[Minor] Update bundled zstd

38 files changed, 8450 insertions, 4388 deletions

diff --git a/contrib/zstd/CMakeLists.txt b/contrib/zstd/CMakeLists.txt
index e3c283c61..2cccee0c1 100644
--- a/contrib/zstd/CMakeLists.txt
+++ b/contrib/zstd/CMakeLists.txt
@@ -1,20 +1,27 @@
 SET(ZSTDSRC
-			cover.c
-			divsufsort.c
-			entropy_common.c
-			error_private.c
-			fse_compress.c
-			fse_decompress.c
-			huf_compress.c
-			huf_decompress.c
-			pool.c
-			threading.c
-			zdict.c
-			zstd_common.c
-			zstd_compress.c
-			zstd_decompress.c)
+		cover.c
+		divsufsort.c
+		entropy_common.c
+		error_private.c
+		fse_compress.c
+		fse_decompress.c
+		huf_compress.c
+		huf_decompress.c
+		pool.c
+		threading.c
+		zdict.c
+		zstd_common.c
+		zstd_compress.c
+		zstd_decompress.c
+		zstd_double_fast.c
+		zstd_fast.c
+		zstd_lazy.c
+		zstd_ldm.c
+		zstdmt_compress.c
+		zstd_opt.c)
 
 ADD_LIBRARY(rspamd-zstd STATIC ${ZSTDSRC})
+ADD_DEFINITIONS(-DZSTD_STATIC_LINKING_ONLY)
 
 IF(ENABLE_FULL_DEBUG MATCHES "OFF")
 if ("${CMAKE_C_COMPILER_ID}" STREQUAL "Clang" OR "${CMAKE_C_COMPILER_ID}" STREQUAL "GNU")
diff --git a/contrib/zstd/bitstream.h b/contrib/zstd/bitstream.h
index 61f45328f..2094823fe 100644
--- a/contrib/zstd/bitstream.h
+++ b/contrib/zstd/bitstream.h
@@ -58,7 +58,9 @@ extern "C" {
 #if defined(BIT_DEBUG) && (BIT_DEBUG>=1)
 #  include <assert.h>
 #else
-#  define assert(condition) ((void)0)
+#  ifndef assert
+#    define assert(condition) ((void)0)
+#  endif
 #endif
 
 
@@ -73,13 +75,14 @@ extern "C" {
 #define STREAM_ACCUMULATOR_MIN_64  57
 #define STREAM_ACCUMULATOR_MIN    ((U32)(MEM_32bits() ? STREAM_ACCUMULATOR_MIN_32 : STREAM_ACCUMULATOR_MIN_64))
 
+
 /*-******************************************
 *  bitStream encoding API (write forward)
 ********************************************/
 /* bitStream can mix input from multiple sources.
-*  A critical property of these streams is that they encode and decode in **reverse** direction.
-*  So the first bit sequence you add will be the last to be read, like a LIFO stack.
-*/
+ * A critical property of these streams is that they encode and decode in **reverse** direction.
+ * So the first bit sequence you add will be the last to be read, like a LIFO stack.
+ */
 typedef struct
 {
     size_t bitContainer;
@@ -166,33 +169,39 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
 ****************************************************************/
 MEM_STATIC unsigned BIT_highbit32 (register U32 val)
 {
+    assert(val != 0);
+    {
 #   if defined(_MSC_VER)   /* Visual */
-    unsigned long r=0;
-    _BitScanReverse ( &r, val );
-    return (unsigned) r;
+        unsigned long r=0;
+        _BitScanReverse ( &r, val );
+        return (unsigned) r;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
-    return 31 - __builtin_clz (val);
+        return 31 - __builtin_clz (val);
 #   else   /* Software version */
-    static const unsigned DeBruijnClz[32] = { 0,  9,  1, 10, 13, 21,  2, 29,
-                                             11, 14, 16, 18, 22, 25,  3, 30,
-                                              8, 12, 20, 28, 15, 17, 24,  7,
-                                             19, 27, 23,  6, 26,  5,  4, 31 };
-    U32 v = val;
-    v |= v >> 1;
-    v |= v >> 2;
-    v |= v >> 4;
-    v |= v >> 8;
-    v |= v >> 16;
-    return DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27];
+        static const unsigned DeBruijnClz[32] = { 0,  9,  1, 10, 13, 21,  2, 29,
+                                                 11, 14, 16, 18, 22, 25,  3, 30,
+                                                  8, 12, 20, 28, 15, 17, 24,  7,
+                                                 19, 27, 23,  6, 26,  5,  4, 31 };
+        U32 v = val;
+        v |= v >> 1;
+        v |= v >> 2;
+        v |= v >> 4;
+        v |= v >> 8;
+        v |= v >> 16;
+        return DeBruijnClz[ (U32) (v * 0x07C4ACDDU) >> 27];
 #   endif
+    }
 }
 
 /*=====    Local Constants   =====*/
-static const unsigned BIT_mask[] = { 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F,
-                                    0xFF, 0x1FF, 0x3FF, 0x7FF, 0xFFF, 0x1FFF, 0x3FFF, 0x7FFF,
-                                    0xFFFF, 0x1FFFF, 0x3FFFF, 0x7FFFF, 0xFFFFF, 0x1FFFFF, 0x3FFFFF, 0x7FFFFF,
-                                    0xFFFFFF, 0x1FFFFFF, 0x3FFFFFF };   /* up to 26 bits */
-
+static const unsigned BIT_mask[] = {
+    0,          1,         3,         7,         0xF,       0x1F,
+    0x3F,       0x7F,      0xFF,      0x1FF,     0x3FF,     0x7FF,
+    0xFFF,      0x1FFF,    0x3FFF,    0x7FFF,    0xFFFF,    0x1FFFF,
+    0x3FFFF,    0x7FFFF,   0xFFFFF,   0x1FFFFF,  0x3FFFFF,  0x7FFFFF,
+    0xFFFFFF,   0x1FFFFFF, 0x3FFFFFF, 0x7FFFFFF, 0xFFFFFFF, 0x1FFFFFFF,
+    0x3FFFFFFF, 0x7FFFFFFF}; /* up to 31 bits */
+#define BIT_MASK_SIZE (sizeof(BIT_mask) / sizeof(BIT_mask[0]))
 
 /*-**************************************************************
 *  bitStream encoding
@@ -200,7 +209,7 @@ static const unsigned BIT_mask[] = { 0, 1, 3, 7, 0xF, 0x1F, 0x3F, 0x7F,
 /*! BIT_initCStream() :
  *  `dstCapacity` must be > sizeof(size_t)
  *  @return : 0 if success,
-              otherwise an error code (can be tested using ERR_isError() ) */
+ *            otherwise an error code (can be tested using ERR_isError()) */
 MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC,
                                   void* startPtr, size_t dstCapacity)
 {
@@ -214,11 +223,14 @@ MEM_STATIC size_t BIT_initCStream(BIT_CStream_t* bitC,
 }
 
 /*! BIT_addBits() :
-    can add up to 26 bits into `bitC`.
-    Does not check for register overflow ! */
+ *  can add up to 31 bits into `bitC`.
+ *  Note : does not check for register overflow ! */
 MEM_STATIC void BIT_addBits(BIT_CStream_t* bitC,
                             size_t value, unsigned nbBits)
 {
+    MEM_STATIC_ASSERT(BIT_MASK_SIZE == 32);
+    assert(nbBits < BIT_MASK_SIZE);
+    assert(nbBits + bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     bitC->bitContainer |= (value & BIT_mask[nbBits]) << bitC->bitPos;
     bitC->bitPos += nbBits;
 }
@@ -229,6 +241,7 @@ MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC,
                                 size_t value, unsigned nbBits)
 {
     assert((value>>nbBits) == 0);
+    assert(nbBits + bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     bitC->bitContainer |= value << bitC->bitPos;
     bitC->bitPos += nbBits;
 }
@@ -239,7 +252,7 @@ MEM_STATIC void BIT_addBitsFast(BIT_CStream_t* bitC,
 MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
-    assert( bitC->bitPos <= (sizeof(bitC->bitContainer)*8) );
+    assert(bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
     assert(bitC->ptr <= bitC->endPtr);
@@ -255,7 +268,7 @@ MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
 MEM_STATIC void BIT_flushBits(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
-    assert( bitC->bitPos <= (sizeof(bitC->bitContainer)*8) );
+    assert(bitC->bitPos < sizeof(bitC->bitContainer) * 8);
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
     if (bitC->ptr > bitC->endPtr) bitC->ptr = bitC->endPtr;
@@ -265,7 +278,7 @@ MEM_STATIC void BIT_flushBits(BIT_CStream_t* bitC)
 
 /*! BIT_closeCStream() :
  *  @return : size of CStream, in bytes,
-              or 0 if it could not fit into dstBuffer */
+ *            or 0 if it could not fit into dstBuffer */
 MEM_STATIC size_t BIT_closeCStream(BIT_CStream_t* bitC)
 {
     BIT_addBitsFast(bitC, 1, 1);   /* endMark */
@@ -276,14 +289,14 @@ MEM_STATIC size_t BIT_closeCStream(BIT_CStream_t* bitC)
 
 
 /*-********************************************************
-* bitStream decoding
+*  bitStream decoding
 **********************************************************/
 /*! BIT_initDStream() :
-*   Initialize a BIT_DStream_t.
-*   `bitD` : a pointer to an already allocated BIT_DStream_t structure.
-*   `srcSize` must be the *exact* size of the bitStream, in bytes.
-*   @return : size of stream (== srcSize) or an errorCode if a problem is detected
-*/
+ *  Initialize a BIT_DStream_t.
+ * `bitD` : a pointer to an already allocated BIT_DStream_t structure.
+ * `srcSize` must be the *exact* size of the bitStream, in bytes.
+ * @return : size of stream (== srcSize), or an errorCode if a problem is detected
+ */
 MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, size_t srcSize)
 {
     if (srcSize < 1) { memset(bitD, 0, sizeof(*bitD)); return ERROR(srcSize_wrong); }
@@ -302,29 +315,30 @@ MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, si
         bitD->bitContainer = *(const BYTE*)(bitD->start);
         switch(srcSize)
         {
-	    case 7: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[6]) << (sizeof(bitD->bitContainer)*8 - 16);
-	            /* fall-through */
+        case 7: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[6]) << (sizeof(bitD->bitContainer)*8 - 16);
+                /* fall-through */
 
-	    case 6: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[5]) << (sizeof(bitD->bitContainer)*8 - 24);
-	            /* fall-through */
+        case 6: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[5]) << (sizeof(bitD->bitContainer)*8 - 24);
+                /* fall-through */
 
-	    case 5: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[4]) << (sizeof(bitD->bitContainer)*8 - 32);
-	            /* fall-through */
+        case 5: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[4]) << (sizeof(bitD->bitContainer)*8 - 32);
+                /* fall-through */
 
-	    case 4: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[3]) << 24;
-	            /* fall-through */
+        case 4: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[3]) << 24;
+                /* fall-through */
 
-	    case 3: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[2]) << 16;
-	            /* fall-through */
+        case 3: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[2]) << 16;
+                /* fall-through */
 
-	    case 2: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[1]) <<  8;
-	            /* fall-through */
+        case 2: bitD->bitContainer += (size_t)(((const BYTE*)(srcBuffer))[1]) <<  8;
+                /* fall-through */
 
-            default: break;
+        default: break;
+        }
+        {   BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
+            bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;
+            if (lastByte == 0) return ERROR(corruption_detected);  /* endMark not present */
         }
-        { BYTE const lastByte = ((const BYTE*)srcBuffer)[srcSize-1];
-          bitD->bitsConsumed = lastByte ? 8 - BIT_highbit32(lastByte) : 0;
-          if (lastByte == 0) return ERROR(GENERIC); /* endMark not present */ }
         bitD->bitsConsumed += (U32)(sizeof(bitD->bitContainer) - srcSize)*8;
     }
 
@@ -346,12 +360,14 @@ MEM_STATIC size_t BIT_getMiddleBits(size_t bitContainer, U32 const start, U32 co
 #  endif
         return _bextr_u32(bitContainer, start, nbBits);
 #else
+    assert(nbBits < BIT_MASK_SIZE);
     return (bitContainer >> start) & BIT_mask[nbBits];
 #endif
 }
 
 MEM_STATIC size_t BIT_getLowerBits(size_t bitContainer, U32 const nbBits)
 {
+    assert(nbBits < BIT_MASK_SIZE);
     return bitContainer & BIT_mask[nbBits];
 }
 
@@ -360,9 +376,8 @@ MEM_STATIC size_t BIT_getLowerBits(size_t bitContainer, U32 const nbBits)
  *  local register is not modified.
  *  On 32-bits, maxNbBits==24.
  *  On 64-bits, maxNbBits==56.
- *  @return : value extracted
- */
- MEM_STATIC size_t BIT_lookBits(const BIT_DStream_t* bitD, U32 nbBits)
+ * @return : value extracted */
+MEM_STATIC size_t BIT_lookBits(const BIT_DStream_t* bitD, U32 nbBits)
 {
 #if defined(__BMI__) && defined(__GNUC__)   /* experimental; fails if bitD->bitsConsumed + nbBits > sizeof(bitD->bitContainer)*8 */
     return BIT_getMiddleBits(bitD->bitContainer, (sizeof(bitD->bitContainer)*8) - bitD->bitsConsumed - nbBits, nbBits);
@@ -389,8 +404,7 @@ MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
 /*! BIT_readBits() :
  *  Read (consume) next n bits from local register and update.
  *  Pay attention to not read more than nbBits contained into local register.
- *  @return : extracted value.
- */
+ * @return : extracted value. */
 MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     size_t const value = BIT_lookBits(bitD, nbBits);
@@ -399,7 +413,7 @@ MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 }
 
 /*! BIT_readBitsFast() :
-*   unsafe version; only works only if nbBits >= 1 */
+ *  unsafe version; only works only if nbBits >= 1 */
 MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
 {
     size_t const value = BIT_lookBitsFast(bitD, nbBits);
@@ -409,10 +423,10 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
 }
 
 /*! BIT_reloadDStream() :
-*   Refill `bitD` from buffer previously set in BIT_initDStream() .
-*   This function is safe, it guarantees it will not read beyond src buffer.
-*   @return : status of `BIT_DStream_t` internal register.
-              if status == BIT_DStream_unfinished, internal register is filled with >= (sizeof(bitD->bitContainer)*8 - 7) bits */
+ *  Refill `bitD` from buffer previously set in BIT_initDStream() .
+ *  This function is safe, it guarantees it will not read beyond src buffer.
+ * @return : status of `BIT_DStream_t` internal register.
+ *           when status == BIT_DStream_unfinished, internal register is filled with at least 25 or 57 bits */
 MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
 {
     if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* overflow detected, like end of stream */
@@ -443,8 +457,8 @@ MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
 }
 
 /*! BIT_endOfDStream() :
-*   @return Tells if DStream has exactly reached its end (all bits consumed).
-*/
+ * @return : 1 if DStream has _exactly_ reached its end (all bits consumed).
+ */
 MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* DStream)
 {
     return ((DStream->ptr == DStream->start) && (DStream->bitsConsumed == sizeof(DStream->bitContainer)*8));
diff --git a/contrib/zstd/compiler.h b/contrib/zstd/compiler.h
new file mode 100644
index 000000000..3a7553c38
--- /dev/null
+++ b/contrib/zstd/compiler.h
@@ -0,0 +1,86 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_COMPILER_H
+#define ZSTD_COMPILER_H
+
+/*-*******************************************************
+*  Compiler specifics
+*********************************************************/
+/* force inlining */
+#if defined (__GNUC__) || defined(__cplusplus) || defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#  define INLINE_KEYWORD inline
+#else
+#  define INLINE_KEYWORD
+#endif
+
+#if defined(__GNUC__)
+#  define FORCE_INLINE_ATTR __attribute__((always_inline))
+#elif defined(_MSC_VER)
+#  define FORCE_INLINE_ATTR __forceinline
+#else
+#  define FORCE_INLINE_ATTR
+#endif
+
+/**
+ * FORCE_INLINE_TEMPLATE is used to define C "templates", which take constant
+ * parameters. They must be inlined for the compiler to elimininate the constant
+ * branches.
+ */
+#define FORCE_INLINE_TEMPLATE static INLINE_KEYWORD FORCE_INLINE_ATTR
+/**
+ * HINT_INLINE is used to help the compiler generate better code. It is *not*
+ * used for "templates", so it can be tweaked based on the compilers
+ * performance.
+ *
+ * gcc-4.8 and gcc-4.9 have been shown to benefit from leaving off the
+ * always_inline attribute.
+ *
+ * clang up to 5.0.0 (trunk) benefit tremendously from the always_inline
+ * attribute.
+ */
+#if !defined(__clang__) && defined(__GNUC__) && __GNUC__ >= 4 && __GNUC_MINOR__ >= 8 && __GNUC__ < 5
+#  define HINT_INLINE static INLINE_KEYWORD
+#else
+#  define HINT_INLINE static INLINE_KEYWORD FORCE_INLINE_ATTR
+#endif
+
+/* force no inlining */
+#ifdef _MSC_VER
+#  define FORCE_NOINLINE static __declspec(noinline)
+#else
+#  ifdef __GNUC__
+#    define FORCE_NOINLINE static __attribute__((__noinline__))
+#  else
+#    define FORCE_NOINLINE static
+#  endif
+#endif
+
+/* prefetch */
+#if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_I86))  /* _mm_prefetch() is not defined outside of x86/x64 */
+#  include <mmintrin.h>   /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */
+#  define PREFETCH(ptr)   _mm_prefetch((const char*)ptr, _MM_HINT_T0)
+#elif defined(__GNUC__)
+#  define PREFETCH(ptr)   __builtin_prefetch(ptr, 0, 0)
+#else
+#  define PREFETCH(ptr)   /* disabled */
+#endif
+
+/* disable warnings */
+#ifdef _MSC_VER    /* Visual Studio */
+#  include <intrin.h>                    /* For Visual 2005 */
+#  pragma warning(disable : 4100)        /* disable: C4100: unreferenced formal parameter */
+#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
+#  pragma warning(disable : 4204)        /* disable: C4204: non-constant aggregate initializer */
+#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
+#  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
+#endif
+
+#endif /* ZSTD_COMPILER_H */
diff --git a/contrib/zstd/cover.c b/contrib/zstd/cover.c
index 0f9622898..f6500b3d8 100644
--- a/contrib/zstd/cover.c
+++ b/contrib/zstd/cover.c
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 /* *****************************************************************************
@@ -382,7 +383,7 @@ static void COVER_group(COVER_ctx_t *ctx, const void *group,
 typedef struct {
   U32 begin;
   U32 end;
-  double score;
+  U32 score;
 } COVER_segment_t;
 
 /**
@@ -398,7 +399,8 @@ typedef struct {
  */
 static COVER_segment_t COVER_selectSegment(const COVER_ctx_t *ctx, U32 *freqs,
                                            COVER_map_t *activeDmers, U32 begin,
-                                           U32 end, COVER_params_t parameters) {
+                                           U32 end,
+                                           ZDICT_cover_params_t parameters) {
   /* Constants */
   const U32 k = parameters.k;
   const U32 d = parameters.d;
@@ -437,7 +439,7 @@ static COVER_segment_t COVER_selectSegment(const COVER_ctx_t *ctx, U32 *freqs,
       U32 *delDmerOcc = COVER_map_at(activeDmers, delDmer);
       activeSegment.begin += 1;
       *delDmerOcc -= 1;
-      /* If this is the last occurrence of the dmer, subtract its score */
+      /* If this is the last occurence of the dmer, subtract its score */
       if (*delDmerOcc == 0) {
         COVER_map_remove(activeDmers, delDmer);
         activeSegment.score -= freqs[delDmer];
@@ -478,11 +480,16 @@ static COVER_segment_t COVER_selectSegment(const COVER_ctx_t *ctx, U32 *freqs,
  * Check the validity of the parameters.
  * Returns non-zero if the parameters are valid and 0 otherwise.
  */
-static int COVER_checkParameters(COVER_params_t parameters) {
+static int COVER_checkParameters(ZDICT_cover_params_t parameters,
+                                 size_t maxDictSize) {
   /* k and d are required parameters */
   if (parameters.d == 0 || parameters.k == 0) {
     return 0;
   }
+  /* k <= maxDictSize */
+  if (parameters.k > maxDictSize) {
+    return 0;
+  }
   /* d <= k */
   if (parameters.d > parameters.k) {
     return 0;
@@ -600,7 +607,7 @@ static int COVER_ctx_init(COVER_ctx_t *ctx, const void *samplesBuffer,
 static size_t COVER_buildDictionary(const COVER_ctx_t *ctx, U32 *freqs,
                                     COVER_map_t *activeDmers, void *dictBuffer,
                                     size_t dictBufferCapacity,
-                                    COVER_params_t parameters) {
+                                    ZDICT_cover_params_t parameters) {
   BYTE *const dict = (BYTE *)dictBuffer;
   size_t tail = dictBufferCapacity;
   /* Divide the data up into epochs of equal size.
@@ -621,9 +628,13 @@ static size_t COVER_buildDictionary(const COVER_ctx_t *ctx, U32 *freqs,
     /* Select a segment */
     COVER_segment_t segment = COVER_selectSegment(
         ctx, freqs, activeDmers, epochBegin, epochEnd, parameters);
-    /* Trim the segment if necessary and if it is empty then we are done */
+    /* If the segment covers no dmers, then we are out of content */
+    if (segment.score == 0) {
+      break;
+    }
+    /* Trim the segment if necessary and if it is too small then we are done */
     segmentSize = MIN(segment.end - segment.begin + parameters.d - 1, tail);
-    if (segmentSize == 0) {
+    if (segmentSize < parameters.d) {
       break;
     }
     /* We fill the dictionary from the back to allow the best segments to be
@@ -639,27 +650,15 @@ static size_t COVER_buildDictionary(const COVER_ctx_t *ctx, U32 *freqs,
   return tail;
 }
 
-/**
- * Translate from COVER_params_t to ZDICT_params_t required for finalizing the
- * dictionary.
- */
-static ZDICT_params_t COVER_translateParams(COVER_params_t parameters) {
-  ZDICT_params_t zdictParams;
-  memset(&zdictParams, 0, sizeof(zdictParams));
-  zdictParams.notificationLevel = 1;
-  zdictParams.dictID = parameters.dictID;
-  zdictParams.compressionLevel = parameters.compressionLevel;
-  return zdictParams;
-}
-
-ZDICTLIB_API size_t COVER_trainFromBuffer(
+ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
     void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
-    const size_t *samplesSizes, unsigned nbSamples, COVER_params_t parameters) {
+    const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t parameters) {
   BYTE *const dict = (BYTE *)dictBuffer;
   COVER_ctx_t ctx;
   COVER_map_t activeDmers;
   /* Checks */
-  if (!COVER_checkParameters(parameters)) {
+  if (!COVER_checkParameters(parameters, dictBufferCapacity)) {
     DISPLAYLEVEL(1, "Cover parameters incorrect\n");
     return ERROR(GENERIC);
   }
@@ -673,7 +672,7 @@ ZDICTLIB_API size_t COVER_trainFromBuffer(
     return ERROR(dstSize_tooSmall);
   }
   /* Initialize global data */
-  g_displayLevel = parameters.notificationLevel;
+  g_displayLevel = parameters.zParams.notificationLevel;
   /* Initialize context and activeDmers */
   if (!COVER_ctx_init(&ctx, samplesBuffer, samplesSizes, nbSamples,
                       parameters.d)) {
@@ -690,10 +689,9 @@ ZDICTLIB_API size_t COVER_trainFromBuffer(
     const size_t tail =
         COVER_buildDictionary(&ctx, ctx.freqs, &activeDmers, dictBuffer,
                               dictBufferCapacity, parameters);
-    ZDICT_params_t zdictParams = COVER_translateParams(parameters);
     const size_t dictionarySize = ZDICT_finalizeDictionary(
         dict, dictBufferCapacity, dict + tail, dictBufferCapacity - tail,
-        samplesBuffer, samplesSizes, nbSamples, zdictParams);
+        samplesBuffer, samplesSizes, nbSamples, parameters.zParams);
     if (!ZSTD_isError(dictionarySize)) {
       DISPLAYLEVEL(2, "Constructed dictionary of size %u\n",
                    (U32)dictionarySize);
@@ -718,7 +716,7 @@ typedef struct COVER_best_s {
   size_t liveJobs;
   void *dict;
   size_t dictSize;
-  COVER_params_t parameters;
+  ZDICT_cover_params_t parameters;
   size_t compressedSize;
 } COVER_best_t;
 
@@ -726,11 +724,9 @@ typedef struct COVER_best_s {
  * Initialize the `COVER_best_t`.
  */
 static void COVER_best_init(COVER_best_t *best) {
-  if (!best) {
-    return;
-  }
-  pthread_mutex_init(&best->mutex, NULL);
-  pthread_cond_init(&best->cond, NULL);
+  if (best==NULL) return; /* compatible with init on NULL */
+  (void)pthread_mutex_init(&best->mutex, NULL);
+  (void)pthread_cond_init(&best->cond, NULL);
   best->liveJobs = 0;
   best->dict = NULL;
   best->dictSize = 0;
@@ -786,7 +782,7 @@ static void COVER_best_start(COVER_best_t *best) {
  * If this dictionary is the best so far save it and its parameters.
  */
 static void COVER_best_finish(COVER_best_t *best, size_t compressedSize,
-                              COVER_params_t parameters, void *dict,
+                              ZDICT_cover_params_t parameters, void *dict,
                               size_t dictSize) {
   if (!best) {
     return;
@@ -830,11 +826,11 @@ typedef struct COVER_tryParameters_data_s {
   const COVER_ctx_t *ctx;
   COVER_best_t *best;
   size_t dictBufferCapacity;
-  COVER_params_t parameters;
+  ZDICT_cover_params_t parameters;
 } COVER_tryParameters_data_t;
 
 /**
- * Tries a set of parameters and updates the COVER_best_t with the results.
+ * Tries a set of parameters and upates the COVER_best_t with the results.
  * This function is thread safe if zstd is compiled with multithreaded support.
  * It takes its parameters as an *OWNING* opaque pointer to support threading.
  */
@@ -842,7 +838,7 @@ static void COVER_tryParameters(void *opaque) {
   /* Save parameters as local variables */
   COVER_tryParameters_data_t *const data = (COVER_tryParameters_data_t *)opaque;
   const COVER_ctx_t *const ctx = data->ctx;
-  const COVER_params_t parameters = data->parameters;
+  const ZDICT_cover_params_t parameters = data->parameters;
   size_t dictBufferCapacity = data->dictBufferCapacity;
   size_t totalCompressedSize = ERROR(GENERIC);
   /* Allocate space for hash table, dict, and freqs */
@@ -863,10 +859,10 @@ static void COVER_tryParameters(void *opaque) {
   {
     const size_t tail = COVER_buildDictionary(ctx, freqs, &activeDmers, dict,
                                               dictBufferCapacity, parameters);
-    const ZDICT_params_t zdictParams = COVER_translateParams(parameters);
     dictBufferCapacity = ZDICT_finalizeDictionary(
         dict, dictBufferCapacity, dict + tail, dictBufferCapacity - tail,
-        ctx->samples, ctx->samplesSizes, (unsigned)ctx->nbSamples, zdictParams);
+        ctx->samples, ctx->samplesSizes, (unsigned)ctx->nbSamples,
+        parameters.zParams);
     if (ZDICT_isError(dictBufferCapacity)) {
       DISPLAYLEVEL(1, "Failed to finalize dictionary\n");
       goto _cleanup;
@@ -892,13 +888,13 @@ static void COVER_tryParameters(void *opaque) {
     }
     /* Create the cctx and cdict */
     cctx = ZSTD_createCCtx();
-    cdict =
-        ZSTD_createCDict(dict, dictBufferCapacity, parameters.compressionLevel);
+    cdict = ZSTD_createCDict(dict, dictBufferCapacity,
+                             parameters.zParams.compressionLevel);
     if (!dst || !cctx || !cdict) {
       goto _compressCleanup;
     }
     /* Compress each sample and sum their sizes (or error) */
-    totalCompressedSize = 0;
+    totalCompressedSize = dictBufferCapacity;
     for (i = 0; i < ctx->nbSamples; ++i) {
       const size_t size = ZSTD_compress_usingCDict(
           cctx, dst, dstCapacity, ctx->samples + ctx->offsets[i],
@@ -930,12 +926,10 @@ _cleanup:
   }
 }
 
