ */
+LZ4LIB_STATIC_API int LZ4_compress_HC_extStateHC_fastReset (
+    void* state,
+    const char* src, char* dst,
+    int srcSize, int dstCapacity,
+    int compressionLevel);
+
+/*! LZ4_attach_HC_dictionary() :
+ *  This is an experimental API that allows for the efficient use of a
+ *  static dictionary many times.
+ *
+ *  Rather than re-loading the dictionary buffer into a working context before
+ *  each compression, or copying a pre-loaded dictionary's LZ4_streamHC_t into a
+ *  working LZ4_streamHC_t, this function introduces a no-copy setup mechanism,
+ *  in which the working stream references the dictionary stream in-place.
+ *
+ *  Several assumptions are made about the state of the dictionary stream.
+ *  Currently, only streams which have been prepared by LZ4_loadDictHC() should
+ *  be expected to work.
+ *
+ *  Alternatively, the provided dictionary stream pointer may be NULL, in which
+ *  case any existing dictionary stream is unset.
+ *
+ *  A dictionary should only be attached to a stream without any history (i.e.,
+ *  a stream that has just been reset).
+ *
+ *  The dictionary will remain attached to the working stream only for the
+ *  current stream session. Calls to LZ4_resetStreamHC(_fast) will remove the
+ *  dictionary context association from the working stream. The dictionary
+ *  stream (and source buffer) must remain in-place / accessible / unchanged
+ *  through the lifetime of the stream session.
+ */
+LZ4LIB_STATIC_API void LZ4_attach_HC_dictionary(
+          LZ4_streamHC_t *working_stream,
+    const LZ4_streamHC_t *dictionary_stream);
 
+#if defined (__cplusplus)
+}
+#endif
 
 #endif   /* LZ4_HC_SLO_098092834 */
 #endif   /* LZ4_HC_STATIC_LINKING_ONLY */

data/contrib/lz4/lib/xxhash.c

@@ -50,20 +50,26 @@
  * Prefer these methods in priority order (0 > 1 > 2)
  */
 #ifndef XXH_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
-#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
+#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) \
+                        || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) \
+                        || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
 #    define XXH_FORCE_MEMORY_ACCESS 2
-#  elif defined(__INTEL_COMPILER) || \
-  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#  elif (defined(__INTEL_COMPILER) && !defined(_WIN32)) || \
+  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) \
+                    || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) \
+                    || defined(__ARM_ARCH_7S__) ))
 #    define XXH_FORCE_MEMORY_ACCESS 1
 #  endif
 #endif
 
 /*!XXH_ACCEPT_NULL_INPUT_POINTER :
- * If the input pointer is a null pointer, xxHash default behavior is to trigger a memory access error, since it is a bad pointer.
- * When this option is enabled, xxHash output for null input pointers will be the same as a null-length input.
- * By default, this option is disabled. To enable it, uncomment below define :
+ * If input pointer is NULL, xxHash default behavior is to dereference it, triggering a segfault.
+ * When this macro is enabled, xxHash actively checks input for null pointer.
+ * It it is, result for null input pointers is the same as a null-length input.
  */
-/* #define XXH_ACCEPT_NULL_INPUT_POINTER 1 */
+#ifndef XXH_ACCEPT_NULL_INPUT_POINTER   /* can be defined externally */
+#  define XXH_ACCEPT_NULL_INPUT_POINTER 0
+#endif
 
 /*!XXH_FORCE_NATIVE_FORMAT :
  * By default, xxHash library provides endian-independent Hash values, based on little-endian convention.
@@ -80,8 +86,9 @@
 /*!XXH_FORCE_ALIGN_CHECK :
  * This is a minor performance trick, only useful with lots of very small keys.
  * It means : check for aligned/unaligned input.
- * The check costs one initial branch per hash; set to 0 when the input data
- * is guaranteed to be aligned.
+ * The check costs one initial branch per hash;
+ * set it to 0 when the input is guaranteed to be aligned,
+ * or when alignment doesn't matter for performance.
  */
 #ifndef XXH_FORCE_ALIGN_CHECK /* can be defined externally */
 #  if defined(__i386) || defined(_M_IX86) || defined(__x86_64__) || defined(_M_X64)
@@ -104,6 +111,8 @@ static void  XXH_free  (void* p)  { free(p); }
 #include <string.h>
 static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcpy(dest,src,size); }
 
+#include <assert.h>   /* assert */
+
 #define XXH_STATIC_LINKING_ONLY
 #include "xxhash.h"
 