-ZDICTLIB_API size_t COVER_optimizeTrainFromBuffer(void *dictBuffer,
-                                                  size_t dictBufferCapacity,
-                                                  const void *samplesBuffer,
-                                                  const size_t *samplesSizes,
-                                                  unsigned nbSamples,
-                                                  COVER_params_t *parameters) {
+ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t *parameters) {
   /* constants */
   const unsigned nbThreads = parameters->nbThreads;
   const unsigned kMinD = parameters->d == 0 ? 6 : parameters->d;
@@ -947,7 +941,7 @@ ZDICTLIB_API size_t COVER_optimizeTrainFromBuffer(void *dictBuffer,
   const unsigned kIterations =
       (1 + (kMaxD - kMinD) / 2) * (1 + (kMaxK - kMinK) / kStepSize);
   /* Local variables */
-  const int displayLevel = parameters->notificationLevel;
+  const int displayLevel = parameters->zParams.notificationLevel;
   unsigned iteration = 1;
   unsigned d;
   unsigned k;
@@ -976,7 +970,7 @@ ZDICTLIB_API size_t COVER_optimizeTrainFromBuffer(void *dictBuffer,
   /* Initialization */
   COVER_best_init(&best);
   /* Turn down global display level to clean up display at level 2 and below */
-  g_displayLevel = parameters->notificationLevel - 1;
+  g_displayLevel = displayLevel == 0 ? 0 : displayLevel - 1;
   /* Loop through d first because each new value needs a new context */
   LOCALDISPLAYLEVEL(displayLevel, 2, "Trying %u different sets of parameters\n",
                     kIterations);
@@ -1010,8 +1004,9 @@ ZDICTLIB_API size_t COVER_optimizeTrainFromBuffer(void *dictBuffer,
       data->parameters.k = k;
       data->parameters.d = d;
       data->parameters.steps = kSteps;
+      data->parameters.zParams.notificationLevel = g_displayLevel;
       /* Check the parameters */
-      if (!COVER_checkParameters(data->parameters)) {
+      if (!COVER_checkParameters(data->parameters, dictBufferCapacity)) {
         DISPLAYLEVEL(1, "Cover parameters incorrect\n");
         free(data);
         continue;
diff --git a/contrib/zstd/error_private.c b/contrib/zstd/error_private.c
index 6bc86da7a..8045e445e 100644
--- a/contrib/zstd/error_private.c
+++ b/contrib/zstd/error_private.c
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 /* The purpose of this file is to have a single list of error strings embedded in binary */
@@ -20,18 +21,17 @@ const char* ERR_getErrorString(ERR_enum code)
     case PREFIX(GENERIC):  return "Error (generic)";
     case PREFIX(prefix_unknown): return "Unknown frame descriptor";
     case PREFIX(version_unsupported): return "Version not supported";
-    case PREFIX(parameter_unknown): return "Unknown parameter type";
     case PREFIX(frameParameter_unsupported): return "Unsupported frame parameter";
-    case PREFIX(frameParameter_unsupportedBy32bits): return "Frame parameter unsupported in 32-bits mode";
     case PREFIX(frameParameter_windowTooLarge): return "Frame requires too much memory for decoding";
-    case PREFIX(compressionParameter_unsupported): return "Compression parameter is out of bound";
+    case PREFIX(corruption_detected): return "Corrupted block detected";
+    case PREFIX(checksum_wrong): return "Restored data doesn't match checksum";
+    case PREFIX(parameter_unsupported): return "Unsupported parameter";
+    case PREFIX(parameter_outOfBound): return "Parameter is out of bound";
     case PREFIX(init_missing): return "Context should be init first";
     case PREFIX(memory_allocation): return "Allocation error : not enough memory";
     case PREFIX(stage_wrong): return "Operation not authorized at current processing stage";
     case PREFIX(dstSize_tooSmall): return "Destination buffer is too small";
     case PREFIX(srcSize_wrong): return "Src size is incorrect";
-    case PREFIX(corruption_detected): return "Corrupted block detected";
-    case PREFIX(checksum_wrong): return "Restored data doesn't match checksum";
     case PREFIX(tableLog_tooLarge): return "tableLog requires too much memory : unsupported";
     case PREFIX(maxSymbolValue_tooLarge): return "Unsupported max Symbol Value : too large";
     case PREFIX(maxSymbolValue_tooSmall): return "Specified maxSymbolValue is too small";
diff --git a/contrib/zstd/error_private.h b/contrib/zstd/error_private.h
index 1bc2e4954..0d2fa7e34 100644
--- a/contrib/zstd/error_private.h
+++ b/contrib/zstd/error_private.h
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 /* Note : this module is expected to remain private, do not expose it */
@@ -48,10 +49,9 @@ typedef ZSTD_ErrorCode ERR_enum;
 /*-****************************************
 *  Error codes handling
 ******************************************/
-#ifdef ERROR
-#  undef ERROR   /* reported already defined on VS 2015 (Rich Geldreich) */
-#endif
-#define ERROR(name) ((size_t)-PREFIX(name))
+#undef ERROR   /* reported already defined on VS 2015 (Rich Geldreich) */
+#define ERROR(name) ZSTD_ERROR(name)
+#define ZSTD_ERROR(name) ((size_t)-PREFIX(name))
 
 ERR_STATIC unsigned ERR_isError(size_t code) { return (code > ERROR(maxCode)); }
 
diff --git a/contrib/zstd/fse.h b/contrib/zstd/fse.h
index 6d5d41def..1c44f8375 100644
--- a/contrib/zstd/fse.h
+++ b/contrib/zstd/fse.h
@@ -31,13 +31,14 @@
    You can contact the author at :
    - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
 ****************************************************************** */
-#ifndef FSE_H
-#define FSE_H
 
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
+#ifndef FSE_H
+#define FSE_H
+
 
 /*-*****************************************
 *  Dependencies
@@ -297,8 +298,10 @@ FSE_decompress_usingDTable() result will tell how many bytes were regenerated (<
 If there is an error, the function will return an error code, which can be tested using FSE_isError(). (ex: dst buffer too small)
 */
 
+#endif  /* FSE_H */
 
-#ifdef FSE_STATIC_LINKING_ONLY
+#if defined(FSE_STATIC_LINKING_ONLY) && !defined(FSE_H_FSE_STATIC_LINKING_ONLY)
+#define FSE_H_FSE_STATIC_LINKING_ONLY
 
 /* *** Dependency *** */
 #include "bitstream.h"
@@ -381,6 +384,11 @@ size_t FSE_buildDTable_rle (FSE_DTable* dt, unsigned char symbolValue);
 size_t FSE_decompress_wksp(void* dst, size_t dstCapacity, const void* cSrc, size_t cSrcSize, FSE_DTable* workSpace, unsigned maxLog);
 /**< same as FSE_decompress(), using an externally allocated `workSpace` produced with `FSE_DTABLE_SIZE_U32(maxLog)` */
 
+typedef enum {
+   FSE_repeat_none,  /**< Cannot use the previous table */
+   FSE_repeat_check, /**< Can use the previous table but it must be checked */
+   FSE_repeat_valid  /**< Can use the previous table and it is asumed to be valid */
+ } FSE_repeat;
 
 /* *****************************************
 *  FSE symbol compression API
@@ -694,5 +702,3 @@ MEM_STATIC unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr)
 #if defined (__cplusplus)
 }
 #endif
-
-#endif  /* FSE_H */
diff --git a/contrib/zstd/fse_compress.c b/contrib/zstd/fse_compress.c
index 26e8052dd..599280b90 100644
--- a/contrib/zstd/fse_compress.c
+++ b/contrib/zstd/fse_compress.c
@@ -33,40 +33,22 @@
 ****************************************************************** */
 
 /* **************************************************************
-*  Compiler specifics
-****************************************************************/
-#ifdef _MSC_VER    /* Visual Studio */
-#  define FORCE_INLINE static __forceinline
-#  include <intrin.h>                    /* For Visual 2005 */
-#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
-#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
-#else
-#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
-#    ifdef __GNUC__
-#      define FORCE_INLINE static inline __attribute__((always_inline))
-#    else
-#      define FORCE_INLINE static inline
-#    endif
-#  else
-#    define FORCE_INLINE static
-#  endif /* __STDC_VERSION__ */
-#endif
-
-
-/* **************************************************************
 *  Includes
 ****************************************************************/
 #include <stdlib.h>     /* malloc, free, qsort */
 #include <string.h>     /* memcpy, memset */
 #include <stdio.h>      /* printf (debug) */
 #include "bitstream.h"
+#include "compiler.h"
 #define FSE_STATIC_LINKING_ONLY
 #include "fse.h"
+#include "error_private.h"
 
 
 /* **************************************************************
 *  Error Management
 ****************************************************************/
+#define FSE_isError ERR_isError
 #define FSE_STATIC_ASSERT(c) { enum { FSE_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
 
 
@@ -479,6 +461,7 @@ static unsigned FSE_minTableLog(size_t srcSize, unsigned maxSymbolValue)
     U32 minBitsSrc = BIT_highbit32((U32)(srcSize - 1)) + 1;
     U32 minBitsSymbols = BIT_highbit32(maxSymbolValue) + 2;
     U32 minBits = minBitsSrc < minBitsSymbols ? minBitsSrc : minBitsSymbols;
+    assert(srcSize > 1); /* Not supported, RLE should be used instead */
     return minBits;
 }
 
@@ -487,6 +470,7 @@ unsigned FSE_optimalTableLog_internal(unsigned maxTableLog, size_t srcSize, unsi
     U32 maxBitsSrc = BIT_highbit32((U32)(srcSize - 1)) - minus;
     U32 tableLog = maxTableLog;
     U32 minBits = FSE_minTableLog(srcSize, maxSymbolValue);
+    assert(srcSize > 1); /* Not supported, RLE should be used instead */
     if (tableLog==0) tableLog = FSE_DEFAULT_TABLELOG;
     if (maxBitsSrc < tableLog) tableLog = maxBitsSrc;   /* Accuracy can be reduced */
     if (minBits > tableLog) tableLog = minBits;   /* Need a minimum to safely represent all symbol values */
@@ -781,7 +765,7 @@ size_t FSE_compress_usingCTable (void* dst, size_t dstSize,
 
 size_t FSE_compressBound(size_t size) { return FSE_COMPRESSBOUND(size); }
 
-#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return f
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return e
 #define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
 
 /* FSE_compress_wksp() :
diff --git a/contrib/zstd/fse_decompress.c b/contrib/zstd/fse_decompress.c
index 8474a4c07..8e3f0035f 100644
--- a/contrib/zstd/fse_decompress.c
+++ b/contrib/zstd/fse_decompress.c
@@ -34,34 +34,15 @@
 
 
 /* **************************************************************
-*  Compiler specifics
-****************************************************************/
-#ifdef _MSC_VER    /* Visual Studio */
-#  define FORCE_INLINE static __forceinline
-#  include <intrin.h>                    /* For Visual 2005 */
-#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
-#  pragma warning(disable : 4214)        /* disable: C4214: non-int bitfields */
-#else
-#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
-#    ifdef __GNUC__
-#      define FORCE_INLINE static inline __attribute__((always_inline))
-#    else
-#      define FORCE_INLINE static inline
-#    endif
-#  else
-#    define FORCE_INLINE static
-#  endif /* __STDC_VERSION__ */
-#endif
-
-
-/* **************************************************************
 *  Includes
 ****************************************************************/
 #include <stdlib.h>     /* malloc, free, qsort */
 #include <string.h>     /* memcpy, memset */
 #include "bitstream.h"
+#include "compiler.h"
 #define FSE_STATIC_LINKING_ONLY
 #include "fse.h"
+#include "error_private.h"
 
 
 /* **************************************************************
@@ -216,7 +197,7 @@ size_t FSE_buildDTable_raw (FSE_DTable* dt, unsigned nbBits)
     return 0;
 }
 
-FORCE_INLINE size_t FSE_decompress_usingDTable_generic(
+FORCE_INLINE_TEMPLATE size_t FSE_decompress_usingDTable_generic(
           void* dst, size_t maxDstSize,
     const void* cSrc, size_t cSrcSize,
     const FSE_DTable* dt, const unsigned fast)
diff --git a/contrib/zstd/huf.h b/contrib/zstd/huf.h
index 7873ca3d4..2b3015a84 100644
--- a/contrib/zstd/huf.h
+++ b/contrib/zstd/huf.h
@@ -31,13 +31,13 @@
    You can contact the author at :
    - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
 ****************************************************************** */
-#ifndef HUF_H_298734234
-#define HUF_H_298734234
 
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
+#ifndef HUF_H_298734234
+#define HUF_H_298734234
 
 /* *** Dependencies *** */
 #include <stddef.h>    /* size_t */
@@ -111,7 +111,20 @@ HUF_PUBLIC_API size_t HUF_compress2 (void* dst, size_t dstCapacity, const void*
 #define HUF_WORKSPACE_SIZE_U32 (HUF_WORKSPACE_SIZE / sizeof(U32))
 HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
 
+/**
+ *  The minimum workspace size for the `workSpace` used in
+ *  HUF_readDTableX2_wksp() and HUF_readDTableX4_wksp().
+ *
+ *  The space used depends on HUF_TABLELOG_MAX, ranging from ~1500 bytes when
+ *  HUF_TABLE_LOG_MAX=12 to ~1850 bytes when HUF_TABLE_LOG_MAX=15.
+ *  Buffer overflow errors may potentially occur if code modifications result in
+ *  a required workspace size greater than that specified in the following
+ *  macro.
+ */
+#define HUF_DECOMPRESS_WORKSPACE_SIZE (2 << 10)
+#define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32))
 
+#endif   /* HUF_H_298734234 */
 
 /* ******************************************************************
  *  WARNING !!
@@ -120,7 +133,8 @@ HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity, const
  *  because they are not guaranteed to remain stable in the future.
  *  Only consider them in association with static linking.
  *******************************************************************/
-#ifdef HUF_STATIC_LINKING_ONLY
+#if defined(HUF_STATIC_LINKING_ONLY) && !defined(HUF_H_HUF_STATIC_LINKING_ONLY)
+#define HUF_H_HUF_STATIC_LINKING_ONLY
 
 /* *** Dependencies *** */
 #include "mem.h"   /* U32 */
@@ -170,8 +184,11 @@ size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cS
 
 size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< decodes RLE and uncompressed */
 size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize); /**< considers RLE and uncompressed as errors */
+size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize); /**< considers RLE and uncompressed as errors */
 size_t HUF_decompress4X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< single-symbol decoder */
 size_t HUF_decompress4X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< double-symbols decoder */
 
 
 /* ****************************************
@@ -243,7 +260,9 @@ HUF_decompress() does the following:
 U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
 
 size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX2_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
 size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize);
+size_t HUF_readDTableX4_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
 
 size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 size_t HUF_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
@@ -266,8 +285,11 @@ size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cS
 size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbol decoder */
 
 size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
+size_t HUF_decompress1X_DCtx_wksp (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);
 size_t HUF_decompress1X2_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< single-symbol decoder */
+size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< single-symbol decoder */
 size_t HUF_decompress1X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /**< double-symbols decoder */
+size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize);   /**< double-symbols decoder */
 
 size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);   /**< automatic selection of sing or double symbol decoder, based on DTable */
 size_t HUF_decompress1X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
@@ -275,9 +297,6 @@ size_t HUF_decompress1X4_usingDTable(void* dst, size_t maxDstSize, const void* c
 
 #endif /* HUF_STATIC_LINKING_ONLY */
 
-
 #if defined (__cplusplus)
 }
 #endif
-
-#endif   /* HUF_H_298734234 */
diff --git a/contrib/zstd/huf_compress.c b/contrib/zstd/huf_compress.c
index 7248c2513..2a47c1820 100644
--- a/contrib/zstd/huf_compress.c
+++ b/contrib/zstd/huf_compress.c
@@ -50,13 +50,15 @@
 #include "fse.h"        /* header compression */
 #define HUF_STATIC_LINKING_ONLY
 #include "huf.h"
+#include "error_private.h"
 
 
 /* **************************************************************
 *  Error Management
 ****************************************************************/
+#define HUF_isError ERR_isError
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
-#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return f
+#define CHECK_V_F(e, f) size_t const e = f; if (ERR_isError(e)) return e
 #define CHECK_F(f)   { CHECK_V_F(_var_err__, f); }
 
 
@@ -266,7 +268,8 @@ static U32 HUF_setMaxHeight(nodeElt* huffNode, U32 lastNonNull, U32 maxNbBits)
                         if (highTotal <= lowTotal) break;
                 }   }
                 /* only triggered when no more rank 1 symbol left => find closest one (note : there is necessarily at least one !) */
-                while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))  /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+                /* HUF_MAX_TABLELOG test just to please gcc 5+; but it should not be necessary */
+                while ((nBitsToDecrease<=HUF_TABLELOG_MAX) && (rankLast[nBitsToDecrease] == noSymbol))
                     nBitsToDecrease ++;
                 totalCost -= 1 << (nBitsToDecrease-1);
                 if (rankLast[nBitsToDecrease-1] == noSymbol)
@@ -435,7 +438,7 @@ static void HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt*
 
 size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
 
-#define HUF_FLUSHBITS(s)  (fast ? BIT_flushBitsFast(s) : BIT_flushBits(s))
+#define HUF_FLUSHBITS(s)  BIT_flushBits(s)
 
 #define HUF_FLUSHBITS_1(stream) \
     if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*2+7) HUF_FLUSHBITS(stream)
@@ -450,7 +453,6 @@ size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, si
     BYTE* const oend = ostart + dstSize;
     BYTE* op = ostart;
     size_t n;
-    const unsigned fast = (dstSize >= HUF_BLOCKBOUND(srcSize));
     BIT_CStream_t bitC;
 
     /* init */
diff --git a/contrib/zstd/huf_decompress.c b/contrib/zstd/huf_decompress.c
index ea35c3620..79ded96bf 100644
--- a/contrib/zstd/huf_decompress.c
+++ b/contrib/zstd/huf_decompress.c
@@ -33,40 +33,30 @@
 ****************************************************************** */
 
 /* **************************************************************
-*  Compiler specifics
-****************************************************************/
-#ifdef _MSC_VER    /* Visual Studio */
-#  define FORCE_INLINE static __forceinline
-#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
-#else
-#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
-#    ifdef __GNUC__
-#      define FORCE_INLINE static inline __attribute__((always_inline))
-#    else
-#      define FORCE_INLINE static inline
-#    endif
-#  else
-#    define FORCE_INLINE static
-#  endif /* __STDC_VERSION__ */
-#endif
-
-
-/* **************************************************************
 *  Dependencies
 ****************************************************************/
 #include <string.h>     /* memcpy, memset */
 #include "bitstream.h"  /* BIT_* */
+#include "compiler.h"
 #include "fse.h"        /* header compression */
 #define HUF_STATIC_LINKING_ONLY
 #include "huf.h"
+#include "error_private.h"
 
 
 /* **************************************************************
 *  Error Management
 ****************************************************************/
+#define HUF_isError ERR_isError
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
 
 
+/* **************************************************************
+*  Byte alignment for workSpace management
+****************************************************************/
+#define HUF_ALIGN(x, a) HUF_ALIGN_MASK((x), (a) - 1)
+#define HUF_ALIGN_MASK(x, mask) (((x) + (mask)) & ~(mask))
+
 /*-***************************/
 /*  generic DTableDesc       */
 /*-***************************/
@@ -87,16 +77,28 @@ static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
 
 typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX2;   /* single-symbol decoding */
 
-size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
+size_t HUF_readDTableX2_wksp(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize)
 {
-    BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
-    U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];   /* large enough for values from 0 to 16 */
     U32 tableLog = 0;
     U32 nbSymbols = 0;
     size_t iSize;
     void* const dtPtr = DTable + 1;
     HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
 
+    U32* rankVal;
+    BYTE* huffWeight;
+    size_t spaceUsed32 = 0;
+
+    rankVal = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_ABSOLUTEMAX + 1;
+    huffWeight = (BYTE *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
+
+    if ((spaceUsed32 << 2) > wkspSize)
+        return ERROR(tableLog_tooLarge);
+    workSpace = (U32 *)workSpace + spaceUsed32;
+    wkspSize -= (spaceUsed32 << 2);
+
     HUF_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
     /* memset(huffWeight, 0, sizeof(huffWeight)); */   /* is not necessary, even though some analyzer complain ... */
 
@@ -135,6 +137,13 @@ size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
     return iSize;
 }
 
+size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_readDTableX2_wksp(DTable, src, srcSize,
+                                 workSpace, sizeof(workSpace));
+}
+
 
 static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
 {
@@ -155,7 +164,7 @@ static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, con
     if (MEM_64bits()) \
         HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
 
-FORCE_INLINE size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
+HINT_INLINE size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
 {
     BYTE* const pStart = p;
 
@@ -212,11 +221,13 @@ size_t HUF_decompress1X2_usingDTable(
     return HUF_decompress1X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
 }
 
-size_t HUF_decompress1X2_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX2 (DCtx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX2_wksp(DCtx, cSrc, cSrcSize, workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -224,6 +235,15 @@ size_t HUF_decompress1X2_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, cons
     return HUF_decompress1X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
 }
 
+
+size_t HUF_decompress1X2_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X2_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
 size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
@@ -335,11 +355,14 @@ size_t HUF_decompress4X2_usingDTable(
 }
 
 
-size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX2 (dctx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX2_wksp (dctx, cSrc, cSrcSize,
+                                                workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -347,6 +370,13 @@ size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, cons
     return HUF_decompress4X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, dctx);
 }
 
+
+size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
 size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
@@ -403,7 +433,8 @@ static void HUF_fillDTableX4Level2(HUF_DEltX4* DTable, U32 sizeLog, const U32 co
     }   }
 }
 
-typedef U32 rankVal_t[HUF_TABLELOG_MAX][HUF_TABLELOG_MAX + 1];
+typedef U32 rankValCol_t[HUF_TABLELOG_MAX + 1];
+typedef rankValCol_t rankVal_t[HUF_TABLELOG_MAX];
 
 static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
                            const sortedSymbol_t* sortedList, const U32 sortedListSize,
@@ -447,20 +478,43 @@ static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
     }
 }
 
-size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize)
+size_t HUF_readDTableX4_wksp(HUF_DTable* DTable, const void* src,
+                             size_t srcSize, void* workSpace,
+                             size_t wkspSize)
 {
-    BYTE weightList[HUF_SYMBOLVALUE_MAX + 1];
-    sortedSymbol_t sortedSymbol[HUF_SYMBOLVALUE_MAX + 1];
-    U32 rankStats[HUF_TABLELOG_MAX + 1] = { 0 };
-    U32 rankStart0[HUF_TABLELOG_MAX + 2] = { 0 };
-    U32* const rankStart = rankStart0+1;
-    rankVal_t rankVal;
     U32 tableLog, maxW, sizeOfSort, nbSymbols;
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     U32 const maxTableLog = dtd.maxTableLog;
     size_t iSize;
     void* dtPtr = DTable+1;   /* force compiler to avoid strict-aliasing */
     HUF_DEltX4* const dt = (HUF_DEltX4*)dtPtr;
+    U32 *rankStart;
+
+    rankValCol_t* rankVal;
+    U32* rankStats;
+    U32* rankStart0;
+    sortedSymbol_t* sortedSymbol;
+    BYTE* weightList;
+    size_t spaceUsed32 = 0;
+
+    rankVal = (rankValCol_t *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += (sizeof(rankValCol_t) * HUF_TABLELOG_MAX) >> 2;
+    rankStats = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_MAX + 1;
+    rankStart0 = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_MAX + 2;
+    sortedSymbol = (sortedSymbol_t *)workSpace + (spaceUsed32 * sizeof(U32)) / sizeof(sortedSymbol_t);
+    spaceUsed32 += HUF_ALIGN(sizeof(sortedSymbol_t) * (HUF_SYMBOLVALUE_MAX + 1), sizeof(U32)) >> 2;
+    weightList = (BYTE *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
+
+    if ((spaceUsed32 << 2) > wkspSize)
+        return ERROR(tableLog_tooLarge);
+    workSpace = (U32 *)workSpace + spaceUsed32;
+    wkspSize -= (spaceUsed32 << 2);
+
+    rankStart = rankStart0 + 1;
+    memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
 
     HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(HUF_DTable));   /* if compiler fails here, assertion is wrong */
     if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
@@ -527,6 +581,12 @@ size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize)
     return iSize;
 }
 
+size_t HUF_readDTableX4(HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+  U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+  return HUF_readDTableX4_wksp(DTable, src, srcSize,
+                               workSpace, sizeof(workSpace));
+}
 
 static U32 HUF_decodeSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
 {
@@ -545,7 +605,8 @@ static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DE
         if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
             BIT_skipBits(DStream, dt[val].nbBits);
             if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
-                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);   /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+                /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);
     }   }
     return 1;
 }
@@ -562,7 +623,7 @@ static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DE
     if (MEM_64bits()) \
         ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
 
-FORCE_INLINE size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
+HINT_INLINE size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
 {
     BYTE* const pStart = p;
 
@@ -626,11 +687,14 @@ size_t HUF_decompress1X4_usingDTable(
     return HUF_decompress1X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
 }
 
-size_t HUF_decompress1X4_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX4 (DCtx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX4_wksp(DCtx, cSrc, cSrcSize,
+                                               workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -638,6 +702,15 @@ size_t HUF_decompress1X4_DCtx (HUF_DTable* DCtx, void* dst, size_t dstSize, cons
     return HUF_decompress1X4_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
 }
 
+
+size_t HUF_decompress1X4_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X4_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
 size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_TABLELOG_MAX);
@@ -748,11 +821,14 @@ size_t HUF_decompress4X4_usingDTable(
 }
 
 
-size_t HUF_decompress4X4_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t hSize = HUF_readDTableX4 (dctx, cSrc, cSrcSize);
+    size_t hSize = HUF_readDTableX4_wksp(dctx, cSrc, cSrcSize,
+                                         workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
@@ -760,6 +836,15 @@ size_t HUF_decompress4X4_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, cons
     return HUF_decompress4X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx);
 }
 
+
+size_t HUF_decompress4X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X4_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
 size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX4(DTable, HUF_TABLELOG_MAX);
@@ -816,11 +901,11 @@ static const algo_time_t algoTime[16 /* Quantization */][3 /* single, double, qu
 *   Tells which decoder is likely to decode faster,
 *   based on a set of pre-determined metrics.
 *   @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
-*   Assumption : 0 < cSrcSize < dstSize <= 128 KB */
+*   Assumption : 0 < cSrcSize, dstSize <= 128 KB */
 U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
 {
     /* decoder timing evaluation */
-    U32 const Q = (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 since dstSize > cSrcSize */
+    U32 const Q = cSrcSize >= dstSize ? 15 : (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 */
     U32 const D256 = (U32)(dstSize >> 8);
     U32 const DTime0 = algoTime[Q][0].tableTime + (algoTime[Q][0].decode256Time * D256);
     U32 DTime1 = algoTime[Q][1].tableTime + (algoTime[Q][1].decode256Time * D256);
@@ -861,19 +946,32 @@ size_t HUF_decompress4X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const
     }
 }
 
-size_t HUF_decompress4X_hufOnly (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X_hufOnly_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                         workSpace, sizeof(workSpace));
+}
+
+
+size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst,
+                                     size_t dstSize, const void* cSrc,
+                                     size_t cSrcSize, void* workSpace,
+                                     size_t wkspSize)
 {
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
-    if ((cSrcSize >= dstSize) || (cSrcSize <= 1)) return ERROR(corruption_detected);   /* invalid */
+    if (cSrcSize == 0) return ERROR(corruption_detected);
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress4X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
-                        HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+        return algoNb ? HUF_decompress4X4_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize):
+                        HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize);
     }
 }
 
-size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                  const void* cSrc, size_t cSrcSize,
+                                  void* workSpace, size_t wkspSize)
 {
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
@@ -882,7 +980,17 @@ size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const
     if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress1X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
-                        HUF_decompress1X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+        return algoNb ? HUF_decompress1X4_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize):
+                        HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize);
     }
 }
+
+size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
+                             const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                      workSpace, sizeof(workSpace));
+}
diff --git a/contrib/zstd/mem.h b/contrib/zstd/mem.h
index 4773a8b93..23335c314 100644
--- a/contrib/zstd/mem.h
+++ b/contrib/zstd/mem.h
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 #ifndef MEM_H_MODULE
@@ -110,7 +111,7 @@ Only use if no other choice to achieve best performance on target platform */
 MEM_STATIC U16 MEM_read16(const void* memPtr) { return *(const U16*) memPtr; }
 MEM_STATIC U32 MEM_read32(const void* memPtr) { return *(const U32*) memPtr; }
 MEM_STATIC U64 MEM_read64(const void* memPtr) { return *(const U64*) memPtr; }
-MEM_STATIC U64 MEM_readST(const void* memPtr) { return *(const size_t*) memPtr; }
+MEM_STATIC size_t MEM_readST(const void* memPtr) { return *(const size_t*) memPtr; }
 
 MEM_STATIC void MEM_write16(void* memPtr, U16 value) { *(U16*)memPtr = value; }
 MEM_STATIC void MEM_write32(void* memPtr, U32 value) { *(U32*)memPtr = value; }
@@ -131,7 +132,7 @@ MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(U64*)memPtr = value; }
 MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
 MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
 MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
-MEM_STATIC U64 MEM_readST(const void* ptr) { return ((const unalign*)ptr)->st; }
+MEM_STATIC size_t MEM_readST(const void* ptr) { return ((const unalign*)ptr)->st; }
 
 MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
 MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; }
@@ -352,20 +353,6 @@ MEM_STATIC void MEM_writeBEST(void* memPtr, size_t val)
 }
 
 
-/* function safe only for comparisons */
-MEM_STATIC U32 MEM_readMINMATCH(const void* memPtr, U32 length)
-{
-    switch (length)
-    {
-    default :
-    case 4 : return MEM_read32(memPtr);
-    case 3 : if (MEM_isLittleEndian())
-                return MEM_read32(memPtr)<<8;
-             else
-                return MEM_read32(memPtr)>>8;
-    }
-}
-
 #if defined (__cplusplus)
 }
 #endif
diff --git a/contrib/zstd/pool.c b/contrib/zstd/pool.c
index e439fe1b0..5f19f331b 100644
--- a/contrib/zstd/pool.c
+++ b/contrib/zstd/pool.c
@@ -1,10 +1,11 @@
-/**
- * Copyright (c) 2016-present, Facebook, Inc.
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 
@@ -25,11 +26,12 @@
 
 /* A job is a function and an opaque argument */
 typedef struct POOL_job_s {
-  POOL_function function;
-  void *opaque;
+    POOL_function function;
+    void *opaque;
 } POOL_job;
 
 struct POOL_ctx_s {
+    ZSTD_customMem customMem;
     /* Keep track of the threads */
     pthread_t *threads;
     size_t numThreads;
@@ -39,6 +41,12 @@ struct POOL_ctx_s {
     size_t queueHead;
     size_t queueTail;
     size_t queueSize;
+
+    /* The number of threads working on jobs */
+    size_t numThreadsBusy;
+    /* Indicates if the queue is empty */
+    int queueEmpty;
+
     /* The mutex protects the queue */
     pthread_mutex_t queueMutex;
     /* Condition variable for pushers to wait on when the queue is full */
@@ -60,48 +68,66 @@ static void* POOL_thread(void* opaque) {
     for (;;) {
         /* Lock the mutex and wait for a non-empty queue or until shutdown */
         pthread_mutex_lock(&ctx->queueMutex);
-        while (ctx->queueHead == ctx->queueTail && !ctx->shutdown) {
+
+        while (ctx->queueEmpty && !ctx->shutdown) {
             pthread_cond_wait(&ctx->queuePopCond, &ctx->queueMutex);
         }
         /* empty => shutting down: so stop */
-        if (ctx->queueHead == ctx->queueTail) {
+        if (ctx->queueEmpty) {
             pthread_mutex_unlock(&ctx->queueMutex);
             return opaque;
         }
         /* Pop a job off the queue */
         {   POOL_job const job = ctx->queue[ctx->queueHead];
             ctx->queueHead = (ctx->queueHead + 1) % ctx->queueSize;
+            ctx->numThreadsBusy++;
+            ctx->queueEmpty = ctx->queueHead == ctx->queueTail;
             /* Unlock the mutex, signal a pusher, and run the job */
             pthread_mutex_unlock(&ctx->queueMutex);
             pthread_cond_signal(&ctx->queuePushCond);
+
             job.function(job.opaque);
-        }
-    }
+
+            /* If the intended queue size was 0, signal after finishing job */
+            if (ctx->queueSize == 1) {
+                pthread_mutex_lock(&ctx->queueMutex);
+                ctx->numThreadsBusy--;
+                pthread_mutex_unlock(&ctx->queueMutex);
+                pthread_cond_signal(&ctx->queuePushCond);
+        }   }
+    }  /* for (;;) */
     /* Unreachable */
 }
 
 POOL_ctx *POOL_create(size_t numThreads, size_t queueSize) {
+    return POOL_create_advanced(numThreads, queueSize, ZSTD_defaultCMem);
+}
+
+POOL_ctx *POOL_create_advanced(size_t numThreads, size_t queueSize, ZSTD_customMem customMem) {
     POOL_ctx *ctx;
     /* Check the parameters */
-    if (!numThreads || !queueSize) { return NULL; }
+    if (!numThreads) { return NULL; }
     /* Allocate the context and zero initialize */
-    ctx = (POOL_ctx *)calloc(1, sizeof(POOL_ctx));
+    ctx = (POOL_ctx *)ZSTD_calloc(sizeof(POOL_ctx), customMem);
     if (!ctx) { return NULL; }
     /* Initialize the job queue.
      * It needs one extra space since one space is wasted to differentiate empty
      * and full queues.
      */
     ctx->queueSize = queueSize + 1;
-    ctx->queue = (POOL_job *)malloc(ctx->queueSize * sizeof(POOL_job));
+    ctx->queue = (POOL_job*) malloc(ctx->queueSize * sizeof(POOL_job));
     ctx->queueHead = 0;
     ctx->queueTail = 0;
-    pthread_mutex_init(&ctx->queueMutex, NULL);
-    pthread_cond_init(&ctx->queuePushCond, NULL);
-    pthread_cond_init(&ctx->queuePopCond, NULL);
+    ctx->numThreadsBusy = 0;
+    ctx->queueEmpty = 1;
+    (void)pthread_mutex_init(&ctx->queueMutex, NULL);
+    (void)pthread_cond_init(&ctx->queuePushCond, NULL);
+    (void)pthread_cond_init(&ctx->queuePopCond, NULL);
     ctx->shutdown = 0;
     /* Allocate space for the thread handles */
-    ctx->threads = (pthread_t *)malloc(numThreads * sizeof(pthread_t));
+    ctx->threads = (pthread_t*)ZSTD_malloc(numThreads * sizeof(pthread_t), customMem);
     ctx->numThreads = 0;
+    ctx->customMem = customMem;
     /* Check for errors */
     if (!ctx->threads || !ctx->queue) { POOL_free(ctx); return NULL; }
     /* Initialize the threads */
@@ -141,27 +167,49 @@ void POOL_free(POOL_ctx *ctx) {
     pthread_mutex_destroy(&ctx->queueMutex);
     pthread_cond_destroy(&ctx->queuePushCond);
     pthread_cond_destroy(&ctx->queuePopCond);
-    if (ctx->queue) free(ctx->queue);
-    if (ctx->threads) free(ctx->threads);
-    free(ctx);
+    ZSTD_free(ctx->queue, ctx->customMem);
+    ZSTD_free(ctx->threads, ctx->customMem);
+    ZSTD_free(ctx, ctx->customMem);
+}
+
+size_t POOL_sizeof(POOL_ctx *ctx) {
+    if (ctx==NULL) return 0;  /* supports sizeof NULL */
+    return sizeof(*ctx)
+        + ctx->queueSize * sizeof(POOL_job)
+        + ctx->numThreads * sizeof(pthread_t);
 }
 
-void POOL_add(void *ctxVoid, POOL_function function, void *opaque) {
-    POOL_ctx *ctx = (POOL_ctx *)ctxVoid;
+/**
+ * Returns 1 if the queue is full and 0 otherwise.
+ *
+ * If the queueSize is 1 (the pool was created with an intended queueSize of 0),
+ * then a queue is empty if there is a thread free and no job is waiting.
+ */
+static int isQueueFull(POOL_ctx const* ctx) {
+    if (ctx->queueSize > 1) {
+        return ctx->queueHead == ((ctx->queueTail + 1) % ctx->queueSize);
+    } else {
+        return ctx->numThreadsBusy == ctx->numThreads ||
+               !ctx->queueEmpty;
+    }
+}
+
+void POOL_add(void* ctxVoid, POOL_function function, void *opaque) {
+    POOL_ctx* const ctx = (POOL_ctx*)ctxVoid;
     if (!ctx) { return; }
 
     pthread_mutex_lock(&ctx->queueMutex);
     {   POOL_job const job = {function, opaque};
+
         /* Wait until there is space in the queue for the new job */
-        size_t newTail = (ctx->queueTail + 1) % ctx->queueSize;
-        while (ctx->queueHead == newTail && !ctx->shutdown) {
+        while (isQueueFull(ctx) && !ctx->shutdown) {
           pthread_cond_wait(&ctx->queuePushCond, &ctx->queueMutex);
-          newTail = (ctx->queueTail + 1) % ctx->queueSize;
         }
         /* The queue is still going => there is space */
         if (!ctx->shutdown) {
+            ctx->queueEmpty = 0;
             ctx->queue[ctx->queueTail] = job;
-            ctx->queueTail = newTail;
+            ctx->queueTail = (ctx->queueTail + 1) % ctx->queueSize;
         }
     }
     pthread_mutex_unlock(&ctx->queueMutex);
@@ -173,22 +221,35 @@ void POOL_add(void *ctxVoid, POOL_function function, void *opaque) {
 
 /* We don't need any data, but if it is empty malloc() might return NULL. */
 struct POOL_ctx_s {
-  int data;
+    int dummy;
 };
+static POOL_ctx g_ctx;
 
-POOL_ctx *POOL_create(size_t numThreads, size_t queueSize) {
-  (void)numThreads;
-  (void)queueSize;
-  return (POOL_ctx *)malloc(sizeof(POOL_ctx));
+POOL_ctx* POOL_create(size_t numThreads, size_t queueSize) {
+    return POOL_create_advanced(numThreads, queueSize, ZSTD_defaultCMem);
 }
 
-void POOL_free(POOL_ctx *ctx) {
-  if (ctx) free(ctx);
+POOL_ctx* POOL_create_advanced(size_t numThreads, size_t queueSize, ZSTD_customMem customMem) {
+    (void)numThreads;
+    (void)queueSize;
+    (void)customMem;
+    return &g_ctx;
+}
+
+void POOL_free(POOL_ctx* ctx) {
+    assert(!ctx || ctx == &g_ctx);
+    (void)ctx;
+}
+
+void POOL_add(void* ctx, POOL_function function, void* opaque) {
+    (void)ctx;
+    function(opaque);
 }
 
-void POOL_add(void *ctx, POOL_function function, void *opaque) {
-  (void)ctx;
-  function(opaque);
+size_t POOL_sizeof(POOL_ctx* ctx) {
+    if (ctx==NULL) return 0;  /* supports sizeof NULL */
+    assert(ctx == &g_ctx);
+    return sizeof(*ctx);
 }
 
 #endif  /* ZSTD_MULTITHREAD */
diff --git a/contrib/zstd/pool.h b/contrib/zstd/pool.h
index 50cb25b12..08c63715a 100644
--- a/contrib/zstd/pool.h
+++ b/contrib/zstd/pool.h
@@ -1,11 +1,13 @@
-/**
- * Copyright (c) 2016-present, Facebook, Inc.
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
+
 #ifndef POOL_H
 #define POOL_H
 
@@ -15,23 +17,30 @@ extern "C" {
 
 
 #include <stddef.h>   /* size_t */
+#include "zstd_internal.h"   /* ZSTD_customMem */
 
 typedef struct POOL_ctx_s POOL_ctx;
 
 /*! POOL_create() :
-    Create a thread pool with at most `numThreads` threads.
-    `numThreads` must be at least 1.
-    The maximum number of queued jobs before blocking is `queueSize`.
-    `queueSize` must be at least 1.
-    @return : The POOL_ctx pointer on success else NULL.
+ *  Create a thread pool with at most `numThreads` threads.
+ * `numThreads` must be at least 1.
+ *  The maximum number of queued jobs before blocking is `queueSize`.
+ * @return : POOL_ctx pointer on success, else NULL.
 */
 POOL_ctx *POOL_create(size_t numThreads, size_t queueSize);
 
+POOL_ctx *POOL_create_advanced(size_t numThreads, size_t queueSize, ZSTD_customMem customMem);
+
 /*! POOL_free() :
     Free a thread pool returned by POOL_create().
 */
 void POOL_free(POOL_ctx *ctx);
 
+/*! POOL_sizeof() :
+    return memory usage of pool returned by POOL_create().
+*/
+size_t POOL_sizeof(POOL_ctx *ctx);
+
 /*! POOL_function :
     The function type that can be added to a thread pool.
 */
diff --git a/contrib/zstd/threading.c b/contrib/zstd/threading.c
index 141376c56..a82c975b2 100644
--- a/contrib/zstd/threading.c
+++ b/contrib/zstd/threading.c
@@ -2,9 +2,9 @@
  * Copyright (c) 2016 Tino Reichardt
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
  *
  * You can contact the author at:
  * - zstdmt source repository: https://github.com/mcmilk/zstdmt
@@ -14,12 +14,8 @@
  * This file will hold wrapper for systems, which do not support pthreads
  */
 
-/* When ZSTD_MULTITHREAD is not defined, this file would become an empty translation unit.
-* Include some ISO C header code to prevent this and portably avoid related warnings.
-* (Visual C++: C4206 / GCC: -Wpedantic / Clang: -Wempty-translation-unit)
-*/
-#include <stddef.h>
-
+/* create fake symbol to avoid empty trnaslation unit warning */
+int g_ZSTD_threading_useles_symbol;
 
 #if defined(ZSTD_MULTITHREAD) && defined(_WIN32)
 
diff --git a/contrib/zstd/threading.h b/contrib/zstd/threading.h
index c0086139e..8194bc6fa 100644
--- a/contrib/zstd/threading.h
+++ b/contrib/zstd/threading.h
@@ -1,11 +1,10 @@
-
 /**
  * Copyright (c) 2016 Tino Reichardt
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
  *
  * You can contact the author at:
  * - zstdmt source repository: https://github.com/mcmilk/zstdmt
@@ -38,18 +37,22 @@ extern "C" {
 #  define WIN32_LEAN_AND_MEAN
 #endif
 
+#undef ERROR   /* reported already defined on VS 2015 (Rich Geldreich) */
 #include <windows.h>
+#undef ERROR
+#define ERROR(name) ZSTD_ERROR(name)
+
 
 /* mutex */
 #define pthread_mutex_t           CRITICAL_SECTION
-#define pthread_mutex_init(a,b)   InitializeCriticalSection((a))
+#define pthread_mutex_init(a,b)   (InitializeCriticalSection((a)), 0)
 #define pthread_mutex_destroy(a)  DeleteCriticalSection((a))
 #define pthread_mutex_lock(a)     EnterCriticalSection((a))
 #define pthread_mutex_unlock(a)   LeaveCriticalSection((a))
 
 /* condition variable */
 #define pthread_cond_t             CONDITION_VARIABLE
-#define pthread_cond_init(a, b)    InitializeConditionVariable((a))
+#define pthread_cond_init(a, b)    (InitializeConditionVariable((a)), 0)
 #define pthread_cond_destroy(a)    /* No delete */
 #define pthread_cond_wait(a, b)    SleepConditionVariableCS((a), (b), INFINITE)
 #define pthread_cond_signal(a)     WakeConditionVariable((a))
@@ -80,14 +83,14 @@ int _pthread_join(pthread_t* thread, void** value_ptr);
 #else  /* ZSTD_MULTITHREAD not defined */
 /* No multithreading support */
 
-#define pthread_mutex_t int   /* #define rather than typedef, as sometimes pthread support is implicit, resulting in duplicated symbols */
-#define pthread_mutex_init(a,b)
+#define pthread_mutex_t int   /* #define rather than typedef, because sometimes pthread support is implicit, resulting in duplicated symbols */
+#define pthread_mutex_init(a,b)    ((void)a, 0)
 #define pthread_mutex_destroy(a)
 #define pthread_mutex_lock(a)
 #define pthread_mutex_unlock(a)
 
 #define pthread_cond_t int
-#define pthread_cond_init(a,b)
+#define pthread_cond_init(a,b)     ((void)a, 0)
 #define pthread_cond_destroy(a)
 #define pthread_cond_wait(a,b)
 #define pthread_cond_signal(a)
diff --git a/contrib/zstd/zdict.c b/contrib/zstd/zdict.c
index 179e02eff..1bb8b0683 100644
--- a/contrib/zstd/zdict.c
+++ b/contrib/zstd/zdict.c
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 
@@ -60,7 +61,7 @@
 
 #define NOISELENGTH 32
 
-static const int g_compressionLevel_default = 6;
+static const int g_compressionLevel_default = 3;
 static const U32 g_selectivity_default = 9;
 
 
@@ -94,7 +95,7 @@ const char* ZDICT_getErrorName(size_t errorCode) { return ERR_getErrorName(error
 unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize)
 {
     if (dictSize < 8) return 0;
-    if (MEM_readLE32(dictBuffer) != ZSTD_DICT_MAGIC) return 0;
+    if (MEM_readLE32(dictBuffer) != ZSTD_MAGIC_DICTIONARY) return 0;
     return MEM_readLE32((const char*)dictBuffer + 4);
 }
 
@@ -374,7 +375,7 @@ static int isIncluded(const void* in, const void* container, size_t length)
     return u==length;
 }
 
-/*! ZDICT_checkMerge
+/*! ZDICT_tryMerge() :
     check if dictItem can be merged, do it if possible
     @return : id of destination elt, 0 if not merged
 */
@@ -439,8 +440,8 @@ static U32 ZDICT_tryMerge(dictItem* table, dictItem elt, U32 eltNbToSkip, const
 
 static void ZDICT_removeDictItem(dictItem* table, U32 id)
 {
-    /* convention : first element is nb of elts */
-    U32 const max = table->pos;
+    /* convention : table[0].pos stores nb of elts */
+    U32 const max = table[0].pos;
     U32 u;
     if (!id) return;   /* protection, should never happen */
     for (u=id; u<max-1; u++)
@@ -487,7 +488,7 @@ static U32 ZDICT_dictSize(const dictItem* dictList)
 }
 
 
-static size_t ZDICT_trainBuffer(dictItem* dictList, U32 dictListSize,
+static size_t ZDICT_trainBuffer_legacy(dictItem* dictList, U32 dictListSize,
                             const void* const buffer, size_t bufferSize,   /* buffer must end with noisy guard band */
                             const size_t* fileSizes, unsigned nbFiles,
                             U32 minRatio, U32 notificationLevel)
@@ -576,7 +577,7 @@ typedef struct
 {
     ZSTD_CCtx* ref;
     ZSTD_CCtx* zc;
-    void* workPlace;   /* must be ZSTD_BLOCKSIZE_ABSOLUTEMAX allocated */
+    void* workPlace;   /* must be ZSTD_BLOCKSIZE_MAX allocated */
 } EStats_ress_t;
 
 #define MAXREPOFFSET 1024
@@ -585,14 +586,14 @@ static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
                             U32* countLit, U32* offsetcodeCount, U32* matchlengthCount, U32* litlengthCount, U32* repOffsets,
                             const void* src, size_t srcSize, U32 notificationLevel)
 {
-    size_t const blockSizeMax = MIN (ZSTD_BLOCKSIZE_ABSOLUTEMAX, 1 << params.cParams.windowLog);
+    size_t const blockSizeMax = MIN (ZSTD_BLOCKSIZE_MAX, 1 << params.cParams.windowLog);
     size_t cSize;
 
     if (srcSize > blockSizeMax) srcSize = blockSizeMax;   /* protection vs large samples */
     {  size_t const errorCode = ZSTD_copyCCtx(esr.zc, esr.ref, 0);
             if (ZSTD_isError(errorCode)) { DISPLAYLEVEL(1, "warning : ZSTD_copyCCtx failed \n"); return; }
     }
-    cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_ABSOLUTEMAX, src, srcSize);
+    cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_MAX, src, srcSize);
     if (ZSTD_isError(cSize)) { DISPLAYLEVEL(3, "warning : could not compress sample size %u \n", (U32)srcSize); return; }
 
     if (cSize) {  /* if == 0; block is not compressible */
@@ -634,17 +635,6 @@ static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
     }   }   }
 }
 
-/*
-static size_t ZDICT_maxSampleSize(const size_t* fileSizes, unsigned nbFiles)
-{
-    unsigned u;
-    size_t max=0;
-    for (u=0; u<nbFiles; u++)
-        if (max < fileSizes[u]) max = fileSizes[u];
-    return max;
-}
-*/
-
 static size_t ZDICT_totalSampleSize(const size_t* fileSizes, unsigned nbFiles)
 {
     size_t total=0;
@@ -700,13 +690,13 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     /* init */
     esr.ref = ZSTD_createCCtx();
     esr.zc = ZSTD_createCCtx();
-    esr.workPlace = malloc(ZSTD_BLOCKSIZE_ABSOLUTEMAX);
+    esr.workPlace = malloc(ZSTD_BLOCKSIZE_MAX);
     if (!esr.ref || !esr.zc || !esr.workPlace) {
         eSize = ERROR(memory_allocation);
         DISPLAYLEVEL(1, "Not enough memory \n");
         goto _cleanup;
     }
-    if (offcodeMax>OFFCODE_MAX) { eSize = ERROR(dictionary_wrong); goto _cleanup; }   /* too large dictionary */
+    if (offcodeMax>OFFCODE_MAX) { eSize = ERROR(dictionaryCreation_failed); goto _cleanup; }   /* too large dictionary */
     for (u=0; u<256; u++) countLit[u] = 1;   /* any character must be described */
     for (u=0; u<=offcodeMax; u++) offcodeCount[u] = 1;
     for (u=0; u<=MaxML; u++) matchLengthCount[u] = 1;
@@ -714,7 +704,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     memset(repOffset, 0, sizeof(repOffset));
     repOffset[1] = repOffset[4] = repOffset[8] = 1;
     memset(bestRepOffset, 0, sizeof(bestRepOffset));
-    if (compressionLevel==0) compressionLevel = g_compressionLevel_default;
+    if (compressionLevel<=0) compressionLevel = g_compressionLevel_default;
     params = ZSTD_getParams(compressionLevel, averageSampleSize, dictBufferSize);
     {   size_t const beginResult = ZSTD_compressBegin_advanced(esr.ref, dictBuffer, dictBufferSize, params, 0);
         if (ZSTD_isError(beginResult)) {
@@ -865,7 +855,7 @@ size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
     if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) return ERROR(dstSize_tooSmall);
 
     /* dictionary header */
-    MEM_writeLE32(header, ZSTD_DICT_MAGIC);
+    MEM_writeLE32(header, ZSTD_MAGIC_DICTIONARY);
     {   U64 const randomID = XXH64(customDictContent, dictContentSize, 0);
         U32 const compliantID = (randomID % ((1U<<31)-32768)) + 32768;
         U32 const dictID = params.dictID ? params.dictID : compliantID;
@@ -917,7 +907,7 @@ size_t ZDICT_addEntropyTablesFromBuffer_advanced(void* dictBuffer, size_t dictCo
     }
 
     /* add dictionary header (after entropy tables) */
-    MEM_writeLE32(dictBuffer, ZSTD_DICT_MAGIC);
+    MEM_writeLE32(dictBuffer, ZSTD_MAGIC_DICTIONARY);
     {   U64 const randomID = XXH64((char*)dictBuffer + dictBufferCapacity - dictContentSize, dictContentSize, 0);
         U32 const compliantID = (randomID % ((1U<<31)-32768)) + 32768;
         U32 const dictID = params.dictID ? params.dictID : compliantID;
@@ -930,14 +920,14 @@ size_t ZDICT_addEntropyTablesFromBuffer_advanced(void* dictBuffer, size_t dictCo
 }
 
 
-/*! ZDICT_trainFromBuffer_unsafe() :
+/*! ZDICT_trainFromBuffer_unsafe_legacy() :
 *   Warning : `samplesBuffer` must be followed by noisy guard band.
 *   @return : size of dictionary, or an error code which can be tested with ZDICT_isError()
 */
-size_t ZDICT_trainFromBuffer_unsafe(
+size_t ZDICT_trainFromBuffer_unsafe_legacy(
                             void* dictBuffer, size_t maxDictSize,
                             const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
-                            ZDICT_params_t params)
+                            ZDICT_legacy_params_t params)
 {
     U32 const dictListSize = MAX(MAX(DICTLISTSIZE_DEFAULT, nbSamples), (U32)(maxDictSize/16));
     dictItem* const dictList = (dictItem*)malloc(dictListSize * sizeof(*dictList));
@@ -946,7 +936,7 @@ size_t ZDICT_trainFromBuffer_unsafe(
     size_t const targetDictSize = maxDictSize;
     size_t const samplesBuffSize = ZDICT_totalSampleSize(samplesSizes, nbSamples);
     size_t dictSize = 0;
-    U32 const notificationLevel = params.notificationLevel;
+    U32 const notificationLevel = params.zParams.notificationLevel;
 
     /* checks */
     if (!dictList) return ERROR(memory_allocation);
@@ -957,13 +947,13 @@ size_t ZDICT_trainFromBuffer_unsafe(
     ZDICT_initDictItem(dictList);
 
     /* build dictionary */
-    ZDICT_trainBuffer(dictList, dictListSize,
-                    samplesBuffer, samplesBuffSize,
-                    samplesSizes, nbSamples,
-                    minRep, notificationLevel);
+    ZDICT_trainBuffer_legacy(dictList, dictListSize,
+                       samplesBuffer, samplesBuffSize,
+                       samplesSizes, nbSamples,
+                       minRep, notificationLevel);
 
     /* display best matches */
-    if (params.notificationLevel>= 3) {
+    if (params.zParams.notificationLevel>= 3) {
         U32 const nb = MIN(25, dictList[0].pos);
         U32 const dictContentSize = ZDICT_dictSize(dictList);
         U32 u;
@@ -1026,7 +1016,7 @@ size_t ZDICT_trainFromBuffer_unsafe(
 
         dictSize = ZDICT_addEntropyTablesFromBuffer_advanced(dictBuffer, dictContentSize, maxDictSize,
                                                              samplesBuffer, samplesSizes, nbSamples,
-                                                             params);
+                                                             params.zParams);
     }
 
     /* clean up */
@@ -1037,9 +1027,9 @@ size_t ZDICT_trainFromBuffer_unsafe(
 
 /* issue : samplesBuffer need to be followed by a noisy guard band.
 *  work around : duplicate the buffer, and add the noise */
-size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacity,
-                                      const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
-                                      ZDICT_params_t params)
+size_t ZDICT_trainFromBuffer_legacy(void* dictBuffer, size_t dictBufferCapacity,
+                              const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
+                              ZDICT_legacy_params_t params)
 {
     size_t result;
     void* newBuff;
@@ -1052,10 +1042,9 @@ size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacit
     memcpy(newBuff, samplesBuffer, sBuffSize);
     ZDICT_fillNoise((char*)newBuff + sBuffSize, NOISELENGTH);   /* guard band, for end of buffer condition */
 
-    result = ZDICT_trainFromBuffer_unsafe(
-                                        dictBuffer, dictBufferCapacity,
-                                        newBuff, samplesSizes, nbSamples,
-                                        params);
+    result =
+        ZDICT_trainFromBuffer_unsafe_legacy(dictBuffer, dictBufferCapacity, newBuff,
+                                            samplesSizes, nbSamples, params);
     free(newBuff);
     return result;
 }
@@ -1064,11 +1053,15 @@ size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacit
 size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
                              const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
 {
-    ZDICT_params_t params;
+    ZDICT_cover_params_t params;
     memset(&params, 0, sizeof(params));
-    return ZDICT_trainFromBuffer_advanced(dictBuffer, dictBufferCapacity,
-                                          samplesBuffer, samplesSizes, nbSamples,
-                                          params);
+    params.d = 8;
+    params.steps = 4;
+    /* Default to level 6 since no compression level information is avaialble */
+    params.zParams.compressionLevel = 6;
+    return ZDICT_optimizeTrainFromBuffer_cover(dictBuffer, dictBufferCapacity,
+                                               samplesBuffer, samplesSizes,
+                                               nbSamples, &params);
 }
 
 size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
diff --git a/contrib/zstd/zdict.h b/contrib/zstd/zdict.h
index 9b53de346..5f0000b1c 100644
--- a/contrib/zstd/zdict.h
+++ b/contrib/zstd/zdict.h
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 #ifndef DICTBUILDER_H_001
@@ -20,10 +21,12 @@ extern "C" {
 
 
 /* =====   ZDICTLIB_API : control library symbols visibility   ===== */
-#if defined(__GNUC__) && (__GNUC__ >= 4)
-#  define ZDICTLIB_VISIBILITY __attribute__ ((visibility ("default")))
-#else
-#  define ZDICTLIB_VISIBILITY
+#ifndef ZDICTLIB_VISIBILITY
+#  if defined(__GNUC__) && (__GNUC__ >= 4)
+#    define ZDICTLIB_VISIBILITY __attribute__ ((visibility ("default")))
+#  else
+#    define ZDICTLIB_VISIBILITY
+#  endif
 #endif
 #if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
 #  define ZDICTLIB_API __declspec(dllexport) ZDICTLIB_VISIBILITY
@@ -34,18 +37,20 @@ extern "C" {
 #endif
 
 
-/*! ZDICT_trainFromBuffer() :
-    Train a dictionary from an array of samples.
-    Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
-    supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
-    The resulting dictionary will be saved into `dictBuffer`.
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
-              or an error code, which can be tested with ZDICT_isError().
-    Tips : In general, a reasonable dictionary has a size of ~ 100 KB.
-           It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
-           In general, it's recommended to provide a few thousands samples, but this can vary a lot.
-           It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
-*/
+/*! ZDICT_trainFromBuffer():
+ * Train a dictionary from an array of samples.
+ * Uses ZDICT_optimizeTrainFromBuffer_cover() single-threaded, with d=8 and steps=4.
+ * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ * The resulting dictionary will be saved into `dictBuffer`.
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ * Note: ZDICT_trainFromBuffer() requires about 9 bytes of memory for each input byte.
+ * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+ */
 ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
                        const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
 
@@ -67,94 +72,78 @@ ZDICTLIB_API const char* ZDICT_getErrorName(size_t errorCode);
  * ==================================================================================== */
 
 typedef struct {
-    unsigned selectivityLevel;   /* 0 means default; larger => select more => larger dictionary */
     int      compressionLevel;   /* 0 means default; target a specific zstd compression level */
     unsigned notificationLevel;  /* Write to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
     unsigned dictID;             /* 0 means auto mode (32-bits random value); other : force dictID value */
-    unsigned reserved[2];        /* reserved space for future parameters */
 } ZDICT_params_t;
 