@@ -113,40 +122,35 @@ static void* XXH_memcpy(void* dest, const void* src, size_t size) { return memcp
 ***************************************/
 #ifdef _MSC_VER    /* Visual Studio */
 #  pragma warning(disable : 4127)      /* disable: C4127: conditional expression is constant */
-#endif
-
-#ifndef XXH_FORCE_INLINE
-#  ifdef _MSC_VER    /* Visual Studio */
-#    define XXH_FORCE_INLINE static __forceinline
-#  else
-#    if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
-#      ifdef __GNUC__
-#        define XXH_FORCE_INLINE static inline __attribute__((always_inline))
-#      else
-#        define XXH_FORCE_INLINE static inline
-#      endif
+#  define FORCE_INLINE static __forceinline
+#else
+#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#    ifdef __GNUC__
+#      define FORCE_INLINE static inline __attribute__((always_inline))
 #    else
-#      define XXH_FORCE_INLINE static
-#    endif /* __STDC_VERSION__ */
-#  endif  /* _MSC_VER */
-#endif /* XXH_FORCE_INLINE */
+#      define FORCE_INLINE static inline
+#    endif
+#  else
+#    define FORCE_INLINE static
+#  endif /* __STDC_VERSION__ */
+#endif
 
 
 /* *************************************
 *  Basic Types
 ***************************************/
 #ifndef MEM_MODULE
-# if !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
+# if !defined (__VMS) \
+  && (defined (__cplusplus) \
+  || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
 #   include <stdint.h>
     typedef uint8_t  BYTE;
     typedef uint16_t U16;
     typedef uint32_t U32;
-    typedef  int32_t S32;
 # else
     typedef unsigned char      BYTE;
     typedef unsigned short     U16;
     typedef unsigned int       U32;
-    typedef   signed int       S32;
 # endif
 #endif
 
@@ -213,8 +217,12 @@ typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
 
 /* XXH_CPU_LITTLE_ENDIAN can be defined externally, for example on the compiler command line */
 #ifndef XXH_CPU_LITTLE_ENDIAN
-    static const int g_one = 1;
-#   define XXH_CPU_LITTLE_ENDIAN   (*(const char*)(&g_one))
+static int XXH_isLittleEndian(void)
+{
+    const union { U32 u; BYTE c[4]; } one = { 1 };   /* don't use static : performance detrimental  */
+    return one.c[0];
+}
+#   define XXH_CPU_LITTLE_ENDIAN   XXH_isLittleEndian()
 #endif
 
 
@@ -223,7 +231,7 @@ typedef enum { XXH_bigEndian=0, XXH_littleEndian=1 } XXH_endianess;
 *****************************/
 typedef enum { XXH_aligned, XXH_unaligned } XXH_alignment;
 
-XXH_FORCE_INLINE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
+FORCE_INLINE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
 {
     if (align==XXH_unaligned)
         return endian==XXH_littleEndian ? XXH_read32(ptr) : XXH_swap32(XXH_read32(ptr));
@@ -231,7 +239,7 @@ XXH_FORCE_INLINE U32 XXH_readLE32_align(const void* ptr, XXH_endianess endian, X
         return endian==XXH_littleEndian ? *(const U32*)ptr : XXH_swap32(*(const U32*)ptr);
 }
 
-XXH_FORCE_INLINE U32 XXH_readLE32(const void* ptr, XXH_endianess endian)
+FORCE_INLINE U32 XXH_readLE32(const void* ptr, XXH_endianess endian)
 {
     return XXH_readLE32_align(ptr, endian, XXH_unaligned);
 }
@@ -245,12 +253,12 @@ static U32 XXH_readBE32(const void* ptr)
 /* *************************************
 *  Macros
 ***************************************/
-#define XXH_STATIC_ASSERT(c)   { enum { XXH_static_assert = 1/(int)(!!(c)) }; }    /* use only *after* variable declarations */
+#define XXH_STATIC_ASSERT(c)  { enum { XXH_sa = 1/(int)(!!(c)) }; }  /* use after variable declarations */
 XXH_PUBLIC_API unsigned XXH_versionNumber (void) { return XXH_VERSION_NUMBER; }
 
 
 /* *******************************************************************
-*  32-bits hash functions
+*  32-bit hash functions
 *********************************************************************/
 static const U32 PRIME32_1 = 2654435761U;
 static const U32 PRIME32_2 = 2246822519U;
@@ -266,14 +274,89 @@ static U32 XXH32_round(U32 seed, U32 input)
     return seed;
 }
 
-XXH_FORCE_INLINE U32 XXH32_endian_align(const void* input, size_t len, U32 seed, XXH_endianess endian, XXH_alignment align)
+/* mix all bits */
+static U32 XXH32_avalanche(U32 h32)
+{
+    h32 ^= h32 >> 15;
+    h32 *= PRIME32_2;
+    h32 ^= h32 >> 13;
+    h32 *= PRIME32_3;
+    h32 ^= h32 >> 16;
+    return(h32);
+}
+
+#define XXH_get32bits(p) XXH_readLE32_align(p, endian, align)
+
+static U32
+XXH32_finalize(U32 h32, const void* ptr, size_t len,
+                XXH_endianess endian, XXH_alignment align)
+
+{
+    const BYTE* p = (const BYTE*)ptr;
+
+#define PROCESS1               \
+    h32 += (*p++) * PRIME32_5; \
+    h32 = XXH_rotl32(h32, 11) * PRIME32_1 ;
+
+#define PROCESS4                         \
+    h32 += XXH_get32bits(p) * PRIME32_3; \
+    p+=4;                                \
+    h32  = XXH_rotl32(h32, 17) * PRIME32_4 ;
+
+    switch(len&15)  /* or switch(bEnd - p) */
+    {
+      case 12:      PROCESS4;
+                    /* fallthrough */
+      case 8:       PROCESS4;
+                    /* fallthrough */
+      case 4:       PROCESS4;
+                    return XXH32_avalanche(h32);
+
+      case 13:      PROCESS4;
+                    /* fallthrough */
+      case 9:       PROCESS4;
+                    /* fallthrough */
+      case 5:       PROCESS4;
+                    PROCESS1;
+                    return XXH32_avalanche(h32);
+
+      case 14:      PROCESS4;
+                    /* fallthrough */
+      case 10:      PROCESS4;
+                    /* fallthrough */
+      case 6:       PROCESS4;
+                    PROCESS1;
+                    PROCESS1;
+                    return XXH32_avalanche(h32);
+
+      case 15:      PROCESS4;
+                    /* fallthrough */
+      case 11:      PROCESS4;
+                    /* fallthrough */
+      case 7:       PROCESS4;
+                    /* fallthrough */
+      case 3:       PROCESS1;
+                    /* fallthrough */
+      case 2:       PROCESS1;
+                    /* fallthrough */
+      case 1:       PROCESS1;
+                    /* fallthrough */
+      case 0:       return XXH32_avalanche(h32);
+    }
+    assert(0);
+    return h32;   /* reaching this point is deemed impossible */
+}
+
+
+FORCE_INLINE U32
+XXH32_endian_align(const void* input, size_t len, U32 seed,
+                    XXH_endianess endian, XXH_alignment align)
 {
     const BYTE* p = (const BYTE*)input;
     const BYTE* bEnd = p + len;
     U32 h32;
-#define XXH_get32bits(p) XXH_readLE32_align(p, endian, align)
 
-#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
     if (p==NULL) {
         len=0;
         bEnd=p=(const BYTE*)(size_t)16;
@@ -281,7 +364,7 @@ XXH_FORCE_INLINE U32 XXH32_endian_align(const void* input, size_t len, U32 seed,
 #endif
 
     if (len>=16) {
-        const BYTE* const limit = bEnd - 16;
+        const BYTE* const limit = bEnd - 15;
         U32 v1 = seed + PRIME32_1 + PRIME32_2;
         U32 v2 = seed + PRIME32_2;
         U32 v3 = seed + 0;
@@ -292,34 +375,17 @@ XXH_FORCE_INLINE U32 XXH32_endian_align(const void* input, size_t len, U32 seed,
             v2 = XXH32_round(v2, XXH_get32bits(p)); p+=4;
             v3 = XXH32_round(v3, XXH_get32bits(p)); p+=4;
             v4 = XXH32_round(v4, XXH_get32bits(p)); p+=4;
-        } while (p<=limit);
+        } while (p < limit);
 
-        h32 = XXH_rotl32(v1, 1) + XXH_rotl32(v2, 7) + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
+        h32 = XXH_rotl32(v1, 1)  + XXH_rotl32(v2, 7)
+            + XXH_rotl32(v3, 12) + XXH_rotl32(v4, 18);
     } else {
         h32  = seed + PRIME32_5;
     }
 
-    h32 += (U32) len;
-
-    while (p+4<=bEnd) {
-        h32 += XXH_get32bits(p) * PRIME32_3;
-        h32  = XXH_rotl32(h32, 17) * PRIME32_4 ;
-        p+=4;
-    }
-
-    while (p<bEnd) {
-        h32 += (*p) * PRIME32_5;
-        h32 = XXH_rotl32(h32, 11) * PRIME32_1 ;
-        p++;
-    }
-
-    h32 ^= h32 >> 15;
-    h32 *= PRIME32_2;
-    h32 ^= h32 >> 13;
-    h32 *= PRIME32_3;
-    h32 ^= h32 >> 16;
+    h32 += (U32)len;
 