-
-/*! ZDICT_trainFromBuffer_advanced() :
-    Same as ZDICT_trainFromBuffer() with control over more parameters.
-    `parameters` is optional and can be provided with values set to 0 to mean "default".
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferSize`),
-              or an error code, which can be tested by ZDICT_isError().
-    note : ZDICT_trainFromBuffer_advanced() will send notifications into stderr if instructed to, using notificationLevel>0.
-*/
-ZDICTLIB_API size_t ZDICT_trainFromBuffer_advanced(void* dictBuffer, size_t dictBufferCapacity,
-                                const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
-                                ZDICT_params_t parameters);
-
-/*! COVER_params_t :
-    For all values 0 means default.
-    k and d are the only required parameters.
-*/
+/*! ZDICT_cover_params_t:
+ *  For all values 0 means default.
+ *  k and d are the only required parameters.
+ */
 typedef struct {
     unsigned k;                  /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
     unsigned d;                  /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
     unsigned steps;              /* Number of steps : Only used for optimization : 0 means default (32) : Higher means more parameters checked */
-
     unsigned nbThreads;          /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
-    unsigned notificationLevel;  /* Write to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
-    unsigned dictID;             /* 0 means auto mode (32-bits random value); other : force dictID value */
-    int      compressionLevel;   /* 0 means default; target a specific zstd compression level */
-} COVER_params_t;
-
-
-/*! COVER_trainFromBuffer() :
-    Train a dictionary from an array of samples using the COVER algorithm.
-    Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
-    supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
-    The resulting dictionary will be saved into `dictBuffer`.
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
-              or an error code, which can be tested with ZDICT_isError().
-    Note : COVER_trainFromBuffer() requires about 9 bytes of memory for each input byte.
-    Tips : In general, a reasonable dictionary has a size of ~ 100 KB.
-           It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
-           In general, it's recommended to provide a few thousands samples, but this can vary a lot.
-           It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
-*/
-ZDICTLIB_API size_t COVER_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
-                              const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
-                              COVER_params_t parameters);
-
-/*! COVER_optimizeTrainFromBuffer() :
-    The same requirements as above hold for all the parameters except `parameters`.
-    This function tries many parameter combinations and picks the best parameters.
-    `*parameters` is filled with the best parameters found, and the dictionary
-    constructed with those parameters is stored in `dictBuffer`.
-
-    All of the parameters d, k, steps are optional.
-    If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8, 10, 12, 14, 16}.
-    if steps is zero it defaults to its default value.
-    If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [16, 2048].
-
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
-              or an error code, which can be tested with ZDICT_isError().
-              On success `*parameters` contains the parameters selected.
-    Note : COVER_optimizeTrainFromBuffer() requires about 8 bytes of memory for each input byte and additionally another 5 bytes of memory for each byte of memory for each thread.
-*/
-ZDICTLIB_API size_t COVER_optimizeTrainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
-                                     const void* samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
-                                     COVER_params_t *parameters);
-
-/*! ZDICT_finalizeDictionary() :
-
-    Given a custom content as a basis for dictionary, and a set of samples,
-    finalize dictionary by adding headers and statistics.
-
-    Samples must be stored concatenated in a flat buffer `samplesBuffer`,
-    supplied with an array of sizes `samplesSizes`, providing the size of each sample in order.
-
-    dictContentSize must be >= ZDICT_CONTENTSIZE_MIN bytes.
-    maxDictSize must be >= dictContentSize, and must be >= ZDICT_DICTSIZE_MIN bytes.
-
-    @return : size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`),
-              or an error code, which can be tested by ZDICT_isError().
-    note : ZDICT_finalizeDictionary() will push notifications into stderr if instructed to, using notificationLevel>0.
-    note 2 : dictBuffer and dictContent can overlap
-*/
+    ZDICT_params_t zParams;
+} ZDICT_cover_params_t;
+
+
+/*! ZDICT_trainFromBuffer_cover():
+ * Train a dictionary from an array of samples using the COVER algorithm.
+ * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ * The resulting dictionary will be saved into `dictBuffer`.
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ * Note: ZDICT_trainFromBuffer_cover() requires about 9 bytes of memory for each input byte.
+ * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+ */
+ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t parameters);
+
+/*! ZDICT_optimizeTrainFromBuffer_cover():
+ * The same requirements as above hold for all the parameters except `parameters`.
+ * This function tries many parameter combinations and picks the best parameters.
+ * `*parameters` is filled with the best parameters found, and the dictionary
+ * constructed with those parameters is stored in `dictBuffer`.
+ *
+ * All of the parameters d, k, steps are optional.
+ * If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8, 10, 12, 14, 16}.
+ * if steps is zero it defaults to its default value.
+ * If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [16, 2048].
+ *
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ *           On success `*parameters` contains the parameters selected.
+ * Note: ZDICT_optimizeTrainFromBuffer_cover() requires about 8 bytes of memory for each input byte and additionally another 5 bytes of memory for each byte of memory for each thread.
+ */
+ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t *parameters);
+
+/*! ZDICT_finalizeDictionary():
+ * Given a custom content as a basis for dictionary, and a set of samples,
+ * finalize dictionary by adding headers and statistics.
+ *
+ * Samples must be stored concatenated in a flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample in order.
+ *
+ * dictContentSize must be >= ZDICT_CONTENTSIZE_MIN bytes.
+ * maxDictSize must be >= dictContentSize, and must be >= ZDICT_DICTSIZE_MIN bytes.
+ *
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`),
+ *           or an error code, which can be tested by ZDICT_isError().
+ * Note: ZDICT_finalizeDictionary() will push notifications into stderr if instructed to, using notificationLevel>0.
+ * Note 2: dictBuffer and dictContent can overlap
+ */
 #define ZDICT_CONTENTSIZE_MIN 128
 #define ZDICT_DICTSIZE_MIN    256
 ZDICTLIB_API size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
@@ -162,7 +151,28 @@ ZDICTLIB_API size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBuffer
                                 const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
                                 ZDICT_params_t parameters);
 
-
+typedef struct {
+    unsigned selectivityLevel;   /* 0 means default; larger => select more => larger dictionary */
+    ZDICT_params_t zParams;
+} ZDICT_legacy_params_t;
+
+/*! ZDICT_trainFromBuffer_legacy():
+ * Train a dictionary from an array of samples.
+ * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ * The resulting dictionary will be saved into `dictBuffer`.
+ * `parameters` is optional and can be provided with values set to 0 to mean "default".
+ * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
+ *           or an error code, which can be tested with ZDICT_isError().
+ * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
+ * Note: ZDICT_trainFromBuffer_legacy() will send notifications into stderr if instructed to, using notificationLevel>0.
+ */
+ZDICTLIB_API size_t ZDICT_trainFromBuffer_legacy(
+    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
+    const size_t *samplesSizes, unsigned nbSamples, ZDICT_legacy_params_t parameters);
 
 /* Deprecation warnings */
 /* It is generally possible to disable deprecation warnings from compiler,
diff --git a/contrib/zstd/zstd.h b/contrib/zstd/zstd.h
index eef2b0b5e..ddb284299 100644
--- a/contrib/zstd/zstd.h
+++ b/contrib/zstd/zstd.h
@@ -2,27 +2,29 @@
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
 #ifndef ZSTD_H_235446
 #define ZSTD_H_235446
-#define ZSTD_STATIC_LINKING_ONLY
+
 /* ======   Dependency   ======*/
 #include <stddef.h>   /* size_t */
 
 
 /* =====   ZSTDLIB_API : control library symbols visibility   ===== */
-#if defined(__GNUC__) && (__GNUC__ >= 4)
-#  define ZSTDLIB_VISIBILITY __attribute__ ((visibility ("default")))
-#else
-#  define ZSTDLIB_VISIBILITY
+#ifndef ZSTDLIB_VISIBILITY
+#  if defined(__GNUC__) && (__GNUC__ >= 4)
+#    define ZSTDLIB_VISIBILITY __attribute__ ((visibility ("default")))
+#  else
+#    define ZSTDLIB_VISIBILITY
+#  endif
 #endif
 #if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
 #  define ZSTDLIB_API __declspec(dllexport) ZSTDLIB_VISIBILITY
@@ -36,35 +38,37 @@ extern "C" {
 /*******************************************************************************************************
   Introduction
 
-  zstd, short for Zstandard, is a fast lossless compression algorithm, targeting real-time compression scenarios
-  at zlib-level and better compression ratios. The zstd compression library provides in-memory compression and
-  decompression functions. The library supports compression levels from 1 up to ZSTD_maxCLevel() which is 22.
+  zstd, short for Zstandard, is a fast lossless compression algorithm,
+  targeting real-time compression scenarios at zlib-level and better compression ratios.
+  The zstd compression library provides in-memory compression and decompression functions.
+  The library supports compression levels from 1 up to ZSTD_maxCLevel() which is currently 22.
   Levels >= 20, labeled `--ultra`, should be used with caution, as they require more memory.
   Compression can be done in:
     - a single step (described as Simple API)
     - a single step, reusing a context (described as Explicit memory management)
     - unbounded multiple steps (described as Streaming compression)
-  The compression ratio achievable on small data can be highly improved using compression with a dictionary in:
+  The compression ratio achievable on small data can be highly improved using a dictionary in:
     - a single step (described as Simple dictionary API)
     - a single step, reusing a dictionary (described as Fast dictionary API)
 
   Advanced experimental functions can be accessed using #define ZSTD_STATIC_LINKING_ONLY before including zstd.h.
-  These APIs shall never be used with a dynamic library.
+  Advanced experimental APIs shall never be used with a dynamic library.
   They are not "stable", their definition may change in the future. Only static linking is allowed.
 *********************************************************************************************************/
 
 /*------   Version   ------*/
 #define ZSTD_VERSION_MAJOR    1
 #define ZSTD_VERSION_MINOR    3
-#define ZSTD_VERSION_RELEASE  0
+#define ZSTD_VERSION_RELEASE  1
+
+#define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
+ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< useful to check dll version */
 
 #define ZSTD_LIB_VERSION ZSTD_VERSION_MAJOR.ZSTD_VERSION_MINOR.ZSTD_VERSION_RELEASE
 #define ZSTD_QUOTE(str) #str
 #define ZSTD_EXPAND_AND_QUOTE(str) ZSTD_QUOTE(str)
 #define ZSTD_VERSION_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_LIB_VERSION)
-
-#define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
-ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< library version number; to be used when checking dll version */
+ZSTDLIB_API const char* ZSTD_versionString(void);   /* v1.3.0 */
 
 
 /***************************************
@@ -81,38 +85,48 @@ ZSTDLIB_API size_t ZSTD_compress( void* dst, size_t dstCapacity,
 
 /*! ZSTD_decompress() :
  *  `compressedSize` : must be the _exact_ size of some number of compressed and/or skippable frames.
- *  `dstCapacity` is an upper bound of originalSize.
+ *  `dstCapacity` is an upper bound of originalSize to regenerate.
  *  If user cannot imply a maximum upper bound, it's better to use streaming mode to decompress data.
  *  @return : the number of bytes decompressed into `dst` (<= `dstCapacity`),
  *            or an errorCode if it fails (which can be tested using ZSTD_isError()). */
 ZSTDLIB_API size_t ZSTD_decompress( void* dst, size_t dstCapacity,
                               const void* src, size_t compressedSize);
 
-/*! ZSTD_getDecompressedSize() :
- *  NOTE: This function is planned to be obsolete, in favour of ZSTD_getFrameContentSize.
- *  ZSTD_getFrameContentSize functions the same way, returning the decompressed size of a single
- *  frame, but distinguishes empty frames from frames with an unknown size, or errors.
- *
- *  Additionally, ZSTD_findDecompressedSize can be used instead.  It can handle multiple
- *  concatenated frames in one buffer, and so is more general.
- *  As a result however, it requires more computation and entire frames to be passed to it,
- *  as opposed to ZSTD_getFrameContentSize which requires only a single frame's header.
- *
- *  'src' is the start of a zstd compressed frame.
- *  @return : content size to be decompressed, as a 64-bits value _if known_, 0 otherwise.
- *   note 1 : decompressed size is an optional field, that may not be present, especially in streaming mode.
- *            When `return==0`, data to decompress could be any size.
+/*! ZSTD_getFrameContentSize() : v1.3.0
+ *  `src` should point to the start of a ZSTD encoded frame.
+ *  `srcSize` must be at least as large as the frame header.
+ *            hint : any size >= `ZSTD_frameHeaderSize_max` is large enough.
+ *  @return : - decompressed size of the frame in `src`, if known
+ *            - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
+ *            - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small)
+ *   note 1 : a 0 return value means the frame is valid but "empty".
+ *   note 2 : decompressed size is an optional field, it may not be present, typically in streaming mode.
+ *            When `return==ZSTD_CONTENTSIZE_UNKNOWN`, data to decompress could be any size.
  *            In which case, it's necessary to use streaming mode to decompress data.
- *            Optionally, application can still use ZSTD_decompress() while relying on implied limits.
- *            (For example, data may be necessarily cut into blocks <= 16 KB).
- *   note 2 : decompressed size is always present when compression is done with ZSTD_compress()
- *   note 3 : decompressed size can be very large (64-bits value),
+ *            Optionally, application can rely on some implicit limit,
+ *            as ZSTD_decompress() only needs an upper bound of decompressed size.
+ *            (For example, data could be necessarily cut into blocks <= 16 KB).
+ *   note 3 : decompressed size is always present when compression is done with ZSTD_compress()
+ *   note 4 : decompressed size can be very large (64-bits value),
  *            potentially larger than what local system can handle as a single memory segment.
  *            In which case, it's necessary to use streaming mode to decompress data.
- *   note 4 : If source is untrusted, decompressed size could be wrong or intentionally modified.
- *            Always ensure result fits within application's authorized limits.
+ *   note 5 : If source is untrusted, decompressed size could be wrong or intentionally modified.
+ *            Always ensure return value fits within application's authorized limits.
  *            Each application can set its own limits.
- *   note 5 : when `return==0`, if precise failure cause is needed, use ZSTD_getFrameParams() to know more. */
+ *   note 6 : This function replaces ZSTD_getDecompressedSize() */
+#define ZSTD_CONTENTSIZE_UNKNOWN (0ULL - 1)
+#define ZSTD_CONTENTSIZE_ERROR   (0ULL - 2)
+ZSTDLIB_API unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize);
+
+/*! ZSTD_getDecompressedSize() :
+ *  NOTE: This function is now obsolete, in favor of ZSTD_getFrameContentSize().
+ *  Both functions work the same way,
+ *  but ZSTD_getDecompressedSize() blends
+ *  "empty", "unknown" and "error" results in the same return value (0),
+ *  while ZSTD_getFrameContentSize() distinguishes them.
+ *
+ *  'src' is the start of a zstd compressed frame.
+ *  @return : content size to be decompressed, as a 64-bits value _if known and not empty_, 0 otherwise. */
 ZSTDLIB_API unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize);
 
 
@@ -137,29 +151,35 @@ ZSTDLIB_API size_t     ZSTD_freeCCtx(ZSTD_CCtx* cctx);
 
 /*! ZSTD_compressCCtx() :
  *  Same as ZSTD_compress(), requires an allocated ZSTD_CCtx (see ZSTD_createCCtx()). */
-ZSTDLIB_API size_t ZSTD_compressCCtx(ZSTD_CCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize, int compressionLevel);
+ZSTDLIB_API size_t ZSTD_compressCCtx(ZSTD_CCtx* ctx,
+                                     void* dst, size_t dstCapacity,
+                               const void* src, size_t srcSize,
+                                     int compressionLevel);
 
 /*= Decompression context
  *  When decompressing many times,
- *  it is recommended to allocate a context just once, and re-use it for each successive compression operation.
+ *  it is recommended to allocate a context only once,
+ *  and re-use it for each successive compression operation.
  *  This will make workload friendlier for system's memory.
- *  Use one context per thread for parallel execution in multi-threaded environments. */
+ *  Use one context per thread for parallel execution. */
 typedef struct ZSTD_DCtx_s ZSTD_DCtx;
 ZSTDLIB_API ZSTD_DCtx* ZSTD_createDCtx(void);
 ZSTDLIB_API size_t     ZSTD_freeDCtx(ZSTD_DCtx* dctx);
 
 /*! ZSTD_decompressDCtx() :
- *  Same as ZSTD_decompress(), requires an allocated ZSTD_DCtx (see ZSTD_createDCtx()). */
-ZSTDLIB_API size_t ZSTD_decompressDCtx(ZSTD_DCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+ *  Same as ZSTD_decompress(), requires an allocated ZSTD_DCtx (see ZSTD_createDCtx()) */
+ZSTDLIB_API size_t ZSTD_decompressDCtx(ZSTD_DCtx* ctx,
+                                       void* dst, size_t dstCapacity,
+                                 const void* src, size_t srcSize);
 
 
 /**************************
 *  Simple dictionary API
 ***************************/
 /*! ZSTD_compress_usingDict() :
-*   Compression using a predefined Dictionary (see dictBuilder/zdict.h).
-*   Note : This function loads the dictionary, resulting in significant startup delay.
-*   Note : When `dict == NULL || dictSize < 8` no dictionary is used. */
+ *  Compression using a predefined Dictionary (see dictBuilder/zdict.h).
+ *  Note : This function loads the dictionary, resulting in significant startup delay.
+ *  Note : When `dict == NULL || dictSize < 8` no dictionary is used. */
 ZSTDLIB_API size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx,
                                            void* dst, size_t dstCapacity,
                                      const void* src, size_t srcSize,
@@ -167,30 +187,31 @@ ZSTDLIB_API size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx,
                                            int compressionLevel);
 
 /*! ZSTD_decompress_usingDict() :
-*   Decompression using a predefined Dictionary (see dictBuilder/zdict.h).
-*   Dictionary must be identical to the one used during compression.
-*   Note : This function loads the dictionary, resulting in significant startup delay.
-*   Note : When `dict == NULL || dictSize < 8` no dictionary is used. */
+ *  Decompression using a predefined Dictionary (see dictBuilder/zdict.h).
+ *  Dictionary must be identical to the one used during compression.
+ *  Note : This function loads the dictionary, resulting in significant startup delay.
+ *  Note : When `dict == NULL || dictSize < 8` no dictionary is used. */
 ZSTDLIB_API size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
                                              void* dst, size_t dstCapacity,
                                        const void* src, size_t srcSize,
                                        const void* dict,size_t dictSize);
 
 
-/****************************
-*  Fast dictionary API
-****************************/
+/**********************************
+ *  Bulk processing dictionary API
+ *********************************/
 typedef struct ZSTD_CDict_s ZSTD_CDict;
 
 /*! ZSTD_createCDict() :
-*   When compressing multiple messages / blocks with the same dictionary, it's recommended to load it just once.
-*   ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup delay.
-*   ZSTD_CDict can be created once and used by multiple threads concurrently, as its usage is read-only.
-*   `dictBuffer` can be released after ZSTD_CDict creation, as its content is copied within CDict */
-ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict(const void* dictBuffer, size_t dictSize, int compressionLevel);
+ *  When compressing multiple messages / blocks with the same dictionary, it's recommended to load it just once.
+ *  ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup delay.
+ *  ZSTD_CDict can be created once and shared by multiple threads concurrently, since its usage is read-only.
+ *  `dictBuffer` can be released after ZSTD_CDict creation, since its content is copied within CDict */
+ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict(const void* dictBuffer, size_t dictSize,
+                                         int compressionLevel);
 
 /*! ZSTD_freeCDict() :
-*   Function frees memory allocated by ZSTD_createCDict(). */
+ *  Function frees memory allocated by ZSTD_createCDict(). */
 ZSTDLIB_API size_t      ZSTD_freeCDict(ZSTD_CDict* CDict);
 
 /*! ZSTD_compress_usingCDict() :
@@ -207,17 +228,17 @@ ZSTDLIB_API size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
 typedef struct ZSTD_DDict_s ZSTD_DDict;
 
 /*! ZSTD_createDDict() :
-*   Create a digested dictionary, ready to start decompression operation without startup delay.
-*   dictBuffer can be released after DDict creation, as its content is copied inside DDict */
+ *  Create a digested dictionary, ready to start decompression operation without startup delay.
+ *  dictBuffer can be released after DDict creation, as its content is copied inside DDict */
 ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict(const void* dictBuffer, size_t dictSize);
 
 /*! ZSTD_freeDDict() :
-*   Function frees memory allocated with ZSTD_createDDict() */
+ *  Function frees memory allocated with ZSTD_createDDict() */
 ZSTDLIB_API size_t      ZSTD_freeDDict(ZSTD_DDict* ddict);
 
 /*! ZSTD_decompress_usingDDict() :
-*   Decompression using a digested Dictionary.
-*   Faster startup than ZSTD_decompress_usingDict(), recommended when same dictionary is used multiple times. */
+ *  Decompression using a digested Dictionary.
+ *  Faster startup than ZSTD_decompress_usingDict(), recommended when same dictionary is used multiple times. */
 ZSTDLIB_API size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
                                               void* dst, size_t dstCapacity,
                                         const void* src, size_t srcSize,
@@ -274,15 +295,17 @@ typedef struct ZSTD_outBuffer_s {
 *  ZSTD_endStream() instructs to finish a frame.
 *  It will perform a flush and write frame epilogue.
 *  The epilogue is required for decoders to consider a frame completed.
-*  Similar to ZSTD_flushStream(), it may not be able to flush the full content if `output->size` is too small.
+*  ZSTD_endStream() may not be able to flush full data if `output->size` is too small.
 *  In which case, call again ZSTD_endStream() to complete the flush.
-*  @return : nb of bytes still present within internal buffer (0 if it's empty, hence compression completed)
+*  @return : 0 if frame fully completed and fully flushed,
+             or >0 if some data is still present within internal buffer
+                  (value is minimum size estimation for remaining data to flush, but it could be more)
 *            or an error code, which can be tested using ZSTD_isError().
 *
 * *******************************************************************/
 
-typedef ZSTD_CCtx ZSTD_CStream;  /**< CCtx and CStream are effectively same object */
-                                 /* Continue due distinghish them for compatibility with versions <= v1.2.0 */
+typedef ZSTD_CCtx ZSTD_CStream;  /**< CCtx and CStream are now effectively same object (>= v1.3.0) */
+                                 /* Continue to distinguish them for compatibility with versions <= v1.2.0 */
 /*===== ZSTD_CStream management functions =====*/
 ZSTDLIB_API ZSTD_CStream* ZSTD_createCStream(void);
 ZSTDLIB_API size_t ZSTD_freeCStream(ZSTD_CStream* zcs);
@@ -320,7 +343,8 @@ ZSTDLIB_API size_t ZSTD_CStreamOutSize(void);   /**< recommended size for output
 *            The return value is a suggested next input size (a hint to improve latency) that will never load more than the current frame.
 * *******************************************************************************/
 
-typedef struct ZSTD_DStream_s ZSTD_DStream;
+typedef ZSTD_DCtx ZSTD_DStream;  /**< DCtx and DStream are now effectively same object (>= v1.3.0) */
+                                 /* Continue to distinguish them for compatibility with versions <= v1.2.0 */
 /*===== ZSTD_DStream management functions =====*/
 ZSTDLIB_API ZSTD_DStream* ZSTD_createDStream(void);
 ZSTDLIB_API size_t ZSTD_freeDStream(ZSTD_DStream* zds);
@@ -335,20 +359,22 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output
 #endif  /* ZSTD_H_235446 */
 
 
-#if defined(ZSTD_STATIC_LINKING_ONLY) && !defined(ZSTD_H_ZSTD_STATIC_LINKING_ONLY)
-#define ZSTD_H_ZSTD_STATIC_LINKING_ONLY
 
 /****************************************************************************************
  * START OF ADVANCED AND EXPERIMENTAL FUNCTIONS
  * The definitions in this section are considered experimental.
- * They should never be used with a dynamic library, as they may change in the future.
- * They are provided for advanced usages.
+ * They should never be used with a dynamic library, as prototypes may change in the future.
+ * They are provided for advanced scenarios.
  * Use them only in association with static linking.
  * ***************************************************************************************/
 
+#if defined(ZSTD_STATIC_LINKING_ONLY) && !defined(ZSTD_H_ZSTD_STATIC_LINKING_ONLY)
+#define ZSTD_H_ZSTD_STATIC_LINKING_ONLY
+
 /* --- Constants ---*/
 #define ZSTD_MAGICNUMBER            0xFD2FB528   /* >= v0.8.0 */
 #define ZSTD_MAGIC_SKIPPABLE_START  0x184D2A50U
+#define ZSTD_MAGIC_DICTIONARY       0xEC30A437   /* v0.7+ */
 
 #define ZSTD_WINDOWLOG_MAX_32  27
 #define ZSTD_WINDOWLOG_MAX_64  27
@@ -365,17 +391,21 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output
 #define ZSTD_SEARCHLENGTH_MIN   3   /* only for ZSTD_btopt, other strategies are limited to 4 */
 #define ZSTD_TARGETLENGTH_MIN   4
 #define ZSTD_TARGETLENGTH_MAX 999
+#define ZSTD_LDM_MINMATCH_MIN 4
+#define ZSTD_LDM_MINMATCH_MAX 4096
+#define ZSTD_LDM_BUCKETSIZELOG_MAX 8
 
 #define ZSTD_FRAMEHEADERSIZE_MAX 18    /* for static allocation */
 #define ZSTD_FRAMEHEADERSIZE_MIN  6
-static const size_t ZSTD_frameHeaderSize_prefix = 5;
-static const size_t ZSTD_frameHeaderSize_min = ZSTD_FRAMEHEADERSIZE_MIN;
+static const size_t ZSTD_frameHeaderSize_prefix = 5;  /* minimum input size to know frame header size */
 static const size_t ZSTD_frameHeaderSize_max = ZSTD_FRAMEHEADERSIZE_MAX;
+static const size_t ZSTD_frameHeaderSize_min = ZSTD_FRAMEHEADERSIZE_MIN;
 static const size_t ZSTD_skippableHeaderSize = 8;  /* magic number + skippable frame length */
 
 
 /*--- Advanced types ---*/
-typedef enum { ZSTD_fast, ZSTD_dfast, ZSTD_greedy, ZSTD_lazy, ZSTD_lazy2, ZSTD_btlazy2, ZSTD_btopt, ZSTD_btultra } ZSTD_strategy;   /* from faster to stronger */
+typedef enum { ZSTD_fast=1, ZSTD_dfast, ZSTD_greedy, ZSTD_lazy, ZSTD_lazy2,
+               ZSTD_btlazy2, ZSTD_btopt, ZSTD_btultra } ZSTD_strategy;   /* from faster to stronger */
 
 typedef struct {
     unsigned windowLog;      /**< largest match distance : larger == more compression, more memory needed during decompression */
@@ -398,17 +428,15 @@ typedef struct {
     ZSTD_frameParameters fParams;
 } ZSTD_parameters;
 
-typedef struct {
-    unsigned long long frameContentSize;
-    unsigned windowSize;
-    unsigned dictID;
-    unsigned checksumFlag;
-} ZSTD_frameHeader;
+typedef struct ZSTD_CCtx_params_s ZSTD_CCtx_params;
 
 /*= Custom memory allocation functions */
 typedef void* (*ZSTD_allocFunction) (void* opaque, size_t size);
 typedef void  (*ZSTD_freeFunction) (void* opaque, void* address);
 typedef struct { ZSTD_allocFunction customAlloc; ZSTD_freeFunction customFree; void* opaque; } ZSTD_customMem;
+/* use this constant to defer to stdlib's functions */
+static const ZSTD_customMem ZSTD_defaultCMem = { NULL, NULL, NULL };
+
 
 /***************************************
 *  Frame size functions
@@ -417,26 +445,15 @@ typedef struct { ZSTD_allocFunction customAlloc; ZSTD_freeFunction customFree; v
 /*! ZSTD_findFrameCompressedSize() :
  *  `src` should point to the start of a ZSTD encoded frame or skippable frame
  *  `srcSize` must be at least as large as the frame
- *  @return : the compressed size of the frame pointed to by `src`,
+ *  @return : the compressed size of the first frame starting at `src`,
  *            suitable to pass to `ZSTD_decompress` or similar,
- *            or an error code if given invalid input. */
+ *            or an error code if input is invalid */
 ZSTDLIB_API size_t ZSTD_findFrameCompressedSize(const void* src, size_t srcSize);
 
-/*! ZSTD_getFrameContentSize() :
- *  `src` should point to the start of a ZSTD encoded frame.
- *  `srcSize` must be at least as large as the frame header.
- *       A value >= `ZSTD_frameHeaderSize_max` is guaranteed to be large enough.
- *  @return : - decompressed size of the frame pointed to be `src` if known
- *            - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
- *            - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */
-#define ZSTD_CONTENTSIZE_UNKNOWN (0ULL - 1)
-#define ZSTD_CONTENTSIZE_ERROR   (0ULL - 2)
-ZSTDLIB_API unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize);
-
 /*! ZSTD_findDecompressedSize() :
  *  `src` should point the start of a series of ZSTD encoded and/or skippable frames
  *  `srcSize` must be the _exact_ size of this series
- *       (i.e. there should be a frame boundary exactly `srcSize` bytes after `src`)
+ *       (i.e. there should be a frame boundary exactly at `srcSize` bytes after `src`)
  *  @return : - decompressed size of all data in all successive frames
  *            - if the decompressed size cannot be determined: ZSTD_CONTENTSIZE_UNKNOWN
  *            - if an error occurred: ZSTD_CONTENTSIZE_ERROR
@@ -444,8 +461,6 @@ ZSTDLIB_API unsigned long long ZSTD_getFrameContentSize(const void *src, size_t
  *   note 1 : decompressed size is an optional field, that may not be present, especially in streaming mode.
  *            When `return==ZSTD_CONTENTSIZE_UNKNOWN`, data to decompress could be any size.
  *            In which case, it's necessary to use streaming mode to decompress data.
- *            Optionally, application can still use ZSTD_decompress() while relying on implied limits.
- *            (For example, data may be necessarily cut into blocks <= 16 KB).
  *   note 2 : decompressed size is always present when compression is done with ZSTD_compress()
  *   note 3 : decompressed size can be very large (64-bits value),
  *            potentially larger than what local system can handle as a single memory segment.
@@ -454,10 +469,16 @@ ZSTDLIB_API unsigned long long ZSTD_getFrameContentSize(const void *src, size_t
  *            Always ensure result fits within application's authorized limits.
  *            Each application can set its own limits.
  *   note 5 : ZSTD_findDecompressedSize handles multiple frames, and so it must traverse the input to
- *            read each contained frame header.  This is efficient as most of the data is skipped,
+ *            read each contained frame header.  This is fast as most of the data is skipped,
  *            however it does mean that all frame data must be present and valid. */
 ZSTDLIB_API unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize);
 
+/*! ZSTD_frameHeaderSize() :
+*   `src` should point to the start of a ZSTD frame
+*   `srcSize` must be >= ZSTD_frameHeaderSize_prefix.
+*   @return : size of the Frame Header */
+ZSTDLIB_API size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
+
 
 /***************************************
 *  Context memory usage
@@ -475,23 +496,50 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
 
 /*! ZSTD_estimate*() :
  *  These functions make it possible to estimate memory usage
- *  of a future target object, before its allocation,
- *  given a set of parameters, which vary depending on target object.
- *  The objective is to guide decision before allocation. */
-ZSTDLIB_API size_t ZSTD_estimateCCtxSize(ZSTD_compressionParameters cParams);
+ *  of a future {D,C}Ctx, before its creation.
+ *  ZSTD_estimateCCtxSize() will provide a budget large enough for any compression level up to selected one.
+ *  It will also consider src size to be arbitrarily "large", which is worst case.
+ *  If srcSize is known to always be small, ZSTD_estimateCCtxSize_advanced_usingCParams() can provide a tighter estimation.
+ *  ZSTD_estimateCCtxSize_advanced_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
+ *  ZSTD_estimateCCtxSize_advanced_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbThreads is > 1.
+ *  Note : CCtx estimation is only correct for single-threaded compression */
+ZSTDLIB_API size_t ZSTD_estimateCCtxSize(int compressionLevel);
+ZSTDLIB_API size_t ZSTD_estimateCCtxSize_advanced_usingCParams(ZSTD_compressionParameters cParams);
+ZSTDLIB_API size_t ZSTD_estimateCCtxSize_advanced_usingCCtxParams(const ZSTD_CCtx_params* params);
 ZSTDLIB_API size_t ZSTD_estimateDCtxSize(void);
 
-/*! ZSTD_estimate?StreamSize() :
+/*! ZSTD_estimateCStreamSize() :
+ *  ZSTD_estimateCStreamSize() will provide a budget large enough for any compression level up to selected one.
+ *  It will also consider src size to be arbitrarily "large", which is worst case.
+ *  If srcSize is known to always be small, ZSTD_estimateCStreamSize_advanced_usingCParams() can provide a tighter estimation.
+ *  ZSTD_estimateCStreamSize_advanced_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
+ *  ZSTD_estimateCStreamSize_advanced_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbThreads is set to a value > 1.
+ *  Note : CStream estimation is only correct for single-threaded compression.
+ *  ZSTD_DStream memory budget depends on window Size.
+ *  This information can be passed manually, using ZSTD_estimateDStreamSize,
+ *  or deducted from a valid frame Header, using ZSTD_estimateDStreamSize_fromFrame();
  *  Note : if streaming is init with function ZSTD_init?Stream_usingDict(),
- *         an internal ?Dict will be created, which size is not estimated.
- *         In this case, get additional size by using ZSTD_estimate?DictSize */
-ZSTDLIB_API size_t ZSTD_estimateCStreamSize(ZSTD_compressionParameters cParams);
-ZSTDLIB_API size_t ZSTD_estimateDStreamSize(ZSTD_frameHeader fHeader);
+ *         an internal ?Dict will be created, which additional size is not estimated here.
+ *         In this case, get total size by adding ZSTD_estimate?DictSize */
+ZSTDLIB_API size_t ZSTD_estimateCStreamSize(int compressionLevel);
+ZSTDLIB_API size_t ZSTD_estimateCStreamSize_advanced_usingCParams(ZSTD_compressionParameters cParams);
+ZSTDLIB_API size_t ZSTD_estimateCStreamSize_advanced_usingCCtxParams(const ZSTD_CCtx_params* params);
+ZSTDLIB_API size_t ZSTD_estimateDStreamSize(size_t windowSize);
+ZSTDLIB_API size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize);
+
+typedef enum {
+    ZSTD_dlm_byCopy = 0,      /* Copy dictionary content internally. */
+    ZSTD_dlm_byRef,           /* Reference dictionary content -- the dictionary buffer must outlives its users. */
+} ZSTD_dictLoadMethod_e;
 
 /*! ZSTD_estimate?DictSize() :
- *  Note : if dictionary is created "byReference", reduce estimation by dictSize */
-ZSTDLIB_API size_t ZSTD_estimateCDictSize(ZSTD_compressionParameters cParams, size_t dictSize);
-ZSTDLIB_API size_t ZSTD_estimateDDictSize(size_t dictSize);
+ *  ZSTD_estimateCDictSize() will bet that src size is relatively "small", and content is copied, like ZSTD_createCDict().
+ *  ZSTD_estimateCStreamSize_advanced_usingCParams() makes it possible to control precisely compression parameters, like ZSTD_createCDict_advanced().
+ *  Note : dictionary created by reference using ZSTD_dlm_byRef are smaller
+ */
+ZSTDLIB_API size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel);
+ZSTDLIB_API size_t ZSTD_estimateCDictSize_advanced(size_t dictSize, ZSTD_compressionParameters cParams, ZSTD_dictLoadMethod_e dictLoadMethod);
+ZSTDLIB_API size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod);
 
 
 /***************************************
@@ -501,14 +549,23 @@ ZSTDLIB_API size_t ZSTD_estimateDDictSize(size_t dictSize);
  *  Create a ZSTD compression context using external alloc and free functions */
 ZSTDLIB_API ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem);
 
-typedef enum {
-    ZSTD_p_forceWindow,   /* Force back-references to remain < windowSize, even when referencing Dictionary content (default:0) */
-    ZSTD_p_forceRawDict   /* Force loading dictionary in "content-only" mode (no header analysis) */
-} ZSTD_CCtxParameter;
-/*! ZSTD_setCCtxParameter() :
- *  Set advanced parameters, selected through enum ZSTD_CCtxParameter
- *  @result : 0, or an error code (which can be tested with ZSTD_isError()) */
-ZSTDLIB_API size_t ZSTD_setCCtxParameter(ZSTD_CCtx* cctx, ZSTD_CCtxParameter param, unsigned value);
+/*! ZSTD_initStaticCCtx() : initialize a fixed-size zstd compression context
+ *  workspace: The memory area to emplace the context into.
+ *             Provided pointer must 8-bytes aligned.
+ *             It must outlive context usage.
+ *  workspaceSize: Use ZSTD_estimateCCtxSize() or ZSTD_estimateCStreamSize()
+ *                 to determine how large workspace must be to support scenario.
+ * @return : pointer to ZSTD_CCtx*, or NULL if error (size too small)
+ *  Note : zstd will never resize nor malloc() when using a static cctx.
+ *         If it needs more memory than available, it will simply error out.
+ *  Note 2 : there is no corresponding "free" function.
+ *           Since workspace was allocated externally, it must be freed externally too.
+ *  Limitation 1 : currently not compatible with internal CDict creation, such as
+ *                 ZSTD_CCtx_loadDictionary() or ZSTD_initCStream_usingDict().
+ *  Limitation 2 : currently not compatible with multi-threading
+ */
+ZSTDLIB_API ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize);
+
 
 /*! ZSTD_createCDict_byReference() :
  *  Create a digested dictionary for compression
@@ -516,10 +573,36 @@ ZSTDLIB_API size_t ZSTD_setCCtxParameter(ZSTD_CCtx* cctx, ZSTD_CCtxParameter par
  *  It is important that dictBuffer outlives CDict, it must remain read accessible throughout the lifetime of CDict */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
 
+typedef enum { ZSTD_dm_auto=0,        /* dictionary is "full" if it starts with ZSTD_MAGIC_DICTIONARY, otherwise it is "rawContent" */
+               ZSTD_dm_rawContent,    /* ensures dictionary is always loaded as rawContent, even if it starts with ZSTD_MAGIC_DICTIONARY */
+               ZSTD_dm_fullDict       /* refuses to load a dictionary if it does not respect Zstandard's specification */
+} ZSTD_dictMode_e;
 /*! ZSTD_createCDict_advanced() :
  *  Create a ZSTD_CDict using external alloc and free, and customized compression parameters */
-ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize, unsigned byReference,
-                                                  ZSTD_compressionParameters cParams, ZSTD_customMem customMem);
+ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize,
+                                                  ZSTD_dictLoadMethod_e dictLoadMethod,
+                                                  ZSTD_dictMode_e dictMode,
+                                                  ZSTD_compressionParameters cParams,
+                                                  ZSTD_customMem customMem);
+
+/*! ZSTD_initStaticCDict_advanced() :
+ *  Generate a digested dictionary in provided memory area.
+ *  workspace: The memory area to emplace the dictionary into.
+ *             Provided pointer must 8-bytes aligned.
+ *             It must outlive dictionary usage.
+ *  workspaceSize: Use ZSTD_estimateCDictSize()
+ *                 to determine how large workspace must be.
+ *  cParams : use ZSTD_getCParams() to transform a compression level
+ *            into its relevants cParams.
+ * @return : pointer to ZSTD_CDict*, or NULL if error (size too small)
+ *  Note : there is no corresponding "free" function.
+ *         Since workspace was allocated externally, it must be freed externally.
+ */
+ZSTDLIB_API ZSTD_CDict* ZSTD_initStaticCDict(
+                            void* workspace, size_t workspaceSize,
+                      const void* dict, size_t dictSize,
+                            ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode,
+                            ZSTD_compressionParameters cParams);
 
 /*! ZSTD_getCParams() :
 *   @return ZSTD_compressionParameters structure for a selected compression level and estimated srcSize.
@@ -536,8 +619,8 @@ ZSTDLIB_API ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long l
 ZSTDLIB_API size_t ZSTD_checkCParams(ZSTD_compressionParameters params);
 
 /*! ZSTD_adjustCParams() :
-*   optimize params for a given `srcSize` and `dictSize`.
-*   both values are optional, select `0` if unknown. */
+ *  optimize params for a given `srcSize` and `dictSize`.
+ *  both values are optional, select `0` if unknown. */
 ZSTDLIB_API ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize);
 
 /*! ZSTD_compress_advanced() :
@@ -569,16 +652,50 @@ ZSTDLIB_API unsigned ZSTD_isFrame(const void* buffer, size_t size);
  *  Create a ZSTD decompression context using external alloc and free functions */
 ZSTDLIB_API ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem);
 
+/*! ZSTD_initStaticDCtx() : initialize a fixed-size zstd decompression context
+ *  workspace: The memory area to emplace the context into.
+ *             Provided pointer must 8-bytes aligned.
+ *             It must outlive context usage.
+ *  workspaceSize: Use ZSTD_estimateDCtxSize() or ZSTD_estimateDStreamSize()
+ *                 to determine how large workspace must be to support scenario.
+ * @return : pointer to ZSTD_DCtx*, or NULL if error (size too small)
+ *  Note : zstd will never resize nor malloc() when using a static dctx.
+ *         If it needs more memory than available, it will simply error out.
+ *  Note 2 : static dctx is incompatible with legacy support
+ *  Note 3 : there is no corresponding "free" function.
+ *           Since workspace was allocated externally, it must be freed externally.
+ *  Limitation : currently not compatible with internal DDict creation,
+ *               such as ZSTD_initDStream_usingDict().
+ */
+ZSTDLIB_API ZSTD_DCtx* ZSTD_initStaticDCtx(void* workspace, size_t workspaceSize);
+
 /*! ZSTD_createDDict_byReference() :
  *  Create a digested dictionary, ready to start decompression operation without startup delay.
- *  Dictionary content is simply referenced, and therefore stays in dictBuffer.
- *  It is important that dictBuffer outlives DDict, it must remain read accessible throughout the lifetime of DDict */
+ *  Dictionary content is referenced, and therefore stays in dictBuffer.
+ *  It is important that dictBuffer outlives DDict,
+ *  it must remain read accessible throughout the lifetime of DDict */
 ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize);
 
 /*! ZSTD_createDDict_advanced() :
  *  Create a ZSTD_DDict using external alloc and free, optionally by reference */
 ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
-                                                  unsigned byReference, ZSTD_customMem customMem);
+                                                  ZSTD_dictLoadMethod_e dictLoadMethod,
+                                                  ZSTD_customMem customMem);
+
+/*! ZSTD_initStaticDDict() :
+ *  Generate a digested dictionary in provided memory area.
+ *  workspace: The memory area to emplace the dictionary into.
+ *             Provided pointer must 8-bytes aligned.
+ *             It must outlive dictionary usage.
+ *  workspaceSize: Use ZSTD_estimateDDictSize()
+ *                 to determine how large workspace must be.
+ * @return : pointer to ZSTD_DDict*, or NULL if error (size too small)
+ *  Note : there is no corresponding "free" function.
+ *         Since workspace was allocated externally, it must be freed externally.
+ */
+ZSTDLIB_API ZSTD_DDict* ZSTD_initStaticDDict(void* workspace, size_t workspaceSize,
+                                             const void* dict, size_t dictSize,
+                                             ZSTD_dictLoadMethod_e dictLoadMethod);
 
 /*! ZSTD_getDictID_fromDict() :
  *  Provides the dictID stored within dictionary.
@@ -601,7 +718,7 @@ ZSTDLIB_API unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict);
  *    Note : this use case also happens when using a non-conformant dictionary.
  *  - `srcSize` is too small, and as a result, the frame header could not be decoded (only possible if `srcSize < ZSTD_FRAMEHEADERSIZE_MAX`).
  *  - This is not a Zstandard frame.
- *  When identifying the exact failure cause, it's possible to use ZSTD_getFrameParams(), which will provide a more precise error code. */
+ *  When identifying the exact failure cause, it's possible to use ZSTD_getFrameHeader(), which will provide a more precise error code. */
 ZSTDLIB_API unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize);
 
 
@@ -611,12 +728,13 @@ ZSTDLIB_API unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize);
 
 /*=====   Advanced Streaming compression functions  =====*/
 ZSTDLIB_API ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem);
+ZSTDLIB_API ZSTD_CStream* ZSTD_initStaticCStream(void* workspace, size_t workspaceSize);    /**< same as ZSTD_initStaticCCtx() */
 ZSTDLIB_API size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize);   /**< pledgedSrcSize must be correct, a size of 0 means unknown.  for a frame size of 0 use initCStream_advanced */
-ZSTDLIB_API size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel); /**< note: a dict will not be used if dict == NULL or dictSize < 8. This result in the creation of an internal CDict */
+ZSTDLIB_API size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel); /**< creates of an internal CDict (incompatible with static CCtx), except if dict == NULL or dictSize < 8, in which case no dict is used. Note: dict is loaded with ZSTD_dm_auto (treated as a full zstd dictionary if it begins with ZSTD_MAGIC_DICTIONARY, else as raw content) and ZSTD_dlm_byCopy.*/
 ZSTDLIB_API size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs, const void* dict, size_t dictSize,
-                                             ZSTD_parameters params, unsigned long long pledgedSrcSize);  /**< pledgedSrcSize is optional and can be 0 (meaning unknown). note: if the contentSizeFlag is set, pledgedSrcSize == 0 means the source size is actually 0 */
+                                             ZSTD_parameters params, unsigned long long pledgedSrcSize);  /**< pledgedSrcSize is optional and can be 0 (meaning unknown). note: if the contentSizeFlag is set, pledgedSrcSize == 0 means the source size is actually 0. dict is loaded with ZSTD_dm_auto and ZSTD_dlm_byCopy. */
 ZSTDLIB_API size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict);  /**< note : cdict will just be referenced, and must outlive compression session */
-ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, unsigned long long pledgedSrcSize, ZSTD_frameParameters fParams);  /**< same as ZSTD_initCStream_usingCDict(), with control over frame parameters */
+ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize);  /**< same as ZSTD_initCStream_usingCDict(), with control over frame parameters */
 
 /*! ZSTD_resetCStream() :
  *  start a new compression job, using same parameters from previous job.
@@ -631,6 +749,7 @@ ZSTDLIB_API size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledg
 /*=====   Advanced Streaming decompression functions  =====*/
 typedef enum { DStream_p_maxWindowSize } ZSTD_DStreamParameter_e;
 ZSTDLIB_API ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem);
+ZSTDLIB_API ZSTD_DStream* ZSTD_initStaticDStream(void* workspace, size_t workspaceSize);    /**< same as ZSTD_initStaticDCtx() */
 ZSTDLIB_API size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds, ZSTD_DStreamParameter_e paramType, unsigned paramValue);
 ZSTDLIB_API size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize); /**< note: a dict will not be used if dict == NULL or dictSize < 8 */
 ZSTDLIB_API size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict);  /**< note : ddict will just be referenced, and must outlive decompression session */
@@ -688,7 +807,6 @@ ZSTDLIB_API size_t ZSTD_compressContinue(ZSTD_CCtx* cctx, void* dst, size_t dstC
 ZSTDLIB_API size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
 
 
-
 /*-
   Buffer-less streaming decompression (synchronous mode)
 
@@ -696,38 +814,54 @@ ZSTDLIB_API size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapaci
   Use ZSTD_createDCtx() / ZSTD_freeDCtx() to manage it.
   A ZSTD_DCtx object can be re-used multiple times.
 
-  First typical operation is to retrieve frame parameters, using ZSTD_getFrameParams().
-  It fills a ZSTD_frameParams structure which provide important information to correctly decode the frame,
-  such as the minimum rolling buffer size to allocate to decompress data (`windowSize`),
-  and the dictionary ID used.
-  (Note : content size is optional, it may not be present. 0 means : content size unknown).
-  Note that these values could be wrong, either because of data malformation, or because an attacker is spoofing deliberate false information.
-  As a consequence, check that values remain within valid application range, especially `windowSize`, before allocation.
-  Each application can set its own limit, depending on local restrictions. For extended interoperability, it is recommended to support at least 8 MB.
-  Frame parameters are extracted from the beginning of the compressed frame.
-  Data fragment must be large enough to ensure successful decoding, typically `ZSTD_frameHeaderSize_max` bytes.
-  @result : 0 : successful decoding, the `ZSTD_frameParams` structure is correctly filled.
+  First typical operation is to retrieve frame parameters, using ZSTD_getFrameHeader().
+  Frame header is extracted from the beginning of compressed frame, so providing only the frame's beginning is enough.
+  Data fragment must be large enough to ensure successful decoding.
+ `ZSTD_frameHeaderSize_max` bytes is guaranteed to always be large enough.
+  @result : 0 : successful decoding, the `ZSTD_frameHeader` structure is correctly filled.
            >0 : `srcSize` is too small, please provide at least @result bytes on next attempt.
            errorCode, which can be tested using ZSTD_isError().
 
-  Start decompression, with ZSTD_decompressBegin() or ZSTD_decompressBegin_usingDict().
-  Alternatively, you can copy a prepared context, using ZSTD_copyDCtx().
+  It fills a ZSTD_frameHeader structure with important information to correctly decode the frame,
+  such as the dictionary ID, content size, or maximum back-reference distance (`windowSize`).
+  Note that these values could be wrong, either because of data corruption, or because a 3rd party deliberately spoofs false information.
+  As a consequence, check that values remain within valid application range.
+  For example, do not allocate memory blindly, check that `windowSize` is within expectation.
+  Each application can set its own limits, depending on local restrictions.
+  For extended interoperability, it is recommended to support `windowSize` of at least 8 MB.
+
+  ZSTD_decompressContinue() needs previous data blocks during decompression, up to `windowSize` bytes.
+  ZSTD_decompressContinue() is very sensitive to contiguity,
+  if 2 blocks don't follow each other, make sure that either the compressor breaks contiguity at the same place,
+  or that previous contiguous segment is large enough to properly handle maximum back-reference distance.
+  There are multiple ways to guarantee this condition.
+
+  The most memory efficient way is to use a round buffer of sufficient size.
+  Sufficient size is determined by invoking ZSTD_decodingBufferSize_min(),
+  which can @return an error code if required value is too large for current system (in 32-bits mode).
+  In a round buffer methodology, ZSTD_decompressContinue() decompresses each block next to previous one,
+  up to the moment there is not enough room left in the buffer to guarantee decoding another full block,
+  which maximum size is provided in `ZSTD_frameHeader` structure, field `blockSizeMax`.
+  At which point, decoding can resume from the beginning of the buffer.
+  Note that already decoded data stored in the buffer should be flushed before being overwritten.
+
+  There are alternatives possible, for example using two or more buffers of size `windowSize` each, though they consume more memory.
+
+  Finally, if you control the compression process, you can also ignore all buffer size rules,
+  as long as the encoder and decoder progress in "lock-step",
+  aka use exactly the same buffer sizes, break contiguity at the same place, etc.
+
+  Once buffers are setup, start decompression, with ZSTD_decompressBegin().
+  If decompression requires a dictionary, use ZSTD_decompressBegin_usingDict() or ZSTD_decompressBegin_usingDDict().
 
   Then use ZSTD_nextSrcSizeToDecompress() and ZSTD_decompressContinue() alternatively.
   ZSTD_nextSrcSizeToDecompress() tells how many bytes to provide as 'srcSize' to ZSTD_decompressContinue().
   ZSTD_decompressContinue() requires this _exact_ amount of bytes, or it will fail.
 
-  @result of ZSTD_decompressContinue() is the number of bytes regenerated within 'dst' (necessarily <= dstCapacity).
-  It can be zero, which is not an error; it just means ZSTD_decompressContinue() has decoded some metadata item.
+ @result of ZSTD_decompressContinue() is the number of bytes regenerated within 'dst' (necessarily <= dstCapacity).
+  It can be zero : it just means ZSTD_decompressContinue() has decoded some metadata item.
   It can also be an error code, which can be tested with ZSTD_isError().
 
-  ZSTD_decompressContinue() needs previous data blocks during decompression, up to `windowSize`.
-  They should preferably be located contiguously, prior to current block.
-  Alternatively, a round buffer of sufficient size is also possible. Sufficient size is determined by frame parameters.
-  ZSTD_decompressContinue() is very sensitive to contiguity,
-  if 2 blocks don't follow each other, make sure that either the compressor breaks contiguity at the same place,
-  or that previous contiguous segment is large enough to properly handle maximum back-reference.
-
   A frame is fully decoded when ZSTD_nextSrcSizeToDecompress() returns zero.
   Context can then be reset to start a new decompression.
 
@@ -737,28 +871,337 @@ ZSTDLIB_API size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapaci
   == Special case : skippable frames ==
 
   Skippable frames allow integration of user-defined data into a flow of concatenated frames.
-  Skippable frames will be ignored (skipped) by a decompressor. The format of skippable frames is as follows :
+  Skippable frames will be ignored (skipped) by decompressor.
+  The format of skippable frames is as follows :
   a) Skippable frame ID - 4 Bytes, Little endian format, any value from 0x184D2A50 to 0x184D2A5F
   b) Frame Size - 4 Bytes, Little endian format, unsigned 32-bits
   c) Frame Content - any content (User Data) of length equal to Frame Size
-  For skippable frames ZSTD_decompressContinue() always returns 0.
-  For skippable frames ZSTD_getFrameParams() returns fparamsPtr->windowLog==0 what means that a frame is skippable.
-    Note : If fparamsPtr->frameContentSize==0, it is ambiguous: the frame might actually be a Zstd encoded frame with no content.
-           For purposes of decompression, it is valid in both cases to skip the frame using
-           ZSTD_findFrameCompressedSize to find its size in bytes.
-  It also returns Frame Size as fparamsPtr->frameContentSize.
+  For skippable frames ZSTD_getFrameHeader() returns zfhPtr->frameType==ZSTD_skippableFrame.
+  For skippable frames ZSTD_decompressContinue() always returns 0 : it only skips the content.
 */
 
 /*=====   Buffer-less streaming decompression functions  =====*/
-ZSTDLIB_API size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize);   /**< doesn't consume input, see details below */
+typedef enum { ZSTD_frame, ZSTD_skippableFrame } ZSTD_frameType_e;
+typedef struct {
+    unsigned long long frameContentSize; /* if == ZSTD_CONTENTSIZE_UNKNOWN, it means this field is not available. 0 means "empty" */
+    unsigned long long windowSize;       /* can be very large, up to <= frameContentSize */
+    unsigned blockSizeMax;
+    ZSTD_frameType_e frameType;          /* if == ZSTD_skippableFrame, frameContentSize is the size of skippable content */
+    unsigned headerSize;
+    unsigned dictID;
+    unsigned checksumFlag;
+} ZSTD_frameHeader;
+ZSTDLIB_API size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t srcSize);   /**< doesn't consume input */
+ZSTDLIB_API size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize);  /**< when frame content size is not known, pass in frameContentSize == ZSTD_CONTENTSIZE_UNKNOWN */
+
 ZSTDLIB_API size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx);
 ZSTDLIB_API size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
-ZSTDLIB_API void   ZSTD_copyDCtx(ZSTD_DCtx* dctx, const ZSTD_DCtx* preparedDCtx);
+ZSTDLIB_API size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
+
 ZSTDLIB_API size_t ZSTD_nextSrcSizeToDecompress(ZSTD_DCtx* dctx);
 ZSTDLIB_API size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
+
+/* misc */
+ZSTDLIB_API void   ZSTD_copyDCtx(ZSTD_DCtx* dctx, const ZSTD_DCtx* preparedDCtx);
 typedef enum { ZSTDnit_frameHeader, ZSTDnit_blockHeader, ZSTDnit_block, ZSTDnit_lastBlock, ZSTDnit_checksum, ZSTDnit_skippableFrame } ZSTD_nextInputType_e;
 ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
 