-    return h32;
+    return XXH32_finalize(h32, p, len&15, endian, align);
 }
 
 
@@ -371,74 +437,81 @@ XXH_PUBLIC_API void XXH32_copyState(XXH32_state_t* dstState, const XXH32_state_t
 XXH_PUBLIC_API XXH_errorcode XXH32_reset(XXH32_state_t* statePtr, unsigned int seed)
 {
     XXH32_state_t state;   /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
-    memset(&state, 0, sizeof(state)-4);   /* do not write into reserved, for future removal */
+    memset(&state, 0, sizeof(state));
     state.v1 = seed + PRIME32_1 + PRIME32_2;
     state.v2 = seed + PRIME32_2;
     state.v3 = seed + 0;
     state.v4 = seed - PRIME32_1;
-    memcpy(statePtr, &state, sizeof(state));
+    /* do not write into reserved, planned to be removed in a future version */
+    memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved));
     return XXH_OK;
 }
 
 
-XXH_FORCE_INLINE XXH_errorcode XXH32_update_endian (XXH32_state_t* state, const void* input, size_t len, XXH_endianess endian)
+FORCE_INLINE XXH_errorcode
+XXH32_update_endian(XXH32_state_t* state, const void* input, size_t len, XXH_endianess endian)
 {
-    const BYTE* p = (const BYTE*)input;
-    const BYTE* const bEnd = p + len;
-
-#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
-    if (input==NULL) return XXH_ERROR;
+    if (input==NULL)
+#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
+        return XXH_OK;
+#else
+        return XXH_ERROR;
 #endif
 
-    state->total_len_32 += (unsigned)len;
-    state->large_len |= (len>=16) | (state->total_len_32>=16);
+    {   const BYTE* p = (const BYTE*)input;
+        const BYTE* const bEnd = p + len;
 
-    if (state->memsize + len < 16)  {   /* fill in tmp buffer */
-        XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len);
-        state->memsize += (unsigned)len;
-        return XXH_OK;
-    }
+        state->total_len_32 += (unsigned)len;
+        state->large_len |= (len>=16) | (state->total_len_32>=16);
 
-    if (state->memsize) {   /* some data left from previous update */
-        XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize);
-        {   const U32* p32 = state->mem32;
-            state->v1 = XXH32_round(state->v1, XXH_readLE32(p32, endian)); p32++;
-            state->v2 = XXH32_round(state->v2, XXH_readLE32(p32, endian)); p32++;
-            state->v3 = XXH32_round(state->v3, XXH_readLE32(p32, endian)); p32++;
-            state->v4 = XXH32_round(state->v4, XXH_readLE32(p32, endian)); p32++;
+        if (state->memsize + len < 16)  {   /* fill in tmp buffer */
+            XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, len);
+            state->memsize += (unsigned)len;
+            return XXH_OK;
         }
-        p += 16-state->memsize;
-        state->memsize = 0;
-    }
-
-    if (p <= bEnd-16) {
-        const BYTE* const limit = bEnd - 16;
-        U32 v1 = state->v1;
-        U32 v2 = state->v2;
-        U32 v3 = state->v3;
-        U32 v4 = state->v4;
 
-        do {
-            v1 = XXH32_round(v1, XXH_readLE32(p, endian)); p+=4;
-            v2 = XXH32_round(v2, XXH_readLE32(p, endian)); p+=4;
-            v3 = XXH32_round(v3, XXH_readLE32(p, endian)); p+=4;
-            v4 = XXH32_round(v4, XXH_readLE32(p, endian)); p+=4;
-        } while (p<=limit);
+        if (state->memsize) {   /* some data left from previous update */
+            XXH_memcpy((BYTE*)(state->mem32) + state->memsize, input, 16-state->memsize);
+            {   const U32* p32 = state->mem32;
+                state->v1 = XXH32_round(state->v1, XXH_readLE32(p32, endian)); p32++;
+                state->v2 = XXH32_round(state->v2, XXH_readLE32(p32, endian)); p32++;
+                state->v3 = XXH32_round(state->v3, XXH_readLE32(p32, endian)); p32++;
+                state->v4 = XXH32_round(state->v4, XXH_readLE32(p32, endian));
+            }
+            p += 16-state->memsize;
+            state->memsize = 0;
+        }
 
-        state->v1 = v1;
-        state->v2 = v2;
-        state->v3 = v3;
-        state->v4 = v4;
-    }
+        if (p <= bEnd-16) {
+            const BYTE* const limit = bEnd - 16;
+            U32 v1 = state->v1;
+            U32 v2 = state->v2;
+            U32 v3 = state->v3;
+            U32 v4 = state->v4;
+
+            do {
+                v1 = XXH32_round(v1, XXH_readLE32(p, endian)); p+=4;
+                v2 = XXH32_round(v2, XXH_readLE32(p, endian)); p+=4;
+                v3 = XXH32_round(v3, XXH_readLE32(p, endian)); p+=4;
+                v4 = XXH32_round(v4, XXH_readLE32(p, endian)); p+=4;
+            } while (p<=limit);
+
+            state->v1 = v1;
+            state->v2 = v2;
+            state->v3 = v3;
+            state->v4 = v4;
+        }
 
-    if (p < bEnd) {
-        XXH_memcpy(state->mem32, p, (size_t)(bEnd-p));
-        state->memsize = (unsigned)(bEnd-p);
+        if (p < bEnd) {
+            XXH_memcpy(state->mem32, p, (size_t)(bEnd-p));
+            state->memsize = (unsigned)(bEnd-p);
+        }
     }
 
     return XXH_OK;
 }
 
+
 XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void* input, size_t len)
 {
     XXH_endianess endian_detected = (XXH_endianess)XXH_CPU_LITTLE_ENDIAN;
@@ -450,40 +523,23 @@ XXH_PUBLIC_API XXH_errorcode XXH32_update (XXH32_state_t* state_in, const void*
 }
 
 
-
-XXH_FORCE_INLINE U32 XXH32_digest_endian (const XXH32_state_t* state, XXH_endianess endian)
+FORCE_INLINE U32
+XXH32_digest_endian (const XXH32_state_t* state, XXH_endianess endian)
 {
-    const BYTE * p = (const BYTE*)state->mem32;
-    const BYTE* const bEnd = (const BYTE*)(state->mem32) + state->memsize;
     U32 h32;
 
     if (state->large_len) {
-        h32 = XXH_rotl32(state->v1, 1) + XXH_rotl32(state->v2, 7) + XXH_rotl32(state->v3, 12) + XXH_rotl32(state->v4, 18);
+        h32 = XXH_rotl32(state->v1, 1)
+            + XXH_rotl32(state->v2, 7)
+            + XXH_rotl32(state->v3, 12)
+            + XXH_rotl32(state->v4, 18);
     } else {
         h32 = state->v3 /* == seed */ + PRIME32_5;
     }
 
     h32 += state->total_len_32;
 
-    while (p+4<=bEnd) {
-        h32 += XXH_readLE32(p, endian) * PRIME32_3;
-        h32  = XXH_rotl32(h32, 17) * PRIME32_4;
-        p+=4;
-    }
-
-    while (p<bEnd) {
-        h32 += (*p) * PRIME32_5;
-        h32  = XXH_rotl32(h32, 11) * PRIME32_1;
-        p++;
-    }
-
-    h32 ^= h32 >> 15;
-    h32 *= PRIME32_2;
-    h32 ^= h32 >> 13;
-    h32 *= PRIME32_3;
-    h32 ^= h32 >> 16;
-
-    return h32;
+    return XXH32_finalize(h32, state->mem32, state->memsize, endian, XXH_aligned);
 }
 
 
@@ -503,7 +559,7 @@ XXH_PUBLIC_API unsigned int XXH32_digest (const XXH32_state_t* state_in)
 /*! Default XXH result types are basic unsigned 32 and 64 bits.
 *   The canonical representation follows human-readable write convention, aka big-endian (large digits first).
 *   These functions allow transformation of hash result into and from its canonical format.
-*   This way, hash values can be written into a file or buffer, and remain comparable across different systems and programs.
+*   This way, hash values can be written into a file or buffer, remaining comparable across different systems.
 */
 
 XXH_PUBLIC_API void XXH32_canonicalFromHash(XXH32_canonical_t* dst, XXH32_hash_t hash)
@@ -522,18 +578,21 @@ XXH_PUBLIC_API XXH32_hash_t XXH32_hashFromCanonical(const XXH32_canonical_t* src
 #ifndef XXH_NO_LONG_LONG
 
 /* *******************************************************************
-*  64-bits hash functions
+*  64-bit hash functions
 *********************************************************************/
 
 /*======   Memory access   ======*/
 
 #ifndef MEM_MODULE
 # define MEM_MODULE
-# if !defined (__VMS) && (defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
+# if !defined (__VMS) \
+  && (defined (__cplusplus) \
+  || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */) )
 #   include <stdint.h>
     typedef uint64_t U64;
 # else
-    typedef unsigned long long U64;   /* if your compiler doesn't support unsigned long long, replace by another 64-bit type here. Note that xxhash.h will also need to be updated. */
+    /* if compiler doesn't support unsigned long long, replace by another 64-bit type */
+    typedef unsigned long long U64;
 # endif
 #endif
 