+
+
+/*===   New advanced API (experimental, and compression only)  ===*/
+
+/* notes on API design :
+ *   In this proposal, parameters are pushed one by one into an existing CCtx,
+ *   and then applied on all subsequent compression jobs.
+ *   When no parameter is ever provided, CCtx is created with compression level ZSTD_CLEVEL_DEFAULT.
+ *
+ *   This API is intended to replace all others experimental API.
+ *   It can basically do all other use cases, and even new ones.
+ *   It stands a good chance to become "stable",
+ *   after a reasonable testing period.
+ */
+
+/* note on naming convention :
+ *   Initially, the API favored names like ZSTD_setCCtxParameter() .
+ *   In this proposal, convention is changed towards ZSTD_CCtx_setParameter() .
+ *   The main driver is that it identifies more clearly the target object type.
+ *   It feels clearer in light of potential variants :
+ *   ZSTD_CDict_setParameter() (rather than ZSTD_setCDictParameter())
+ *   ZSTD_DCtx_setParameter()  (rather than ZSTD_setDCtxParameter() )
+ *   Left variant feels easier to distinguish.
+ */
+
+/* note on enum design :
+ * All enum will be manually set to explicit values before reaching "stable API" status */
+
+typedef enum {
+    /* compression parameters */
+    ZSTD_p_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
+                              * Default level is ZSTD_CLEVEL_DEFAULT==3.
+                              * Special: value 0 means "do not change cLevel". */
+    ZSTD_p_windowLog,        /* Maximum allowed back-reference distance, expressed as power of 2.
+                              * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
+                              * Special: value 0 means "do not change windowLog". */
+    ZSTD_p_hashLog,          /* Size of the probe table, as a power of 2.
+                              * Resulting table size is (1 << (hashLog+2)).
+                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
+                              * Larger tables improve compression ratio of strategies <= dFast,
+                              * and improve speed of strategies > dFast.
+                              * Special: value 0 means "do not change hashLog". */
+    ZSTD_p_chainLog,         /* Size of the full-search table, as a power of 2.
+                              * Resulting table size is (1 << (chainLog+2)).
+                              * Larger tables result in better and slower compression.
+                              * This parameter is useless when using "fast" strategy.
+                              * Special: value 0 means "do not change chainLog". */
+    ZSTD_p_searchLog,        /* Number of search attempts, as a power of 2.
+                              * More attempts result in better and slower compression.
+                              * This parameter is useless when using "fast" and "dFast" strategies.
+                              * Special: value 0 means "do not change searchLog". */
+    ZSTD_p_minMatch,         /* Minimum size of searched matches (note : repCode matches can be smaller).
+                              * Larger values make faster compression and decompression, but decrease ratio.
+                              * Must be clamped between ZSTD_SEARCHLENGTH_MIN and ZSTD_SEARCHLENGTH_MAX.
+                              * Note that currently, for all strategies < btopt, effective minimum is 4.
+                              * Note that currently, for all strategies > fast, effective maximum is 6.
+                              * Special: value 0 means "do not change minMatchLength". */
+    ZSTD_p_targetLength,     /* Only useful for strategies >= btopt.
+                              * Length of Match considered "good enough" to stop search.
+                              * Larger values make compression stronger and slower.
+                              * Special: value 0 means "do not change targetLength". */
+    ZSTD_p_compressionStrategy, /* See ZSTD_strategy enum definition.
+                              * Cast selected strategy as unsigned for ZSTD_CCtx_setParameter() compatibility.
+                              * The higher the value of selected strategy, the more complex it is,
+                              * resulting in stronger and slower compression.
+                              * Special: value 0 means "do not change strategy". */
+
+    /* frame parameters */
+    ZSTD_p_contentSizeFlag=200, /* Content size is written into frame header _whenever known_ (default:1)
+                              * note that content size must be known at the beginning,
+                              * it is sent using ZSTD_CCtx_setPledgedSrcSize() */
+    ZSTD_p_checksumFlag,     /* A 32-bits checksum of content is written at end of frame (default:0) */
+    ZSTD_p_dictIDFlag,       /* When applicable, dictID of dictionary is provided in frame header (default:1) */
+
+    /* multi-threading parameters */
+    ZSTD_p_nbThreads=400,    /* Select how many threads a compression job can spawn (default:1)
+                              * More threads improve speed, but also increase memory usage.
+                              * Can only receive a value > 1 if ZSTD_MULTITHREAD is enabled.
+                              * Special: value 0 means "do not change nbThreads" */
+    ZSTD_p_jobSize,          /* Size of a compression job. Each compression job is completed in parallel.
+                              * 0 means default, which is dynamically determined based on compression parameters.
+                              * Job size must be a minimum of overlapSize, or 1 KB, whichever is largest
+                              * The minimum size is automatically and transparently enforced */
+    ZSTD_p_overlapSizeLog,   /* Size of previous input reloaded at the beginning of each job.
+                              * 0 => no overlap, 6(default) => use 1/8th of windowSize, >=9 => use full windowSize */
+
+    /* advanced parameters - may not remain available after API update */
+    ZSTD_p_forceMaxWindow=1100, /* Force back-reference distances to remain < windowSize,
+                              * even when referencing into Dictionary content (default:0) */
+    ZSTD_p_enableLongDistanceMatching=1200,  /* Enable long distance matching.
+                                         * This parameter is designed to improve the compression
+                                         * ratio for large inputs with long distance matches.
+                                         * This increases the memory usage as well as window size.
+                                         * Note: setting this parameter sets all the LDM parameters
+                                         * as well as ZSTD_p_windowLog. It should be set after
+                                         * ZSTD_p_compressionLevel and before ZSTD_p_windowLog and
+                                         * other LDM parameters. Setting the compression level
+                                         * after this parameter overrides the window log, though LDM
+                                         * will remain enabled until explicitly disabled. */
+    ZSTD_p_ldmHashLog,   /* Size of the table for long distance matching, as a power of 2.
+                          * Larger values increase memory usage and compression ratio, but decrease
+                          * compression speed.
+                          * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
+                          * (default: 20). */
+    ZSTD_p_ldmMinMatch,  /* Minimum size of searched matches for long distance matcher.
+                          * Larger/too small values usually decrease compression ratio.
+                          * Must be clamped between ZSTD_LDM_MINMATCH_MIN
+                          * and ZSTD_LDM_MINMATCH_MAX (default: 64). */
+    ZSTD_p_ldmBucketSizeLog,  /* Log size of each bucket in the LDM hash table for collision resolution.
+                               * Larger values usually improve collision resolution but may decrease
+                               * compression speed.
+                               * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX (default: 3). */
+    ZSTD_p_ldmHashEveryLog,  /* Frequency of inserting/looking up entries in the LDM hash table.
+                              * The default is MAX(0, (windowLog - ldmHashLog)) to
+                              * optimize hash table usage.
+                              * Larger values improve compression speed. Deviating far from the
+                              * default value will likely result in a decrease in compression ratio.
+                              * Must be clamped between 0 and ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN. */
+
+} ZSTD_cParameter;
+
+
+/*! ZSTD_CCtx_setParameter() :
+ *  Set one compression parameter, selected by enum ZSTD_cParameter.
+ *  Note : when `value` is an enum, cast it to unsigned for proper type checking.
+ *  @result : 0, or an error code (which can be tested with ZSTD_isError()). */
+ZSTDLIB_API size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned value);
+
+/*! ZSTD_CCtx_setPledgedSrcSize() :
+ *  Total input data size to be compressed as a single frame.
+ *  This value will be controlled at the end, and result in error if not respected.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Note 1 : 0 means zero, empty.
+ *           In order to mean "unknown content size", pass constant ZSTD_CONTENTSIZE_UNKNOWN.
+ *           Note that ZSTD_CONTENTSIZE_UNKNOWN is default value for new compression jobs.
+ *  Note 2 : If all data is provided and consumed in a single round,
+ *           this value is overriden by srcSize instead. */
+ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize);
+
+/*! ZSTD_CCtx_loadDictionary() :
+ *  Create an internal CDict from dict buffer.
+ *  Decompression will have to use same buffer.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special : Adding a NULL (or 0-size) dictionary invalidates any previous dictionary,
+ *            meaning "return to no-dictionary mode".
+ *  Note 1 : `dict` content will be copied internally. Use
+ *            ZSTD_CCtx_loadDictionary_byReference() to reference dictionary
+ *            content instead. The dictionary buffer must then outlive its
+ *            users.
+ *  Note 2 : Loading a dictionary involves building tables, which are dependent on compression parameters.
+ *           For this reason, compression parameters cannot be changed anymore after loading a dictionary.
+ *           It's also a CPU-heavy operation, with non-negligible impact on latency.
+ *  Note 3 : Dictionary will be used for all future compression jobs.
+ *           To return to "no-dictionary" situation, load a NULL dictionary
+ *  Note 5 : Use ZSTD_CCtx_loadDictionary_advanced() to select how dictionary
+ *           content will be interpreted.
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode);
+
+
+/*! ZSTD_CCtx_refCDict() :
+ *  Reference a prepared dictionary, to be used for all next compression jobs.
+ *  Note that compression parameters are enforced from within CDict,
+ *  and supercede any compression parameter previously set within CCtx.
+ *  The dictionary will remain valid for future compression jobs using same CCtx.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special : adding a NULL CDict means "return to no-dictionary mode".
+ *  Note 1 : Currently, only one dictionary can be managed.
+ *           Adding a new dictionary effectively "discards" any previous one.
+ *  Note 2 : CDict is just referenced, its lifetime must outlive CCtx.
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
+
+/*! ZSTD_CCtx_refPrefix() :
+ *  Reference a prefix (single-usage dictionary) for next compression job.
+ *  Decompression need same prefix to properly regenerate data.
+ *  Prefix is **only used once**. Tables are discarded at end of compression job.
+ *  Subsequent compression jobs will be done without prefix (if none is explicitly referenced).
+ *  If there is a need to use same prefix multiple times, consider embedding it into a ZSTD_CDict instead.
+ * @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ *  Special : Adding any prefix (including NULL) invalidates any previous prefix or dictionary
+ *  Note 1 : Prefix buffer is referenced. It must outlive compression job.
+ *  Note 2 : Referencing a prefix involves building tables, which are dependent on compression parameters.
+ *           It's a CPU-heavy operation, with non-negligible impact on latency.
+ *  Note 3 : By default, the prefix is treated as raw content
+ *           (ZSTD_dm_rawContent). Use ZSTD_CCtx_refPrefix_advanced() to alter
+ *           dictMode. */
+ZSTDLIB_API size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize);
+ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictMode_e dictMode);
+
+
+
+typedef enum {
+    ZSTD_e_continue=0, /* collect more data, encoder transparently decides when to output result, for optimal conditions */
+    ZSTD_e_flush,      /* flush any data provided so far - frame will continue, future data can still reference previous data for better compression */
+    ZSTD_e_end         /* flush any remaining data and ends current frame. Any future compression starts a new frame. */
+} ZSTD_EndDirective;
+
+/*! ZSTD_compress_generic() :
+ *  Behave about the same as ZSTD_compressStream. To note :
+ *  - Compression parameters are pushed into CCtx before starting compression, using ZSTD_CCtx_setParameter()
+ *  - Compression parameters cannot be changed once compression is started.
+ *  - *dstPos must be <= dstCapacity, *srcPos must be <= srcSize
+ *  - *dspPos and *srcPos will be updated. They are guaranteed to remain below their respective limit.
+ *  - @return provides the minimum amount of data still to flush from internal buffers
+ *            or an error code, which can be tested using ZSTD_isError().
+ *            if @return != 0, flush is not fully completed, there is some data left within internal buffers.
+ *  - after a ZSTD_e_end directive, if internal buffer is not fully flushed,
+ *            only ZSTD_e_end or ZSTD_e_flush operations are allowed.
+ *            It is necessary to fully flush internal buffers
+ *            before starting a new compression job, or changing compression parameters.
+ */
+ZSTDLIB_API size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
+                                          ZSTD_outBuffer* output,
+                                          ZSTD_inBuffer* input,
+                                          ZSTD_EndDirective endOp);
+
+/*! ZSTD_CCtx_reset() :
+ *  Return a CCtx to clean state.
+ *  Useful after an error, or to interrupt an ongoing compression job and start a new one.
+ *  Any internal data not yet flushed is cancelled.
+ *  Dictionary (if any) is dropped.
+ *  It's possible to modify compression parameters after a reset.
+ */
+ZSTDLIB_API void ZSTD_CCtx_reset(ZSTD_CCtx* cctx);   /* Not ready yet ! */
+
+
+/*! ZSTD_compress_generic_simpleArgs() :
+ *  Same as ZSTD_compress_generic(),
+ *  but using only integral types as arguments.
+ *  Argument list is larger and less expressive than ZSTD_{in,out}Buffer,
+ *  but can be helpful for binders from dynamic languages
+ *  which have troubles handling structures containing memory pointers.
+ */
+size_t ZSTD_compress_generic_simpleArgs (
+                            ZSTD_CCtx* cctx,
+                            void* dst, size_t dstCapacity, size_t* dstPos,
+                      const void* src, size_t srcSize, size_t* srcPos,
+                            ZSTD_EndDirective endOp);
+
+
+/** ZSTD_CCtx_params
+ *
+ *  - ZSTD_createCCtxParams() : Create a ZSTD_CCtx_params structure
+ *  - ZSTD_CCtxParam_setParameter() : Push parameters one by one into an
+ *  existing ZSTD_CCtx_params structure. This is similar to
+ *  ZSTD_CCtx_setParameter().
+ *  - ZSTD_CCtx_setParametersUsingCCtxParams() : Apply parameters to an existing CCtx. These
+ *  parameters will be applied to all subsequent compression jobs.
+ *  - ZSTD_compress_generic() : Do compression using the CCtx.
+ *  - ZSTD_freeCCtxParams() : Free the memory.
+ *
+ *  This can be used with ZSTD_estimateCCtxSize_opaque() for static allocation
+ *  for single-threaded compression.
+ */
+ZSTDLIB_API ZSTD_CCtx_params* ZSTD_createCCtxParams(void);
+
+/*! ZSTD_resetCCtxParams() :
+ *  Reset params to default, with the default compression level.
+ */
+ZSTDLIB_API size_t ZSTD_resetCCtxParams(ZSTD_CCtx_params* params);
+
+/*! ZSTD_initCCtxParams() :
+ *  Initializes the compression parameters of cctxParams according to
+ *  compression level. All other parameters are reset to their default values.
+ */
+ZSTDLIB_API size_t ZSTD_initCCtxParams(ZSTD_CCtx_params* cctxParams, int compressionLevel);
+
+/*! ZSTD_initCCtxParams_advanced() :
+ *  Initializes the compression and frame parameters of cctxParams according to
+ *  params. All other parameters are reset to their default values.
+ */
+ZSTDLIB_API size_t ZSTD_initCCtxParams_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
+
+ZSTDLIB_API size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
+
+/*! ZSTD_CCtxParam_setParameter() :
+ *  Similar to ZSTD_CCtx_setParameter.
+ *  Set one compression parameter, selected by enum ZSTD_cParameter.
+ *  Parameters must be applied to a ZSTD_CCtx using ZSTD_CCtx_setParametersUsingCCtxParams().
+ *  Note : when `value` is an enum, cast it to unsigned for proper type checking.
+ *  @result : 0, or an error code (which can be tested with ZSTD_isError()).
+ */
+ZSTDLIB_API size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* params, ZSTD_cParameter param, unsigned value);
+
+/*! ZSTD_CCtx_setParametersUsingCCtxParams() :
+ * Apply a set of ZSTD_CCtx_params to the compression context.
+ * This must be done before the dictionary is loaded.
+ * The pledgedSrcSize is treated as unknown.
+ * Multithreading parameters are applied only if nbThreads > 1.
+ */
+ZSTDLIB_API size_t ZSTD_CCtx_setParametersUsingCCtxParams(
+        ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params);
+
 /**
     Block functions
 
@@ -773,7 +1216,7 @@ ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
       + compression : any ZSTD_compressBegin*() variant, including with dictionary
       + decompression : any ZSTD_decompressBegin*() variant, including with dictionary
       + copyCCtx() and copyDCtx() can be used too
-    - Block size is limited, it must be <= ZSTD_getBlockSizeMax() <= ZSTD_BLOCKSIZE_ABSOLUTEMAX
+    - Block size is limited, it must be <= ZSTD_getBlockSize() <= ZSTD_BLOCKSIZE_MAX
       + If input is larger than a block size, it's necessary to split input data into multiple blocks
       + For inputs larger than a single block size, consider using the regular ZSTD_compress() instead.
         Frame metadata is not that costly, and quickly becomes negligible as source size grows larger.
@@ -786,12 +1229,13 @@ ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
         Use ZSTD_insertBlock() for such a case.
 */
 
-#define ZSTD_BLOCKSIZE_ABSOLUTEMAX (128 * 1024)   /* define, for static allocation */
+#define ZSTD_BLOCKSIZELOG_MAX 17
+#define ZSTD_BLOCKSIZE_MAX   (1<<ZSTD_BLOCKSIZELOG_MAX)   /* define, for static allocation */
 /*=====   Raw zstd block functions  =====*/
-ZSTDLIB_API size_t ZSTD_getBlockSizeMax(ZSTD_CCtx* cctx);
+ZSTDLIB_API size_t ZSTD_getBlockSize   (const ZSTD_CCtx* cctx);
 ZSTDLIB_API size_t ZSTD_compressBlock  (ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
 ZSTDLIB_API size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
-ZSTDLIB_API size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize);  /**< insert block into `dctx` history. Useful for uncompressed blocks */
+ZSTDLIB_API size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize);  /**< insert uncompressed block into `dctx` history. Useful for multi-blocks decompression */
 
 
 #endif   /* ZSTD_H_ZSTD_STATIC_LINKING_ONLY */
diff --git a/contrib/zstd/zstd_common.c b/contrib/zstd/zstd_common.c
index 8408a589a..c2041053b 100644
--- a/contrib/zstd/zstd_common.c
+++ b/contrib/zstd/zstd_common.c
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 
@@ -12,16 +13,18 @@
 /*-*************************************
 *  Dependencies
 ***************************************/
-#include <stdlib.h>         /* malloc */
+#include <stdlib.h>      /* malloc, calloc, free */
+#include <string.h>      /* memset */
 #include "error_private.h"
-#define ZSTD_STATIC_LINKING_ONLY
-#include "zstd.h"           /* declaration of ZSTD_isError, ZSTD_getErrorName, ZSTD_getErrorCode, ZSTD_getErrorString, ZSTD_versionNumber */
+#include "zstd_internal.h"
 
 
 /*-****************************************
 *  Version
 ******************************************/
-unsigned ZSTD_versionNumber (void) { return ZSTD_VERSION_NUMBER; }
+unsigned ZSTD_versionNumber(void) { return ZSTD_VERSION_NUMBER; }
+
+const char* ZSTD_versionString(void) { return ZSTD_VERSION_STRING; }
 
 
 /*-****************************************
@@ -47,27 +50,31 @@ const char* ZSTD_getErrorString(ZSTD_ErrorCode code) { return ERR_getErrorString
 /*=**************************************************************
 *  Custom allocator
 ****************************************************************/
-/* default uses stdlib */
-void* ZSTD_defaultAllocFunction(void* opaque, size_t size)
-{
-    void* address = malloc(size);
-    (void)opaque;
-    return address;
-}
-
-void ZSTD_defaultFreeFunction(void* opaque, void* address)
+void* ZSTD_malloc(size_t size, ZSTD_customMem customMem)
 {
-    (void)opaque;
-    free(address);
+    if (customMem.customAlloc)
+        return customMem.customAlloc(customMem.opaque, size);
+    return malloc(size);
 }
 
-void* ZSTD_malloc(size_t size, ZSTD_customMem customMem)
+void* ZSTD_calloc(size_t size, ZSTD_customMem customMem)
 {
-    return customMem.customAlloc(customMem.opaque, size);
+    if (customMem.customAlloc) {
+        /* calloc implemented as malloc+memset;
+         * not as efficient as calloc, but next best guess for custom malloc */
+        void* const ptr = customMem.customAlloc(customMem.opaque, size);
+        memset(ptr, 0, size);
+        return ptr;
+    }
+    return calloc(1, size);
 }
 
 void ZSTD_free(void* ptr, ZSTD_customMem customMem)
 {
-    if (ptr!=NULL)
-        customMem.customFree(customMem.opaque, ptr);
+    if (ptr!=NULL) {
+        if (customMem.customFree)
+            customMem.customFree(customMem.opaque, ptr);
+        else
+            free(ptr);
+    }
 }
diff --git a/contrib/zstd/zstd_compress.c b/contrib/zstd/zstd_compress.c
index edf4609f7..b0e9195dd 100644
--- a/contrib/zstd/zstd_compress.c
+++ b/contrib/zstd/zstd_compress.c
@@ -1,14 +1,23 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 
 /*-*************************************
+*  Tuning parameters
+***************************************/
+#ifndef ZSTD_CLEVEL_DEFAULT
+#  define ZSTD_CLEVEL_DEFAULT 3
+#endif
+
+
+/*-*************************************
 *  Dependencies
 ***************************************/
 #include <string.h>         /* memset */
@@ -17,42 +26,12 @@
 #include "fse.h"
 #define HUF_STATIC_LINKING_ONLY
 #include "huf.h"
-#include "zstd_internal.h"  /* includes zstd.h */
-
-
-/*-*************************************
-*  Debug
-***************************************/
-#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
-#  include <assert.h>
-#else
-#  define assert(condition) ((void)0)
-#endif
-
-#define ZSTD_STATIC_ASSERT(c) { enum { ZSTD_static_assert = 1/(int)(!!(c)) }; }
-
-#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
-#  include <stdio.h>
-   static unsigned g_debugLevel = ZSTD_DEBUG;
-#  define DEBUGLOG(l, ...) if (l<=g_debugLevel) { fprintf(stderr, __FILE__ ": "); fprintf(stderr, __VA_ARGS__); fprintf(stderr, " \n"); }
-#else
-#  define DEBUGLOG(l, ...)      {}    /* disabled */
-#endif
-
-
-/*-*************************************
-*  Constants
-***************************************/
-static const U32 g_searchStrength = 8;   /* control skip over incompressible data */
-#define HASH_READ_SIZE 8
-typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
-
-/* entropy tables always have same size */
-static size_t const hufCTable_size = HUF_CTABLE_SIZE(255);
-static size_t const litlengthCTable_size = FSE_CTABLE_SIZE(LLFSELog, MaxLL);
-static size_t const offcodeCTable_size = FSE_CTABLE_SIZE(OffFSELog, MaxOff);
-static size_t const matchlengthCTable_size = FSE_CTABLE_SIZE(MLFSELog, MaxML);
-static size_t const entropyScratchSpace_size = HUF_WORKSPACE_SIZE;
+#include "zstd_compress.h"
+#include "zstd_fast.h"
+#include "zstd_double_fast.h"
+#include "zstd_lazy.h"
+#include "zstd_opt.h"
+#include "zstd_ldm.h"
 
 
 /*-*************************************
@@ -79,141 +58,568 @@ static void ZSTD_resetSeqStore(seqStore_t* ssPtr)
 /*-*************************************
 *  Context memory management
 ***************************************/
-typedef enum { zcss_init, zcss_load, zcss_flush, zcss_final } ZSTD_cStreamStage;
-
-struct ZSTD_CCtx_s {
-    const BYTE* nextSrc;    /* next block here to continue on current prefix */
-    const BYTE* base;       /* All regular indexes relative to this position */
-    const BYTE* dictBase;   /* extDict indexes relative to this position */
-    U32   dictLimit;        /* below that point, need extDict */
-    U32   lowLimit;         /* below that point, no more data */
-    U32   nextToUpdate;     /* index from which to continue dictionary update */
-    U32   nextToUpdate3;    /* index from which to continue dictionary update */
-    U32   hashLog3;         /* dispatch table : larger == faster, more memory */
-    U32   loadedDictEnd;    /* index of end of dictionary */
-    U32   forceWindow;      /* force back-references to respect limit of 1<<wLog, even for dictionary */
-    U32   forceRawDict;     /* Force loading dictionary in "content-only" mode (no header analysis) */
-    ZSTD_compressionStage_e stage;
-    U32   rep[ZSTD_REP_NUM];
-    U32   repToConfirm[ZSTD_REP_NUM];
-    U32   dictID;
-    ZSTD_parameters params;
-    void* workSpace;
-    size_t workSpaceSize;
-    size_t blockSize;
-    U64 frameContentSize;
-    U64 consumedSrcSize;
-    XXH64_state_t xxhState;
-    ZSTD_customMem customMem;
-
-    seqStore_t seqStore;    /* sequences storage ptrs */
-    U32* hashTable;
-    U32* hashTable3;
-    U32* chainTable;
-    HUF_repeat hufCTable_repeatMode;
-    HUF_CElt* hufCTable;
-    U32 fseCTables_ready;
-    FSE_CTable* offcodeCTable;
-    FSE_CTable* matchlengthCTable;
-    FSE_CTable* litlengthCTable;
-    unsigned* entropyScratchSpace;
-
-    /* streaming */
-    ZSTD_CDict* cdictLocal;
-    const ZSTD_CDict* cdict;
-    char*  inBuff;
-    size_t inBuffSize;
-    size_t inToCompress;
-    size_t inBuffPos;
-    size_t inBuffTarget;
-    char*  outBuff;
-    size_t outBuffSize;
-    size_t outBuffContentSize;
-    size_t outBuffFlushedSize;
-    ZSTD_cStreamStage streamStage;
-    U32    frameEnded;
-    U64    pledgedSrcSize;
-};
+struct ZSTD_CDict_s {
+    void* dictBuffer;
+    const void* dictContent;
+    size_t dictContentSize;
+    ZSTD_CCtx* refContext;
+};  /* typedef'd to ZSTD_CDict within "zstd.h" */
 
 ZSTD_CCtx* ZSTD_createCCtx(void)
 {
-    return ZSTD_createCCtx_advanced(defaultCustomMem);
+    return ZSTD_createCCtx_advanced(ZSTD_defaultCMem);
 }
 
 ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)
 {
     ZSTD_CCtx* cctx;
 
-    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
-    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
 
-    cctx = (ZSTD_CCtx*) ZSTD_malloc(sizeof(ZSTD_CCtx), customMem);
+    cctx = (ZSTD_CCtx*) ZSTD_calloc(sizeof(ZSTD_CCtx), customMem);
     if (!cctx) return NULL;
-    memset(cctx, 0, sizeof(ZSTD_CCtx));
     cctx->customMem = customMem;
+    cctx->requestedParams.compressionLevel = ZSTD_CLEVEL_DEFAULT;
+    ZSTD_STATIC_ASSERT(zcss_init==0);
+    ZSTD_STATIC_ASSERT(ZSTD_CONTENTSIZE_UNKNOWN==(0ULL - 1));
+    return cctx;
+}
+
+ZSTD_CCtx* ZSTD_initStaticCCtx(void *workspace, size_t workspaceSize)
+{
+    ZSTD_CCtx* const cctx = (ZSTD_CCtx*) workspace;
+    if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL;  /* minimum size */
+    if ((size_t)workspace & 7) return NULL;  /* must be 8-aligned */
+    memset(workspace, 0, workspaceSize);   /* may be a bit generous, could memset be smaller ? */
+    cctx->staticSize = workspaceSize;
+    cctx->workSpace = (void*)(cctx+1);
+    cctx->workSpaceSize = workspaceSize - sizeof(ZSTD_CCtx);
+
+    /* entropy space (never moves) */
+    if (cctx->workSpaceSize < sizeof(ZSTD_entropyCTables_t)) return NULL;
+    assert(((size_t)cctx->workSpace & (sizeof(void*)-1)) == 0);   /* ensure correct alignment */
+    cctx->entropy = (ZSTD_entropyCTables_t*)cctx->workSpace;
+
     return cctx;
 }
 
 size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)
 {
     if (cctx==NULL) return 0;   /* support free on NULL */
+    if (cctx->staticSize) return ERROR(memory_allocation);   /* not compatible with static CCtx */
     ZSTD_free(cctx->workSpace, cctx->customMem);
     cctx->workSpace = NULL;
     ZSTD_freeCDict(cctx->cdictLocal);
     cctx->cdictLocal = NULL;
-    ZSTD_free(cctx->inBuff, cctx->customMem);
-    cctx->inBuff = NULL;
-    ZSTD_free(cctx->outBuff, cctx->customMem);
-    cctx->outBuff = NULL;
+#ifdef ZSTD_MULTITHREAD
+    ZSTDMT_freeCCtx(cctx->mtctx);
+    cctx->mtctx = NULL;
+#endif
     ZSTD_free(cctx, cctx->customMem);
     return 0;   /* reserved as a potential error code in the future */
 }
 