@@ -583,7 +642,7 @@ static U64 XXH_swap64 (U64 x)
 }
 #endif
 
-XXH_FORCE_INLINE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
+FORCE_INLINE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, XXH_alignment align)
 {
     if (align==XXH_unaligned)
         return endian==XXH_littleEndian ? XXH_read64(ptr) : XXH_swap64(XXH_read64(ptr));
@@ -591,7 +650,7 @@ XXH_FORCE_INLINE U64 XXH_readLE64_align(const void* ptr, XXH_endianess endian, X
         return endian==XXH_littleEndian ? *(const U64*)ptr : XXH_swap64(*(const U64*)ptr);
 }
 
-XXH_FORCE_INLINE U64 XXH_readLE64(const void* ptr, XXH_endianess endian)
+FORCE_INLINE U64 XXH_readLE64(const void* ptr, XXH_endianess endian)
 {
     return XXH_readLE64_align(ptr, endian, XXH_unaligned);
 }
@@ -626,14 +685,137 @@ static U64 XXH64_mergeRound(U64 acc, U64 val)
     return acc;
 }
 
-XXH_FORCE_INLINE U64 XXH64_endian_align(const void* input, size_t len, U64 seed, XXH_endianess endian, XXH_alignment align)
+static U64 XXH64_avalanche(U64 h64)
+{
+    h64 ^= h64 >> 33;
+    h64 *= PRIME64_2;
+    h64 ^= h64 >> 29;
+    h64 *= PRIME64_3;
+    h64 ^= h64 >> 32;
+    return h64;
+}
+
+
+#define XXH_get64bits(p) XXH_readLE64_align(p, endian, align)
+
+static U64
+XXH64_finalize(U64 h64, const void* ptr, size_t len,
+               XXH_endianess endian, XXH_alignment align)
+{
+    const BYTE* p = (const BYTE*)ptr;
+
+#define PROCESS1_64            \
+    h64 ^= (*p++) * PRIME64_5; \
+    h64 = XXH_rotl64(h64, 11) * PRIME64_1;
+
+#define PROCESS4_64          \
+    h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1; \
+    p+=4;                    \
+    h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
+
+#define PROCESS8_64 {        \
+    U64 const k1 = XXH64_round(0, XXH_get64bits(p)); \
+    p+=8;                    \
+    h64 ^= k1;               \
+    h64  = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4; \
+}
+
+    switch(len&31) {
+      case 24: PROCESS8_64;
+                    /* fallthrough */
+      case 16: PROCESS8_64;
+                    /* fallthrough */
+      case  8: PROCESS8_64;
+               return XXH64_avalanche(h64);
+
+      case 28: PROCESS8_64;
+                    /* fallthrough */
+      case 20: PROCESS8_64;
+                    /* fallthrough */
+      case 12: PROCESS8_64;
+                    /* fallthrough */
+      case  4: PROCESS4_64;
+               return XXH64_avalanche(h64);
+
+      case 25: PROCESS8_64;
+                    /* fallthrough */
+      case 17: PROCESS8_64;
+                    /* fallthrough */
+      case  9: PROCESS8_64;
+               PROCESS1_64;
+               return XXH64_avalanche(h64);
+
+      case 29: PROCESS8_64;
+                    /* fallthrough */
+      case 21: PROCESS8_64;
+                    /* fallthrough */
+      case 13: PROCESS8_64;
+                    /* fallthrough */
+      case  5: PROCESS4_64;
+               PROCESS1_64;
+               return XXH64_avalanche(h64);
+
+      case 26: PROCESS8_64;
+                    /* fallthrough */
+      case 18: PROCESS8_64;
+                    /* fallthrough */
+      case 10: PROCESS8_64;
+               PROCESS1_64;
+               PROCESS1_64;
+               return XXH64_avalanche(h64);
+
+      case 30: PROCESS8_64;
+                    /* fallthrough */
+      case 22: PROCESS8_64;
+                    /* fallthrough */
+      case 14: PROCESS8_64;
+                    /* fallthrough */
+      case  6: PROCESS4_64;
+               PROCESS1_64;
+               PROCESS1_64;
+               return XXH64_avalanche(h64);
+
+      case 27: PROCESS8_64;
+                    /* fallthrough */
+      case 19: PROCESS8_64;
+                    /* fallthrough */
+      case 11: PROCESS8_64;
+               PROCESS1_64;
+               PROCESS1_64;
+               PROCESS1_64;
+               return XXH64_avalanche(h64);
+
+      case 31: PROCESS8_64;
+                    /* fallthrough */
+      case 23: PROCESS8_64;
+                    /* fallthrough */
+      case 15: PROCESS8_64;
+                    /* fallthrough */
+      case  7: PROCESS4_64;
+                    /* fallthrough */
+      case  3: PROCESS1_64;
+                    /* fallthrough */
+      case  2: PROCESS1_64;
+                    /* fallthrough */
+      case  1: PROCESS1_64;
+                    /* fallthrough */
+      case  0: return XXH64_avalanche(h64);
+    }
+
+    /* impossible to reach */
+    assert(0);
+    return 0;  /* unreachable, but some compilers complain without it */
+}
+
+FORCE_INLINE U64
+XXH64_endian_align(const void* input, size_t len, U64 seed,
+                XXH_endianess endian, XXH_alignment align)
 {
     const BYTE* p = (const BYTE*)input;
-    const BYTE* const bEnd = p + len;
+    const BYTE* bEnd = p + len;
     U64 h64;
-#define XXH_get64bits(p) XXH_readLE64_align(p, endian, align)
 