+
+static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)
+{
+#ifdef ZSTD_MULTITHREAD
+    return ZSTDMT_sizeof_CCtx(cctx->mtctx);
+#else
+    (void) cctx;
+    return 0;
+#endif
+}
+
+
 size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx)
 {
     if (cctx==NULL) return 0;   /* support sizeof on NULL */
+    DEBUGLOG(3, "sizeof(*cctx) : %u", (U32)sizeof(*cctx));
+    DEBUGLOG(3, "workSpaceSize (including streaming buffers): %u", (U32)cctx->workSpaceSize);
+    DEBUGLOG(3, "inner cdict : %u", (U32)ZSTD_sizeof_CDict(cctx->cdictLocal));
+    DEBUGLOG(3, "inner MTCTX : %u", (U32)ZSTD_sizeof_mtctx(cctx));
     return sizeof(*cctx) + cctx->workSpaceSize
            + ZSTD_sizeof_CDict(cctx->cdictLocal)
-           + cctx->outBuffSize + cctx->inBuffSize;
+           + ZSTD_sizeof_mtctx(cctx);
+}
+
+size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs)
+{
+    return ZSTD_sizeof_CCtx(zcs);  /* same object */
+}
+
+/* private API call, for dictBuilder only */
+const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx) { return &(ctx->seqStore); }
+
+#define ZSTD_CLEVEL_CUSTOM 999
+
+static ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
+        ZSTD_CCtx_params params, U64 srcSizeHint, size_t dictSize)
+{
+    return (params.compressionLevel == ZSTD_CLEVEL_CUSTOM ?
+                    params.cParams :
+                    ZSTD_getCParams(params.compressionLevel, srcSizeHint, dictSize));
+}
+
+static void ZSTD_cLevelToCCtxParams_srcSize(ZSTD_CCtx_params* params, U64 srcSize)
+{
+    params->cParams = ZSTD_getCParamsFromCCtxParams(*params, srcSize, 0);
+    params->compressionLevel = ZSTD_CLEVEL_CUSTOM;
+}
+
+static void ZSTD_cLevelToCParams(ZSTD_CCtx* cctx)
+{
+    ZSTD_cLevelToCCtxParams_srcSize(
+            &cctx->requestedParams, cctx->pledgedSrcSizePlusOne-1);
+}
+
+static void ZSTD_cLevelToCCtxParams(ZSTD_CCtx_params* params)
+{
+    ZSTD_cLevelToCCtxParams_srcSize(params, 0);
+}
+
+static ZSTD_CCtx_params ZSTD_makeCCtxParamsFromCParams(
+        ZSTD_compressionParameters cParams)
+{
+    ZSTD_CCtx_params cctxParams;
+    memset(&cctxParams, 0, sizeof(cctxParams));
+    cctxParams.cParams = cParams;
+    cctxParams.compressionLevel = ZSTD_CLEVEL_CUSTOM;
+    return cctxParams;
+}
+
+static ZSTD_CCtx_params* ZSTD_createCCtxParams_advanced(
+        ZSTD_customMem customMem)
+{
+    ZSTD_CCtx_params* params;
+    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
+    params = (ZSTD_CCtx_params*)ZSTD_calloc(
+            sizeof(ZSTD_CCtx_params), customMem);
+    if (!params) { return NULL; }
+    params->customMem = customMem;
+    params->compressionLevel = ZSTD_CLEVEL_DEFAULT;
+    return params;
+}
+
+ZSTD_CCtx_params* ZSTD_createCCtxParams(void)
+{
+    return ZSTD_createCCtxParams_advanced(ZSTD_defaultCMem);
+}
+
+size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params)
+{
+    if (params == NULL) { return 0; }
+    ZSTD_free(params, params->customMem);
+    return 0;
+}
+
+size_t ZSTD_resetCCtxParams(ZSTD_CCtx_params* params)
+{
+    return ZSTD_initCCtxParams(params, ZSTD_CLEVEL_DEFAULT);
+}
+
+size_t ZSTD_initCCtxParams(ZSTD_CCtx_params* cctxParams, int compressionLevel) {
+    if (!cctxParams) { return ERROR(GENERIC); }
+    memset(cctxParams, 0, sizeof(*cctxParams));
+    cctxParams->compressionLevel = compressionLevel;
+    return 0;
+}
+
+size_t ZSTD_initCCtxParams_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params)
+{
+    if (!cctxParams) { return ERROR(GENERIC); }
+    CHECK_F( ZSTD_checkCParams(params.cParams) );
+    memset(cctxParams, 0, sizeof(*cctxParams));
+    cctxParams->cParams = params.cParams;
+    cctxParams->fParams = params.fParams;
+    cctxParams->compressionLevel = ZSTD_CLEVEL_CUSTOM;
+    return 0;
+}
+
+static ZSTD_CCtx_params ZSTD_assignParamsToCCtxParams(
+        ZSTD_CCtx_params cctxParams, ZSTD_parameters params)
+{
+    ZSTD_CCtx_params ret = cctxParams;
+    ret.cParams = params.cParams;
+    ret.fParams = params.fParams;
+    return ret;
+}
+
+#define CLAMPCHECK(val,min,max) {            \
+    if (((val)<(min)) | ((val)>(max))) {     \
+        return ERROR(parameter_outOfBound);  \
+}   }
+
+size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned value)
+{
+    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+
+    switch(param)
+    {
+    case ZSTD_p_compressionLevel:
+        if (value == 0) return 0;  /* special value : 0 means "don't change anything" */
+        if (cctx->cdict) return ERROR(stage_wrong);
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_windowLog:
+    case ZSTD_p_hashLog:
+    case ZSTD_p_chainLog:
+    case ZSTD_p_searchLog:
+    case ZSTD_p_minMatch:
+    case ZSTD_p_targetLength:
+    case ZSTD_p_compressionStrategy:
+        if (value == 0) return 0;  /* special value : 0 means "don't change anything" */
+        if (cctx->cdict) return ERROR(stage_wrong);
+        ZSTD_cLevelToCParams(cctx);  /* Can optimize if srcSize is known */
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_contentSizeFlag:
+    case ZSTD_p_checksumFlag:
+    case ZSTD_p_dictIDFlag:
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_forceMaxWindow :  /* Force back-references to remain < windowSize,
+                                   * even when referencing into Dictionary content
+                                   * default : 0 when using a CDict, 1 when using a Prefix */
+        cctx->loadedDictEnd = 0;
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_nbThreads:
+        if (value==0) return 0;
+        DEBUGLOG(5, " setting nbThreads : %u", value);
+        if (value > 1 && cctx->staticSize) {
+            return ERROR(parameter_unsupported);  /* MT not compatible with static alloc */
+        }
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_jobSize:
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_overlapSizeLog:
+        DEBUGLOG(5, " setting overlap with nbThreads == %u", cctx->requestedParams.nbThreads);
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_enableLongDistanceMatching:
+        if (cctx->cdict) return ERROR(stage_wrong);
+        if (value != 0) {
+            ZSTD_cLevelToCParams(cctx);
+        }
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_ldmHashLog:
+    case ZSTD_p_ldmMinMatch:
+        if (value == 0) return 0;  /* special value : 0 means "don't change anything" */
+        if (cctx->cdict) return ERROR(stage_wrong);
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    case ZSTD_p_ldmBucketSizeLog:
+    case ZSTD_p_ldmHashEveryLog:
+        if (cctx->cdict) return ERROR(stage_wrong);
+        return ZSTD_CCtxParam_setParameter(&cctx->requestedParams, param, value);
+
+    default: return ERROR(parameter_unsupported);
+    }
 }
 
-size_t ZSTD_setCCtxParameter(ZSTD_CCtx* cctx, ZSTD_CCtxParameter param, unsigned value)
+size_t ZSTD_CCtxParam_setParameter(
+        ZSTD_CCtx_params* params, ZSTD_cParameter param, unsigned value)
 {
     switch(param)
     {
-    case ZSTD_p_forceWindow : cctx->forceWindow = value>0; cctx->loadedDictEnd = 0; return 0;
-    case ZSTD_p_forceRawDict : cctx->forceRawDict = value>0; return 0;
-    default: return ERROR(parameter_unknown);
+    case ZSTD_p_compressionLevel :
+        if ((int)value > ZSTD_maxCLevel()) value = ZSTD_maxCLevel();
+        if (value == 0) return 0;
+        params->compressionLevel = value;
+        return 0;
+
+    case ZSTD_p_windowLog :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.windowLog = value;
+        return 0;
+
+    case ZSTD_p_hashLog :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.hashLog = value;
+        return 0;
+
+    case ZSTD_p_chainLog :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.chainLog = value;
+        return 0;
+
+    case ZSTD_p_searchLog :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.searchLog = value;
+        return 0;
+
+    case ZSTD_p_minMatch :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_SEARCHLENGTH_MIN, ZSTD_SEARCHLENGTH_MAX);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.searchLength = value;
+        return 0;
+
+    case ZSTD_p_targetLength :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_TARGETLENGTH_MIN, ZSTD_TARGETLENGTH_MAX);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.targetLength = value;
+        return 0;
+
+    case ZSTD_p_compressionStrategy :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, (unsigned)ZSTD_fast, (unsigned)ZSTD_btultra);
+        ZSTD_cLevelToCCtxParams(params);
+        params->cParams.strategy = (ZSTD_strategy)value;
+        return 0;
+
+    case ZSTD_p_contentSizeFlag :
+        /* Content size written in frame header _when known_ (default:1) */
+        DEBUGLOG(5, "set content size flag = %u", (value>0));
+        params->fParams.contentSizeFlag = value > 0;
+        return 0;
+
+    case ZSTD_p_checksumFlag :
+        /* A 32-bits content checksum will be calculated and written at end of frame (default:0) */
+        params->fParams.checksumFlag = value > 0;
+        return 0;
+
+    case ZSTD_p_dictIDFlag : /* When applicable, dictionary's dictID is provided in frame header (default:1) */
+        DEBUGLOG(5, "set dictIDFlag = %u", (value>0));
+        params->fParams.noDictIDFlag = (value == 0);
+        return 0;
+
+    case ZSTD_p_forceMaxWindow :
+        params->forceWindow = value > 0;
+        return 0;
+
+    case ZSTD_p_nbThreads :
+        if (value == 0) return 0;
+#ifndef ZSTD_MULTITHREAD
+        if (value > 1) return ERROR(parameter_unsupported);
+        return 0;
+#else
+        return ZSTDMT_initializeCCtxParameters(params, value);
+#endif
+
+    case ZSTD_p_jobSize :
+#ifndef ZSTD_MULTITHREAD
+        return ERROR(parameter_unsupported);
+#else
+        if (params->nbThreads <= 1) return ERROR(parameter_unsupported);
+        return ZSTDMT_CCtxParam_setMTCtxParameter(params, ZSTDMT_p_sectionSize, value);
+#endif
+
+    case ZSTD_p_overlapSizeLog :
+#ifndef ZSTD_MULTITHREAD
+        return ERROR(parameter_unsupported);
+#else
+        if (params->nbThreads <= 1) return ERROR(parameter_unsupported);
+        return ZSTDMT_CCtxParam_setMTCtxParameter(params, ZSTDMT_p_overlapSectionLog, value);
+#endif
+
+    case ZSTD_p_enableLongDistanceMatching :
+        if (value != 0) {
+            ZSTD_cLevelToCCtxParams(params);
+            params->cParams.windowLog = ZSTD_LDM_WINDOW_LOG;
+        }
+        return ZSTD_ldm_initializeParameters(&params->ldmParams, value);
+
+    case ZSTD_p_ldmHashLog :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
+        params->ldmParams.hashLog = value;
+        return 0;
+
+    case ZSTD_p_ldmMinMatch :
+        if (value == 0) return 0;
+        CLAMPCHECK(value, ZSTD_LDM_MINMATCH_MIN, ZSTD_LDM_MINMATCH_MAX);
+        params->ldmParams.minMatchLength = value;
+        return 0;
+
+    case ZSTD_p_ldmBucketSizeLog :
+        if (value > ZSTD_LDM_BUCKETSIZELOG_MAX) {
+            return ERROR(parameter_outOfBound);
+        }
+        params->ldmParams.bucketSizeLog = value;
+        return 0;
+
+    case ZSTD_p_ldmHashEveryLog :
+        if (value > ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN) {
+            return ERROR(parameter_outOfBound);
+        }
+        params->ldmParams.hashEveryLog = value;
+        return 0;
+
+    default: return ERROR(parameter_unsupported);
     }
 }
 
-const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx)   /* hidden interface */
+/**
+ * This function should be updated whenever ZSTD_CCtx_params is updated.
+ * Parameters are copied manually before the dictionary is loaded.
+ * The multithreading parameters jobSize and overlapSizeLog are set only if
+ * nbThreads > 1.
+ *
+ * Pledged srcSize is treated as unknown.
+ */
+size_t ZSTD_CCtx_setParametersUsingCCtxParams(
+        ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params)
+{
+    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+    if (cctx->cdict) return ERROR(stage_wrong);
+
+    /* Assume the compression and frame parameters are validated */
+    cctx->requestedParams.cParams = params->cParams;
+    cctx->requestedParams.fParams = params->fParams;
+    cctx->requestedParams.compressionLevel = params->compressionLevel;
+
+    /* Set force window explicitly since it sets cctx->loadedDictEnd */
+    CHECK_F( ZSTD_CCtx_setParameter(
+                   cctx, ZSTD_p_forceMaxWindow, params->forceWindow) );
+
+    /* Set multithreading parameters explicitly */
+    CHECK_F( ZSTD_CCtx_setParameter(cctx, ZSTD_p_nbThreads, params->nbThreads) );
+    if (params->nbThreads > 1) {
+        CHECK_F( ZSTD_CCtx_setParameter(cctx, ZSTD_p_jobSize, params->jobSize) );
+        CHECK_F( ZSTD_CCtx_setParameter(
+                    cctx, ZSTD_p_overlapSizeLog, params->overlapSizeLog) );
+    }
+
+    /* Copy long distance matching parameters */
+    cctx->requestedParams.ldmParams = params->ldmParams;
+
+    /* customMem is used only for create/free params and can be ignored */
+    return 0;
+}
+
+ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize)
+{
+    DEBUGLOG(5, " setting pledgedSrcSize to %u", (U32)pledgedSrcSize);
+    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+    cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;
+    return 0;
+}
+
+size_t ZSTD_CCtx_loadDictionary_advanced(
+        ZSTD_CCtx* cctx, const void* dict, size_t dictSize,
+        ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode)
+{
+    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+    if (cctx->staticSize) return ERROR(memory_allocation);  /* no malloc for static CCtx */
+    DEBUGLOG(5, "load dictionary of size %u", (U32)dictSize);
+    ZSTD_freeCDict(cctx->cdictLocal);  /* in case one already exists */
+    if (dict==NULL || dictSize==0) {   /* no dictionary mode */
+        cctx->cdictLocal = NULL;
+        cctx->cdict = NULL;
+    } else {
+        ZSTD_compressionParameters const cParams =
+                ZSTD_getCParamsFromCCtxParams(cctx->requestedParams, 0, dictSize);
+        cctx->cdictLocal = ZSTD_createCDict_advanced(
+                                dict, dictSize,
+                                dictLoadMethod, dictMode,
+                                cParams, cctx->customMem);
+        cctx->cdict = cctx->cdictLocal;
+        if (cctx->cdictLocal == NULL)
+            return ERROR(memory_allocation);
+    }
+    return 0;
+}
+
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference(
+      ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
 {
-    return &(ctx->seqStore);
+    return ZSTD_CCtx_loadDictionary_advanced(
+            cctx, dict, dictSize, ZSTD_dlm_byRef, ZSTD_dm_auto);
 }
 
-static ZSTD_parameters ZSTD_getParamsFromCCtx(const ZSTD_CCtx* cctx)
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
 {
-    return cctx->params;
+    return ZSTD_CCtx_loadDictionary_advanced(
+            cctx, dict, dictSize, ZSTD_dlm_byCopy, ZSTD_dm_auto);
 }
 
 
-/** ZSTD_checkParams() :
-    ensure param values remain within authorized range.
+size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
+{
+    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+    cctx->cdict = cdict;
+    memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));  /* exclusive */
+    return 0;
+}
+
+size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize)
+{
+    return ZSTD_CCtx_refPrefix_advanced(cctx, prefix, prefixSize, ZSTD_dm_rawContent);
+}
+
+size_t ZSTD_CCtx_refPrefix_advanced(
+        ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictMode_e dictMode)
+{
+    if (cctx->streamStage != zcss_init) return ERROR(stage_wrong);
+    cctx->cdict = NULL;   /* prefix discards any prior cdict */
+    cctx->prefixDict.dict = prefix;
+    cctx->prefixDict.dictSize = prefixSize;
+    cctx->prefixDict.dictMode = dictMode;
+    return 0;
+}
+
+static void ZSTD_startNewCompression(ZSTD_CCtx* cctx)
+{
+    cctx->streamStage = zcss_init;
+    cctx->pledgedSrcSizePlusOne = 0;
+}
+
+/*! ZSTD_CCtx_reset() :
+ *  Also dumps dictionary */
+void ZSTD_CCtx_reset(ZSTD_CCtx* cctx)
+{
+    ZSTD_startNewCompression(cctx);
+    cctx->cdict = NULL;
+}
+
+/** ZSTD_checkCParams() :
+    control CParam values remain within authorized range.
     @return : 0, or an error code if one value is beyond authorized range */
 size_t ZSTD_checkCParams(ZSTD_compressionParameters cParams)
 {
-#   define CLAMPCHECK(val,min,max) { if ((val<min) | (val>max)) return ERROR(compressionParameter_unsupported); }
     CLAMPCHECK(cParams.windowLog, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
     CLAMPCHECK(cParams.chainLog, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
     CLAMPCHECK(cParams.hashLog, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
     CLAMPCHECK(cParams.searchLog, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
     CLAMPCHECK(cParams.searchLength, ZSTD_SEARCHLENGTH_MIN, ZSTD_SEARCHLENGTH_MAX);
     CLAMPCHECK(cParams.targetLength, ZSTD_TARGETLENGTH_MIN, ZSTD_TARGETLENGTH_MAX);
-    if ((U32)(cParams.strategy) > (U32)ZSTD_btultra) return ERROR(compressionParameter_unsupported);
+    if ((U32)(cParams.strategy) > (U32)ZSTD_btultra)
+        return ERROR(parameter_unsupported);
     return 0;
 }
 
+/** ZSTD_clampCParams() :
+ *  make CParam values within valid range.
+ *  @return : valid CParams */
+static ZSTD_compressionParameters ZSTD_clampCParams(ZSTD_compressionParameters cParams)
+{
+#   define CLAMP(val,min,max) {      \
+        if (val<min) val=min;        \
+        else if (val>max) val=max;   \
+    }
+    CLAMP(cParams.windowLog, ZSTD_WINDOWLOG_MIN, ZSTD_WINDOWLOG_MAX);
+    CLAMP(cParams.chainLog, ZSTD_CHAINLOG_MIN, ZSTD_CHAINLOG_MAX);
+    CLAMP(cParams.hashLog, ZSTD_HASHLOG_MIN, ZSTD_HASHLOG_MAX);
+    CLAMP(cParams.searchLog, ZSTD_SEARCHLOG_MIN, ZSTD_SEARCHLOG_MAX);
+    CLAMP(cParams.searchLength, ZSTD_SEARCHLENGTH_MIN, ZSTD_SEARCHLENGTH_MAX);
+    CLAMP(cParams.targetLength, ZSTD_TARGETLENGTH_MIN, ZSTD_TARGETLENGTH_MAX);
+    if ((U32)(cParams.strategy) > (U32)ZSTD_btultra) cParams.strategy = ZSTD_btultra;
+    return cParams;
+}
 
 /** ZSTD_cycleLog() :
  *  condition for correct operation : hashLog > 1 */
@@ -223,21 +629,23 @@ static U32 ZSTD_cycleLog(U32 hashLog, ZSTD_strategy strat)
     return hashLog - btScale;
 }
 
-/** ZSTD_adjustCParams() :
+/** ZSTD_adjustCParams_internal() :
     optimize `cPar` for a given input (`srcSize` and `dictSize`).
     mostly downsizing to reduce memory consumption and initialization.
     Both `srcSize` and `dictSize` are optional (use 0 if unknown),
     but if both are 0, no optimization can be done.
     Note : cPar is considered validated at this stage. Use ZSTD_checkParams() to ensure that. */
-ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize)
+ZSTD_compressionParameters ZSTD_adjustCParams_internal(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize)
 {
+    assert(ZSTD_checkCParams(cPar)==0);
     if (srcSize+dictSize == 0) return cPar;   /* no size information available : no adjustment */
 
     /* resize params, to use less memory when necessary */
     {   U32 const minSrcSize = (srcSize==0) ? 500 : 0;
         U64 const rSize = srcSize + dictSize + minSrcSize;
         if (rSize < ((U64)1<<ZSTD_WINDOWLOG_MAX)) {
-            U32 const srcLog = MAX(ZSTD_HASHLOG_MIN, ZSTD_highbit32((U32)(rSize)-1) + 1);
+            U32 const srcLog =
+                    MAX(ZSTD_HASHLOG_MIN, (rSize==1) ? 1 : ZSTD_highbit32((U32)(rSize)-1) + 1);
             if (cPar.windowLog > srcLog) cPar.windowLog = srcLog;
     }   }
     if (cPar.hashLog > cPar.windowLog) cPar.hashLog = cPar.windowLog;
@@ -250,100 +658,212 @@ ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, u
     return cPar;
 }
 
+ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize)
+{
+    cPar = ZSTD_clampCParams(cPar);
+    return ZSTD_adjustCParams_internal(cPar, srcSize, dictSize);
+}
 
-size_t ZSTD_estimateCCtxSize(ZSTD_compressionParameters cParams)
+size_t ZSTD_estimateCCtxSize_advanced_usingCCtxParams(const ZSTD_CCtx_params* params)
 {
-    size_t const blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, (size_t)1 << cParams.windowLog);
-    U32    const divider = (cParams.searchLength==3) ? 3 : 4;
-    size_t const maxNbSeq = blockSize / divider;
-    size_t const tokenSpace = blockSize + 11*maxNbSeq;
+    /* Estimate CCtx size is supported for single-threaded compression only. */
+    if (params->nbThreads > 1) { return ERROR(GENERIC); }
+    {   ZSTD_compressionParameters const cParams =
+                ZSTD_getCParamsFromCCtxParams(*params, 0, 0);
+        size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
+        U32    const divider = (cParams.searchLength==3) ? 3 : 4;
+        size_t const maxNbSeq = blockSize / divider;
+        size_t const tokenSpace = blockSize + 11*maxNbSeq;
+        size_t const chainSize =
+                (cParams.strategy == ZSTD_fast) ? 0 : (1 << cParams.chainLog);
+        size_t const hSize = ((size_t)1) << cParams.hashLog;
+        U32    const hashLog3 = (cParams.searchLength>3) ?
+                                0 : MIN(ZSTD_HASHLOG3_MAX, cParams.windowLog);
+        size_t const h3Size = ((size_t)1) << hashLog3;
+        size_t const entropySpace = sizeof(ZSTD_entropyCTables_t);
+        size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
 
-    size_t const chainSize = (cParams.strategy == ZSTD_fast) ? 0 : (1 << cParams.chainLog);
-    size_t const hSize = ((size_t)1) << cParams.hashLog;
-    U32    const hashLog3 = (cParams.searchLength>3) ? 0 : MIN(ZSTD_HASHLOG3_MAX, cParams.windowLog);
-    size_t const h3Size = ((size_t)1) << hashLog3;
-    size_t const entropySpace = hufCTable_size + litlengthCTable_size
-                              + offcodeCTable_size + matchlengthCTable_size
-                              + entropyScratchSpace_size;
-    size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
+        size_t const optBudget =
+                ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<<Litbits))*sizeof(U32)
+                + (ZSTD_OPT_NUM+1)*(sizeof(ZSTD_match_t) + sizeof(ZSTD_optimal_t));
+        size_t const optSpace = ((cParams.strategy == ZSTD_btopt) || (cParams.strategy == ZSTD_btultra)) ? optBudget : 0;
 
-    size_t const optBudget = ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<<Litbits))*sizeof(U32)
-                          + (ZSTD_OPT_NUM+1)*(sizeof(ZSTD_match_t) + sizeof(ZSTD_optimal_t));
-    size_t const optSpace = ((cParams.strategy == ZSTD_btopt) || (cParams.strategy == ZSTD_btultra)) ? optBudget : 0;
-    size_t const neededSpace = entropySpace + tableSpace + tokenSpace + optSpace;
+        size_t const ldmSpace = params->ldmParams.enableLdm ?
+            ZSTD_ldm_getTableSize(params->ldmParams.hashLog,
+                                  params->ldmParams.bucketSizeLog) : 0;
 
-    return sizeof(ZSTD_CCtx) + neededSpace;
+        size_t const neededSpace = entropySpace + tableSpace + tokenSpace +
+                                   optSpace + ldmSpace;
+
+        DEBUGLOG(5, "sizeof(ZSTD_CCtx) : %u", (U32)sizeof(ZSTD_CCtx));
+        DEBUGLOG(5, "estimate workSpace : %u", (U32)neededSpace);
+        return sizeof(ZSTD_CCtx) + neededSpace;
+    }
+}
+
+size_t ZSTD_estimateCCtxSize_advanced_usingCParams(ZSTD_compressionParameters cParams)
+{
+    ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams);
+    return ZSTD_estimateCCtxSize_advanced_usingCCtxParams(&params);
+}
+
+size_t ZSTD_estimateCCtxSize(int compressionLevel)
+{
+    ZSTD_compressionParameters const cParams = ZSTD_getCParams(compressionLevel, 0, 0);
+    return ZSTD_estimateCCtxSize_advanced_usingCParams(cParams);
+}
+
+size_t ZSTD_estimateCStreamSize_advanced_usingCCtxParams(const ZSTD_CCtx_params* params)
+{
+    if (params->nbThreads > 1) { return ERROR(GENERIC); }
+    {   size_t const CCtxSize = ZSTD_estimateCCtxSize_advanced_usingCCtxParams(params);
+        size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << params->cParams.windowLog);
+        size_t const inBuffSize = ((size_t)1 << params->cParams.windowLog) + blockSize;
+        size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1;
+        size_t const streamingSize = inBuffSize + outBuffSize;
+
+        return CCtxSize + streamingSize;
+    }
+}
+
+size_t ZSTD_estimateCStreamSize_advanced_usingCParams(ZSTD_compressionParameters cParams)
+{
+    ZSTD_CCtx_params const params = ZSTD_makeCCtxParamsFromCParams(cParams);
+    return ZSTD_estimateCStreamSize_advanced_usingCCtxParams(&params);
+}
+
+size_t ZSTD_estimateCStreamSize(int compressionLevel) {
+    ZSTD_compressionParameters const cParams = ZSTD_getCParams(compressionLevel, 0, 0);
+    return ZSTD_estimateCStreamSize_advanced_usingCParams(cParams);
+}
+
+static U32 ZSTD_equivalentCParams(ZSTD_compressionParameters cParams1,
+                                  ZSTD_compressionParameters cParams2)
+{
+    U32 bslog1 = MIN(cParams1.windowLog, ZSTD_BLOCKSIZELOG_MAX);
+    U32 bslog2 = MIN(cParams2.windowLog, ZSTD_BLOCKSIZELOG_MAX);
+    return (bslog1 == bslog2)   /* same block size */
+         & (cParams1.hashLog  == cParams2.hashLog)
+         & (cParams1.chainLog == cParams2.chainLog)
+         & (cParams1.strategy == cParams2.strategy)   /* opt parser space */
+         & ((cParams1.searchLength==3) == (cParams2.searchLength==3));  /* hashlog3 space */
 }
 
+/** The parameters are equivalent if ldm is not enabled in both sets or
+ *  all the parameters are equivalent. */
+static U32 ZSTD_equivalentLdmParams(ldmParams_t ldmParams1,
+                                    ldmParams_t ldmParams2)
+{
+    return (!ldmParams1.enableLdm && !ldmParams2.enableLdm) ||
+           (ldmParams1.enableLdm == ldmParams2.enableLdm &&
+            ldmParams1.hashLog == ldmParams2.hashLog &&
+            ldmParams1.bucketSizeLog == ldmParams2.bucketSizeLog &&
+            ldmParams1.minMatchLength == ldmParams2.minMatchLength &&
+            ldmParams1.hashEveryLog == ldmParams2.hashEveryLog);
+}
 
-static U32 ZSTD_equivalentParams(ZSTD_parameters param1, ZSTD_parameters param2)
+/** Equivalence for resetCCtx purposes */
+static U32 ZSTD_equivalentParams(ZSTD_CCtx_params params1,
+                                 ZSTD_CCtx_params params2)
 {
-    return (param1.cParams.hashLog  == param2.cParams.hashLog)
-         & (param1.cParams.chainLog == param2.cParams.chainLog)
-         & (param1.cParams.strategy == param2.cParams.strategy)   /* opt parser space */
-         & ((param1.cParams.searchLength==3) == (param2.cParams.searchLength==3));  /* hashlog3 space */
+    return ZSTD_equivalentCParams(params1.cParams, params2.cParams) &&
+           ZSTD_equivalentLdmParams(params1.ldmParams, params2.ldmParams);
 }
 
 /*! ZSTD_continueCCtx() :
-    reuse CCtx without reset (note : requires no dictionary) */
-static size_t ZSTD_continueCCtx(ZSTD_CCtx* cctx, ZSTD_parameters params, U64 frameContentSize)
+ *  reuse CCtx without reset (note : requires no dictionary) */
+static size_t ZSTD_continueCCtx(ZSTD_CCtx* cctx, ZSTD_CCtx_params params, U64 pledgedSrcSize)
 {
     U32 const end = (U32)(cctx->nextSrc - cctx->base);
-    cctx->params = params;
-    cctx->frameContentSize = frameContentSize;
+    DEBUGLOG(5, "continue mode");
+    cctx->appliedParams = params;
+    cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;
     cctx->consumedSrcSize = 0;
+    if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)
+        cctx->appliedParams.fParams.contentSizeFlag = 0;
+    DEBUGLOG(5, "pledged content size : %u ; flag : %u",
+        (U32)pledgedSrcSize, cctx->appliedParams.fParams.contentSizeFlag);
     cctx->lowLimit = end;
     cctx->dictLimit = end;
     cctx->nextToUpdate = end+1;
     cctx->stage = ZSTDcs_init;
     cctx->dictID = 0;
     cctx->loadedDictEnd = 0;
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) cctx->rep[i] = repStartValue[i]; }
-    cctx->seqStore.litLengthSum = 0;  /* force reset of btopt stats */
+    { int i; for (i=0; i<ZSTD_REP_NUM; i++) cctx->seqStore.rep[i] = repStartValue[i]; }
+    cctx->optState.litLengthSum = 0;  /* force reset of btopt stats */
     XXH64_reset(&cctx->xxhState, 0);
     return 0;
 }
 
-typedef enum { ZSTDcrp_continue, ZSTDcrp_noMemset, ZSTDcrp_fullReset } ZSTD_compResetPolicy_e;
+typedef enum { ZSTDcrp_continue, ZSTDcrp_noMemset } ZSTD_compResetPolicy_e;
+typedef enum { ZSTDb_not_buffered, ZSTDb_buffered } ZSTD_buffered_policy_e;
 
 /*! ZSTD_resetCCtx_internal() :
-    note : `params` must be validated */
-static size_t ZSTD_resetCCtx_internal (ZSTD_CCtx* zc,
-                                       ZSTD_parameters params, U64 frameContentSize,
-                                       ZSTD_compResetPolicy_e const crp)
-{
-    DEBUGLOG(5, "ZSTD_resetCCtx_internal \n");
-
-    if (crp == ZSTDcrp_continue)
-        if (ZSTD_equivalentParams(params, zc->params)) {
-            DEBUGLOG(5, "ZSTD_equivalentParams()==1 \n");
-            zc->fseCTables_ready = 0;
-            zc->hufCTable_repeatMode = HUF_repeat_none;
-            return ZSTD_continueCCtx(zc, params, frameContentSize);
-        }
+    note : `params` are assumed fully validated at this stage */
+static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
+                                      ZSTD_CCtx_params params, U64 pledgedSrcSize,
+                                      ZSTD_compResetPolicy_e const crp,
+                                      ZSTD_buffered_policy_e const zbuff)
+{
+    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+
+    if (crp == ZSTDcrp_continue) {
+        if (ZSTD_equivalentParams(params, zc->appliedParams)) {
+            DEBUGLOG(5, "ZSTD_equivalentParams()==1");
+            assert(!(params.ldmParams.enableLdm &&
+                     params.ldmParams.hashEveryLog == ZSTD_LDM_HASHEVERYLOG_NOTSET));
+            zc->entropy->hufCTable_repeatMode = HUF_repeat_none;
+            zc->entropy->offcode_repeatMode = FSE_repeat_none;
+            zc->entropy->matchlength_repeatMode = FSE_repeat_none;
+            zc->entropy->litlength_repeatMode = FSE_repeat_none;
+            return ZSTD_continueCCtx(zc, params, pledgedSrcSize);
+    }   }
 