-#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
+#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
     if (p==NULL) {
         len=0;
         bEnd=p=(const BYTE*)(size_t)32;
@@ -666,32 +848,7 @@ XXH_FORCE_INLINE U64 XXH64_endian_align(const void* input, size_t len, U64 seed,
 
     h64 += (U64) len;
 
-    while (p+8<=bEnd) {
-        U64 const k1 = XXH64_round(0, XXH_get64bits(p));
-        h64 ^= k1;
-        h64  = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
-        p+=8;
-    }
-
-    if (p+4<=bEnd) {
-        h64 ^= (U64)(XXH_get32bits(p)) * PRIME64_1;
-        h64 = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
-        p+=4;
-    }
-
-    while (p<bEnd) {
-        h64 ^= (*p) * PRIME64_5;
-        h64 = XXH_rotl64(h64, 11) * PRIME64_1;
-        p++;
-    }
-
-    h64 ^= h64 >> 33;
-    h64 *= PRIME64_2;
-    h64 ^= h64 >> 29;
-    h64 *= PRIME64_3;
-    h64 ^= h64 >> 32;
-
-    return h64;
+    return XXH64_finalize(h64, p, len, endian, align);
 }
 
 
@@ -741,65 +898,71 @@ XXH_PUBLIC_API void XXH64_copyState(XXH64_state_t* dstState, const XXH64_state_t
 XXH_PUBLIC_API XXH_errorcode XXH64_reset(XXH64_state_t* statePtr, unsigned long long seed)
 {
     XXH64_state_t state;   /* using a local state to memcpy() in order to avoid strict-aliasing warnings */
-    memset(&state, 0, sizeof(state)-8);   /* do not write into reserved, for future removal */
+    memset(&state, 0, sizeof(state));
     state.v1 = seed + PRIME64_1 + PRIME64_2;
     state.v2 = seed + PRIME64_2;
     state.v3 = seed + 0;
     state.v4 = seed - PRIME64_1;
-    memcpy(statePtr, &state, sizeof(state));
+     /* do not write into reserved, planned to be removed in a future version */
+    memcpy(statePtr, &state, sizeof(state) - sizeof(state.reserved));
     return XXH_OK;
 }
 
-XXH_FORCE_INLINE XXH_errorcode XXH64_update_endian (XXH64_state_t* state, const void* input, size_t len, XXH_endianess endian)
+FORCE_INLINE XXH_errorcode
+XXH64_update_endian (XXH64_state_t* state, const void* input, size_t len, XXH_endianess endian)
 {
-    const BYTE* p = (const BYTE*)input;
-    const BYTE* const bEnd = p + len;
-
-#ifdef XXH_ACCEPT_NULL_INPUT_POINTER
-    if (input==NULL) return XXH_ERROR;
+    if (input==NULL)
+#if defined(XXH_ACCEPT_NULL_INPUT_POINTER) && (XXH_ACCEPT_NULL_INPUT_POINTER>=1)
+        return XXH_OK;
+#else
+        return XXH_ERROR;
 #endif
 
-    state->total_len += len;
-
-    if (state->memsize + len < 32) {  /* fill in tmp buffer */
-        XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
-        state->memsize += (U32)len;
-        return XXH_OK;
-    }
+    {   const BYTE* p = (const BYTE*)input;
+        const BYTE* const bEnd = p + len;
 
-    if (state->memsize) {   /* tmp buffer is full */
-        XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize);
-        state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0, endian));
-        state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1, endian));
-        state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2, endian));
-        state->v4 = XXH64_round(state->v4, XXH_readLE64(state->mem64+3, endian));
-        p += 32-state->memsize;
-        state->memsize = 0;
-    }
+        state->total_len += len;
 
-    if (p+32 <= bEnd) {
-        const BYTE* const limit = bEnd - 32;
-        U64 v1 = state->v1;
-        U64 v2 = state->v2;
-        U64 v3 = state->v3;
-        U64 v4 = state->v4;
+        if (state->memsize + len < 32) {  /* fill in tmp buffer */
+            XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, len);
+            state->memsize += (U32)len;
+            return XXH_OK;
+        }
 
-        do {
-            v1 = XXH64_round(v1, XXH_readLE64(p, endian)); p+=8;
-            v2 = XXH64_round(v2, XXH_readLE64(p, endian)); p+=8;
-            v3 = XXH64_round(v3, XXH_readLE64(p, endian)); p+=8;
-            v4 = XXH64_round(v4, XXH_readLE64(p, endian)); p+=8;
-        } while (p<=limit);
+        if (state->memsize) {   /* tmp buffer is full */
+            XXH_memcpy(((BYTE*)state->mem64) + state->memsize, input, 32-state->memsize);
+            state->v1 = XXH64_round(state->v1, XXH_readLE64(state->mem64+0, endian));
+            state->v2 = XXH64_round(state->v2, XXH_readLE64(state->mem64+1, endian));
+            state->v3 = XXH64_round(state->v3, XXH_readLE64(state->mem64+2, endian));
+            state->v4 = XXH64_round(state->v4, XXH_readLE64(state->mem64+3, endian));
+            p += 32-state->memsize;
+            state->memsize = 0;
+        }
 
-        state->v1 = v1;
-        state->v2 = v2;
-        state->v3 = v3;
-        state->v4 = v4;
-    }
+        if (p+32 <= bEnd) {
+            const BYTE* const limit = bEnd - 32;
+            U64 v1 = state->v1;
+            U64 v2 = state->v2;
+            U64 v3 = state->v3;
+            U64 v4 = state->v4;
+
+            do {
+                v1 = XXH64_round(v1, XXH_readLE64(p, endian)); p+=8;
+                v2 = XXH64_round(v2, XXH_readLE64(p, endian)); p+=8;
+                v3 = XXH64_round(v3, XXH_readLE64(p, endian)); p+=8;
+                v4 = XXH64_round(v4, XXH_readLE64(p, endian)); p+=8;
+            } while (p<=limit);
+
+            state->v1 = v1;
+            state->v2 = v2;
+            state->v3 = v3;
+            state->v4 = v4;
+        }
 
-    if (p < bEnd) {
-        XXH_memcpy(state->mem64, p, (size_t)(bEnd-p));
-        state->memsize = (unsigned)(bEnd-p);
+        if (p < bEnd) {
+            XXH_memcpy(state->mem64, p, (size_t)(bEnd-p));
+            state->memsize = (unsigned)(bEnd-p);
+        }
     }
 
     return XXH_OK;
@@ -815,10 +978,8 @@ XXH_PUBLIC_API XXH_errorcode XXH64_update (XXH64_state_t* state_in, const void*
         return XXH64_update_endian(state_in, input, len, XXH_bigEndian);
 }
 
-XXH_FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endianess endian)
+FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endianess endian)
 {
-    const BYTE * p = (const BYTE*)state->mem64;
-    const BYTE* const bEnd = (const BYTE*)state->mem64 + state->memsize;
     U64 h64;
 
     if (state->total_len >= 32) {
@@ -833,37 +994,12 @@ XXH_FORCE_INLINE U64 XXH64_digest_endian (const XXH64_state_t* state, XXH_endian
         h64 = XXH64_mergeRound(h64, v3);
         h64 = XXH64_mergeRound(h64, v4);
     } else {
-        h64  = state->v3 + PRIME64_5;
+        h64  = state->v3 /*seed*/ + PRIME64_5;
     }
 
     h64 += (U64) state->total_len;
 
-    while (p+8<=bEnd) {
-        U64 const k1 = XXH64_round(0, XXH_readLE64(p, endian));
-        h64 ^= k1;
-        h64  = XXH_rotl64(h64,27) * PRIME64_1 + PRIME64_4;
-        p+=8;
-    }
-
-    if (p+4<=bEnd) {
-        h64 ^= (U64)(XXH_readLE32(p, endian)) * PRIME64_1;
-        h64  = XXH_rotl64(h64, 23) * PRIME64_2 + PRIME64_3;
-        p+=4;
-    }
-
-    while (p<bEnd) {
-        h64 ^= (*p) * PRIME64_5;
-        h64  = XXH_rotl64(h64, 11) * PRIME64_1;
-        p++;
-    }
-
-    h64 ^= h64 >> 33;
-    h64 *= PRIME64_2;
-    h64 ^= h64 >> 29;
-    h64 *= PRIME64_3;
-    h64 ^= h64 >> 32;
-
-    return h64;
+    return XXH64_finalize(h64, state->mem64, (size_t)state->total_len, endian, XXH_aligned);
 }
 
 XXH_PUBLIC_API unsigned long long XXH64_digest (const XXH64_state_t* state_in)