-    {   size_t const blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, (size_t)1 << params.cParams.windowLog);
+    if (params.ldmParams.enableLdm) {
+        /* Adjust long distance matching parameters */
+        ZSTD_ldm_adjustParameters(&params.ldmParams, params.cParams.windowLog);
+        assert(params.ldmParams.hashLog >= params.ldmParams.bucketSizeLog);
+        assert(params.ldmParams.hashEveryLog < 32);
+        zc->ldmState.hashPower =
+                ZSTD_ldm_getHashPower(params.ldmParams.minMatchLength);
+    }
+
+    {   size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << params.cParams.windowLog);
         U32    const divider = (params.cParams.searchLength==3) ? 3 : 4;
         size_t const maxNbSeq = blockSize / divider;
         size_t const tokenSpace = blockSize + 11*maxNbSeq;
-        size_t const chainSize = (params.cParams.strategy == ZSTD_fast) ? 0 : (1 << params.cParams.chainLog);
+        size_t const chainSize = (params.cParams.strategy == ZSTD_fast) ?
+                                0 : (1 << params.cParams.chainLog);
         size_t const hSize = ((size_t)1) << params.cParams.hashLog;
-        U32    const hashLog3 = (params.cParams.searchLength>3) ? 0 : MIN(ZSTD_HASHLOG3_MAX, params.cParams.windowLog);
+        U32    const hashLog3 = (params.cParams.searchLength>3) ?
+                                0 : MIN(ZSTD_HASHLOG3_MAX, params.cParams.windowLog);
         size_t const h3Size = ((size_t)1) << hashLog3;
         size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
+        size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0;
+        size_t const buffInSize = (zbuff==ZSTDb_buffered) ? ((size_t)1 << params.cParams.windowLog) + blockSize : 0;
         void* ptr;
 
         /* Check if workSpace is large enough, alloc a new one if needed */
-        {   size_t const entropySpace = hufCTable_size + litlengthCTable_size
-                                  + offcodeCTable_size + matchlengthCTable_size
-                                  + entropyScratchSpace_size;
+        {   size_t const entropySpace = sizeof(ZSTD_entropyCTables_t);
             size_t const optPotentialSpace = ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<<Litbits)) * sizeof(U32)
                                   + (ZSTD_OPT_NUM+1) * (sizeof(ZSTD_match_t)+sizeof(ZSTD_optimal_t));
-            size_t const optSpace = ((params.cParams.strategy == ZSTD_btopt) || (params.cParams.strategy == ZSTD_btultra)) ? optPotentialSpace : 0;
-            size_t const neededSpace = entropySpace + optSpace + tableSpace + tokenSpace;
-            if (zc->workSpaceSize < neededSpace) {
+            size_t const optSpace = ( (params.cParams.strategy == ZSTD_btopt)
+                                    || (params.cParams.strategy == ZSTD_btultra)) ?
+                                    optPotentialSpace : 0;
+            size_t const bufferSpace = buffInSize + buffOutSize;
+            size_t const ldmSpace = params.ldmParams.enableLdm
+                ? ZSTD_ldm_getTableSize(params.ldmParams.hashLog, params.ldmParams.bucketSizeLog)
+                : 0;
+            size_t const neededSpace = entropySpace + optSpace + ldmSpace +
+                                       tableSpace + tokenSpace + bufferSpace;
+
+            if (zc->workSpaceSize < neededSpace) {  /* too small : resize */
                 DEBUGLOG(5, "Need to update workSpaceSize from %uK to %uK \n",
-                            (unsigned)zc->workSpaceSize>>10, (unsigned)neededSpace>>10);
+                            (unsigned)zc->workSpaceSize>>10,
+                            (unsigned)neededSpace>>10);
+                /* static cctx : no resize, error out */
+                if (zc->staticSize) return ERROR(memory_allocation);
+
                 zc->workSpaceSize = 0;
                 ZSTD_free(zc->workSpace, zc->customMem);
                 zc->workSpace = ZSTD_malloc(neededSpace, zc->customMem);
@@ -352,62 +872,64 @@ static size_t ZSTD_resetCCtx_internal (ZSTD_CCtx* zc,
                 ptr = zc->workSpace;
 
                 /* entropy space */
-                zc->hufCTable = (HUF_CElt*)ptr;
-                ptr = (char*)zc->hufCTable + hufCTable_size;  /* note : HUF_CElt* is incomplete type, size is estimated via macro */
-                zc->offcodeCTable = (FSE_CTable*) ptr;
-                ptr = (char*)ptr + offcodeCTable_size;
-                zc->matchlengthCTable = (FSE_CTable*) ptr;
-                ptr = (char*)ptr + matchlengthCTable_size;
-                zc->litlengthCTable = (FSE_CTable*) ptr;
-                ptr = (char*)ptr + litlengthCTable_size;
-                assert(((size_t)ptr & 3) == 0);   /* ensure correct alignment */
-                zc->entropyScratchSpace = (unsigned*) ptr;
+                assert(((size_t)zc->workSpace & 3) == 0);   /* ensure correct alignment */
+                assert(zc->workSpaceSize >= sizeof(ZSTD_entropyCTables_t));
+                zc->entropy = (ZSTD_entropyCTables_t*)zc->workSpace;
         }   }
 
         /* init params */
-        zc->params = params;
-        zc->blockSize = blockSize;
-        DEBUGLOG(5, "blockSize = %uK \n", (U32)blockSize>>10);
-        zc->frameContentSize = frameContentSize;
+        zc->appliedParams = params;
+        zc->pledgedSrcSizePlusOne = pledgedSrcSize+1;
         zc->consumedSrcSize = 0;
+        if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)
+            zc->appliedParams.fParams.contentSizeFlag = 0;
+        DEBUGLOG(5, "pledged content size : %u ; flag : %u",
+            (U32)pledgedSrcSize, zc->appliedParams.fParams.contentSizeFlag);
+        zc->blockSize = blockSize;
 
         XXH64_reset(&zc->xxhState, 0);
         zc->stage = ZSTDcs_init;
         zc->dictID = 0;
         zc->loadedDictEnd = 0;
-        zc->fseCTables_ready = 0;
-        zc->hufCTable_repeatMode = HUF_repeat_none;
+        zc->entropy->hufCTable_repeatMode = HUF_repeat_none;
+        zc->entropy->offcode_repeatMode = FSE_repeat_none;
+        zc->entropy->matchlength_repeatMode = FSE_repeat_none;
+        zc->entropy->litlength_repeatMode = FSE_repeat_none;
         zc->nextToUpdate = 1;
         zc->nextSrc = NULL;
         zc->base = NULL;
         zc->dictBase = NULL;
         zc->dictLimit = 0;
         zc->lowLimit = 0;
-        { int i; for (i=0; i<ZSTD_REP_NUM; i++) zc->rep[i] = repStartValue[i]; }
+        { int i; for (i=0; i<ZSTD_REP_NUM; i++) zc->seqStore.rep[i] = repStartValue[i]; }
         zc->hashLog3 = hashLog3;
-        zc->seqStore.litLengthSum = 0;
+        zc->optState.litLengthSum = 0;
 
-        /* ensure entropy tables are close together at the beginning */
-        assert((void*)zc->hufCTable == zc->workSpace);
-        assert((char*)zc->offcodeCTable == (char*)zc->hufCTable + hufCTable_size);
-        assert((char*)zc->matchlengthCTable == (char*)zc->offcodeCTable + offcodeCTable_size);
-        assert((char*)zc->litlengthCTable == (char*)zc->matchlengthCTable + matchlengthCTable_size);
-        assert((char*)zc->entropyScratchSpace == (char*)zc->litlengthCTable + litlengthCTable_size);
-        ptr = (char*)zc->entropyScratchSpace + entropyScratchSpace_size;
+        ptr = zc->entropy + 1;
 
         /* opt parser space */
         if ((params.cParams.strategy == ZSTD_btopt) || (params.cParams.strategy == ZSTD_btultra)) {
-            DEBUGLOG(5, "reserving optimal parser space ");
+            DEBUGLOG(5, "reserving optimal parser space");
             assert(((size_t)ptr & 3) == 0);  /* ensure ptr is properly aligned */
-            zc->seqStore.litFreq = (U32*)ptr;
-            zc->seqStore.litLengthFreq = zc->seqStore.litFreq + (1<<Litbits);
-            zc->seqStore.matchLengthFreq = zc->seqStore.litLengthFreq + (MaxLL+1);
-            zc->seqStore.offCodeFreq = zc->seqStore.matchLengthFreq + (MaxML+1);
-            ptr = zc->seqStore.offCodeFreq + (MaxOff+1);
-            zc->seqStore.matchTable = (ZSTD_match_t*)ptr;
-            ptr = zc->seqStore.matchTable + ZSTD_OPT_NUM+1;
-            zc->seqStore.priceTable = (ZSTD_optimal_t*)ptr;
-            ptr = zc->seqStore.priceTable + ZSTD_OPT_NUM+1;
+            zc->optState.litFreq = (U32*)ptr;
+            zc->optState.litLengthFreq = zc->optState.litFreq + (1<<Litbits);
+            zc->optState.matchLengthFreq = zc->optState.litLengthFreq + (MaxLL+1);
+            zc->optState.offCodeFreq = zc->optState.matchLengthFreq + (MaxML+1);
+            ptr = zc->optState.offCodeFreq + (MaxOff+1);
+            zc->optState.matchTable = (ZSTD_match_t*)ptr;
+            ptr = zc->optState.matchTable + ZSTD_OPT_NUM+1;
+            zc->optState.priceTable = (ZSTD_optimal_t*)ptr;
+            ptr = zc->optState.priceTable + ZSTD_OPT_NUM+1;
+        }
+
+        /* ldm hash table */
+        /* initialize bucketOffsets table later for pointer alignment */
+        if (params.ldmParams.enableLdm) {
+            size_t const ldmHSize = ((size_t)1) << params.ldmParams.hashLog;
+            memset(ptr, 0, ldmHSize * sizeof(ldmEntry_t));
+            assert(((size_t)ptr & 3) == 0); /* ensure ptr is properly aligned */
+            zc->ldmState.hashTable = (ldmEntry_t*)ptr;
+            ptr = zc->ldmState.hashTable + ldmHSize;
         }
 
         /* table Space */
@@ -425,6 +947,23 @@ static size_t ZSTD_resetCCtx_internal (ZSTD_CCtx* zc,
         zc->seqStore.mlCode = zc->seqStore.llCode + maxNbSeq;
         zc->seqStore.ofCode = zc->seqStore.mlCode + maxNbSeq;
         zc->seqStore.litStart = zc->seqStore.ofCode + maxNbSeq;
+        ptr = zc->seqStore.litStart + blockSize;
+
+        /* ldm bucketOffsets table */
+        if (params.ldmParams.enableLdm) {
+            size_t const ldmBucketSize =
+                  ((size_t)1) << (params.ldmParams.hashLog -
+                                  params.ldmParams.bucketSizeLog);
+            memset(ptr, 0, ldmBucketSize);
+            zc->ldmState.bucketOffsets = (BYTE*)ptr;
+            ptr = zc->ldmState.bucketOffsets + ldmBucketSize;
+        }
+
+        /* buffers */
+        zc->inBuffSize = buffInSize;
+        zc->inBuff = (char*)ptr;
+        zc->outBuffSize = buffOutSize;
+        zc->outBuff = zc->inBuff + buffInSize;
 
         return 0;
     }
@@ -436,31 +975,38 @@ static size_t ZSTD_resetCCtx_internal (ZSTD_CCtx* zc,
  *        do not use with extDict variant ! */
 void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx) {
     int i;
-    for (i=0; i<ZSTD_REP_NUM; i++) cctx->rep[i] = 0;
+    for (i=0; i<ZSTD_REP_NUM; i++) cctx->seqStore.rep[i] = 0;
 }
 
 
 /*! ZSTD_copyCCtx_internal() :
  *  Duplicate an existing context `srcCCtx` into another one `dstCCtx`.
+ *  The "context", in this case, refers to the hash and chain tables, entropy
+ *  tables, and dictionary offsets.
  *  Only works during stage ZSTDcs_init (i.e. after creation, but before first call to ZSTD_compressContinue()).
  *  pledgedSrcSize=0 means "empty" if fParams.contentSizeFlag=1
  *  @return : 0, or an error code */
-size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx,
-                              ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize)
+static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
+                            const ZSTD_CCtx* srcCCtx,
+                            ZSTD_frameParameters fParams,
+                            unsigned long long pledgedSrcSize,
+                            ZSTD_buffered_policy_e zbuff)
 {
-    DEBUGLOG(5, "ZSTD_copyCCtx_internal \n");
+    DEBUGLOG(5, "ZSTD_copyCCtx_internal");
     if (srcCCtx->stage!=ZSTDcs_init) return ERROR(stage_wrong);
 
     memcpy(&dstCCtx->customMem, &srcCCtx->customMem, sizeof(ZSTD_customMem));
-    {   ZSTD_parameters params = srcCCtx->params;
+    {   ZSTD_CCtx_params params = dstCCtx->requestedParams;
+        /* Copy only compression parameters related to tables. */
+        params.cParams = srcCCtx->appliedParams.cParams;
         params.fParams = fParams;
-        DEBUGLOG(5, "ZSTD_resetCCtx_internal : dictIDFlag : %u \n", !fParams.noDictIDFlag);
-        ZSTD_resetCCtx_internal(dstCCtx, params, pledgedSrcSize, ZSTDcrp_noMemset);
+        ZSTD_resetCCtx_internal(dstCCtx, params, pledgedSrcSize,
+                                ZSTDcrp_noMemset, zbuff);
     }
 
     /* copy tables */
-    {   size_t const chainSize = (srcCCtx->params.cParams.strategy == ZSTD_fast) ? 0 : (1 << srcCCtx->params.cParams.chainLog);
-        size_t const hSize =  (size_t)1 << srcCCtx->params.cParams.hashLog;
+    {   size_t const chainSize = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : (1 << srcCCtx->appliedParams.cParams.chainLog);
+        size_t const hSize =  (size_t)1 << srcCCtx->appliedParams.cParams.hashLog;
         size_t const h3Size = (size_t)1 << srcCCtx->hashLog3;
         size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
         assert((U32*)dstCCtx->chainTable == (U32*)dstCCtx->hashTable + hSize);  /* chainTable must follow hashTable */
@@ -480,16 +1026,7 @@ size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx,
     dstCCtx->dictID       = srcCCtx->dictID;
 
     /* copy entropy tables */
-    dstCCtx->fseCTables_ready = srcCCtx->fseCTables_ready;
-    if (srcCCtx->fseCTables_ready) {
-        memcpy(dstCCtx->litlengthCTable, srcCCtx->litlengthCTable, litlengthCTable_size);
-        memcpy(dstCCtx->matchlengthCTable, srcCCtx->matchlengthCTable, matchlengthCTable_size);
-        memcpy(dstCCtx->offcodeCTable, srcCCtx->offcodeCTable, offcodeCTable_size);
-    }
-    dstCCtx->hufCTable_repeatMode = srcCCtx->hufCTable_repeatMode;
-    if (srcCCtx->hufCTable_repeatMode) {
-        memcpy(dstCCtx->hufCTable, srcCCtx->hufCTable, hufCTable_size);
-    }
+    memcpy(dstCCtx->entropy, srcCCtx->entropy, sizeof(ZSTD_entropyCTables_t));
 
     return 0;
 }
@@ -502,9 +1039,11 @@ size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx,
 size_t ZSTD_copyCCtx(ZSTD_CCtx* dstCCtx, const ZSTD_CCtx* srcCCtx, unsigned long long pledgedSrcSize)
 {
     ZSTD_frameParameters fParams = { 1 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
+    ZSTD_buffered_policy_e const zbuff = (ZSTD_buffered_policy_e)(srcCCtx->inBuffSize>0);
+    ZSTD_STATIC_ASSERT((U32)ZSTDb_buffered==1);
     fParams.contentSizeFlag = pledgedSrcSize>0;
 
-    return ZSTD_copyCCtx_internal(dstCCtx, srcCCtx, fParams, pledgedSrcSize);
+    return ZSTD_copyCCtx_internal(dstCCtx, srcCCtx, fParams, pledgedSrcSize, zbuff);
 }
 
 
@@ -519,18 +1058,36 @@ static void ZSTD_reduceTable (U32* const table, U32 const size, U32 const reduce
     }
 }
 
+/*! ZSTD_ldm_reduceTable() :
+ *  reduce table indexes by `reducerValue` */
+static void ZSTD_ldm_reduceTable(ldmEntry_t* const table, U32 const size,
+                                 U32 const reducerValue)
+{
+    U32 u;
+    for (u = 0; u < size; u++) {
+        if (table[u].offset < reducerValue) table[u].offset = 0;
+        else table[u].offset -= reducerValue;
+    }
+}
+
 /*! ZSTD_reduceIndex() :
 *   rescale all indexes to avoid future overflow (indexes are U32) */
 static void ZSTD_reduceIndex (ZSTD_CCtx* zc, const U32 reducerValue)
 {
-    { U32 const hSize = 1 << zc->params.cParams.hashLog;
+    { U32 const hSize = (U32)1 << zc->appliedParams.cParams.hashLog;
       ZSTD_reduceTable(zc->hashTable, hSize, reducerValue); }
 
-    { U32 const chainSize = (zc->params.cParams.strategy == ZSTD_fast) ? 0 : (1 << zc->params.cParams.chainLog);
+    { U32 const chainSize = (zc->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : ((U32)1 << zc->appliedParams.cParams.chainLog);
       ZSTD_reduceTable(zc->chainTable, chainSize, reducerValue); }
 
-    { U32 const h3Size = (zc->hashLog3) ? 1 << zc->hashLog3 : 0;
+    { U32 const h3Size = (zc->hashLog3) ? (U32)1 << zc->hashLog3 : 0;
       ZSTD_reduceTable(zc->hashTable3, h3Size, reducerValue); }
+
+    { if (zc->appliedParams.ldmParams.enableLdm) {
+          U32 const ldmHSize = (U32)1 << zc->appliedParams.ldmParams.hashLog;
+          ZSTD_ldm_reduceTable(zc->ldmState.hashTable, ldmHSize, reducerValue);
+      }
+    }
 }
 
 
@@ -564,10 +1121,11 @@ static size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void
         case 2: /* 2 - 2 - 12 */
             MEM_writeLE16(ostart, (U16)((U32)set_basic + (1<<2) + (srcSize<<4)));
             break;
-        default:   /*note : should not be necessary : flSize is within {1,2,3} */
         case 3: /* 2 - 2 - 20 */
             MEM_writeLE32(ostart, (U32)((U32)set_basic + (3<<2) + (srcSize<<4)));
             break;
+        default:   /* not necessary : flSize is {1,2,3} */
+            assert(0);
     }
 
     memcpy(ostart + flSize, src, srcSize);
@@ -589,10 +1147,11 @@ static size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, cons
         case 2: /* 2 - 2 - 12 */
             MEM_writeLE16(ostart, (U16)((U32)set_rle + (1<<2) + (srcSize<<4)));
             break;
-        default:   /*note : should not be necessary : flSize is necessarily within {1,2,3} */
         case 3: /* 2 - 2 - 20 */
             MEM_writeLE32(ostart, (U32)((U32)set_rle + (3<<2) + (srcSize<<4)));
             break;
+        default:   /* not necessary : flSize is {1,2,3} */
+            assert(0);
     }
 
     ostart[flSize] = *(const BYTE*)src;
@@ -602,7 +1161,8 @@ static size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, cons
 
 static size_t ZSTD_minGain(size_t srcSize) { return (srcSize >> 6) + 2; }
 
-static size_t ZSTD_compressLiterals (ZSTD_CCtx* zc,
+static size_t ZSTD_compressLiterals (ZSTD_entropyCTables_t * entropy,
+                                     ZSTD_strategy strategy,
                                      void* dst, size_t dstCapacity,
                                const void* src, size_t srcSize)
 {
@@ -616,28 +1176,28 @@ static size_t ZSTD_compressLiterals (ZSTD_CCtx* zc,
 
     /* small ? don't even attempt compression (speed opt) */
 #   define LITERAL_NOENTROPY 63
-    {   size_t const minLitSize = zc->hufCTable_repeatMode == HUF_repeat_valid ? 6 : LITERAL_NOENTROPY;
+    {   size_t const minLitSize = entropy->hufCTable_repeatMode == HUF_repeat_valid ? 6 : LITERAL_NOENTROPY;
         if (srcSize <= minLitSize) return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
     }
 
     if (dstCapacity < lhSize+1) return ERROR(dstSize_tooSmall);   /* not enough space for compression */
-    {   HUF_repeat repeat = zc->hufCTable_repeatMode;
-        int const preferRepeat = zc->params.cParams.strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
+    {   HUF_repeat repeat = entropy->hufCTable_repeatMode;
+        int const preferRepeat = strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
         if (repeat == HUF_repeat_valid && lhSize == 3) singleStream = 1;
         cLitSize = singleStream ? HUF_compress1X_repeat(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11,
-                                      zc->entropyScratchSpace, entropyScratchSpace_size, zc->hufCTable, &repeat, preferRepeat)
+                                      entropy->workspace, sizeof(entropy->workspace), (HUF_CElt*)entropy->hufCTable, &repeat, preferRepeat)
                                 : HUF_compress4X_repeat(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11,
-                                      zc->entropyScratchSpace, entropyScratchSpace_size, zc->hufCTable, &repeat, preferRepeat);
+                                      entropy->workspace, sizeof(entropy->workspace), (HUF_CElt*)entropy->hufCTable, &repeat, preferRepeat);
         if (repeat != HUF_repeat_none) { hType = set_repeat; }    /* reused the existing table */
-        else { zc->hufCTable_repeatMode = HUF_repeat_check; }       /* now have a table to reuse */
+        else { entropy->hufCTable_repeatMode = HUF_repeat_check; }       /* now have a table to reuse */
     }
 
-    if ((cLitSize==0) | (cLitSize >= srcSize - minGain)) {
-        zc->hufCTable_repeatMode = HUF_repeat_none;
+    if ((cLitSize==0) | (cLitSize >= srcSize - minGain) | ERR_isError(cLitSize)) {
+        entropy->hufCTable_repeatMode = HUF_repeat_none;
         return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
     }
     if (cLitSize==1) {
-        zc->hufCTable_repeatMode = HUF_repeat_none;
+        entropy->hufCTable_repeatMode = HUF_repeat_none;
         return ZSTD_compressRleLiteralsBlock(dst, dstCapacity, src, srcSize);
     }
 
@@ -654,35 +1214,18 @@ static size_t ZSTD_compressLiterals (ZSTD_CCtx* zc,
             MEM_writeLE32(ostart, lhc);
             break;
         }
-    default:   /* should not be necessary, lhSize is only {3,4,5} */
     case 5: /* 2 - 2 - 18 - 18 */
         {   U32 const lhc = hType + (3 << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<22);
             MEM_writeLE32(ostart, lhc);
             ostart[4] = (BYTE)(cLitSize >> 10);
             break;
         }
+    default:   /* not possible : lhSize is {3,4,5} */
+        assert(0);
     }
     return lhSize+cLitSize;
 }
 
-static const BYTE LL_Code[64] = {  0,  1,  2,  3,  4,  5,  6,  7,
-                                   8,  9, 10, 11, 12, 13, 14, 15,
-                                  16, 16, 17, 17, 18, 18, 19, 19,
-                                  20, 20, 20, 20, 21, 21, 21, 21,
-                                  22, 22, 22, 22, 22, 22, 22, 22,
-                                  23, 23, 23, 23, 23, 23, 23, 23,
-                                  24, 24, 24, 24, 24, 24, 24, 24,
-                                  24, 24, 24, 24, 24, 24, 24, 24 };
-
-static const BYTE ML_Code[128] = { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
-                                  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
-                                  32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
-                                  38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39,
-                                  40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
-                                  41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
-                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
-                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42 };
-
 
 void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
 {
@@ -707,17 +1250,154 @@ void ZSTD_seqToCodes(const seqStore_t* seqStorePtr)
         mlCodeTable[seqStorePtr->longLengthPos] = MaxML;
 }
 
-MEM_STATIC size_t ZSTD_compressSequences (ZSTD_CCtx* zc,
-                              void* dst, size_t dstCapacity,
-                              size_t srcSize)
+MEM_STATIC symbolEncodingType_e ZSTD_selectEncodingType(FSE_repeat* repeatMode,
+        size_t const mostFrequent, size_t nbSeq, U32 defaultNormLog)
 {
-    const int longOffsets = zc->params.cParams.windowLog > STREAM_ACCUMULATOR_MIN;
-    const seqStore_t* seqStorePtr = &(zc->seqStore);
+#define MIN_SEQ_FOR_DYNAMIC_FSE   64
+#define MAX_SEQ_FOR_STATIC_FSE  1000
+
+    if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
+        *repeatMode = FSE_repeat_check;
+        return set_rle;
+    }
+    if ((*repeatMode == FSE_repeat_valid) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
+        return set_repeat;
+    }
+    if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (defaultNormLog-1)))) {
+        *repeatMode = FSE_repeat_valid;
+        return set_basic;
+    }
+    *repeatMode = FSE_repeat_check;
+    return set_compressed;
+}
+
+MEM_STATIC size_t ZSTD_buildCTable(void* dst, size_t dstCapacity,
+        FSE_CTable* CTable, U32 FSELog, symbolEncodingType_e type,
+        U32* count, U32 max,
+        BYTE const* codeTable, size_t nbSeq,
+        S16 const* defaultNorm, U32 defaultNormLog, U32 defaultMax,
+        void* workspace, size_t workspaceSize)
+{
+    BYTE* op = (BYTE*)dst;
+    BYTE const* const oend = op + dstCapacity;
+
+    switch (type) {
+    case set_rle:
+        *op = codeTable[0];
+        CHECK_F(FSE_buildCTable_rle(CTable, (BYTE)max));
+        return 1;
+    case set_repeat:
+        return 0;
+    case set_basic:
+        CHECK_F(FSE_buildCTable_wksp(CTable, defaultNorm, defaultMax, defaultNormLog, workspace, workspaceSize));
+        return 0;
+    case set_compressed: {
+        S16 norm[MaxSeq + 1];
+        size_t nbSeq_1 = nbSeq;
+        const U32 tableLog = FSE_optimalTableLog(FSELog, nbSeq, max);
+        if (count[codeTable[nbSeq-1]] > 1) {
+            count[codeTable[nbSeq-1]]--;
+            nbSeq_1--;
+        }
+        CHECK_F(FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max));
+        {   size_t const NCountSize = FSE_writeNCount(op, oend - op, norm, max, tableLog);   /* overflow protected */
+            if (FSE_isError(NCountSize)) return NCountSize;
+            CHECK_F(FSE_buildCTable_wksp(CTable, norm, max, tableLog, workspace, workspaceSize));
+            return NCountSize;
+        }
+    }
+    default: return assert(0), ERROR(GENERIC);
+    }
+}
+
+MEM_STATIC size_t ZSTD_encodeSequences(void* dst, size_t dstCapacity,
+    FSE_CTable const* CTable_MatchLength, BYTE const* mlCodeTable,
+    FSE_CTable const* CTable_OffsetBits, BYTE const* ofCodeTable,
+    FSE_CTable const* CTable_LitLength, BYTE const* llCodeTable,
+    seqDef const* sequences, size_t nbSeq, int longOffsets)
+{
+    BIT_CStream_t blockStream;
+    FSE_CState_t  stateMatchLength;
+    FSE_CState_t  stateOffsetBits;
+    FSE_CState_t  stateLitLength;
+
+    CHECK_E(BIT_initCStream(&blockStream, dst, dstCapacity), dstSize_tooSmall); /* not enough space remaining */
+
+    /* first symbols */
+    FSE_initCState2(&stateMatchLength, CTable_MatchLength, mlCodeTable[nbSeq-1]);
+    FSE_initCState2(&stateOffsetBits,  CTable_OffsetBits,  ofCodeTable[nbSeq-1]);
+    FSE_initCState2(&stateLitLength,   CTable_LitLength,   llCodeTable[nbSeq-1]);
+    BIT_addBits(&blockStream, sequences[nbSeq-1].litLength, LL_bits[llCodeTable[nbSeq-1]]);
+    if (MEM_32bits()) BIT_flushBits(&blockStream);
+    BIT_addBits(&blockStream, sequences[nbSeq-1].matchLength, ML_bits[mlCodeTable[nbSeq-1]]);
+    if (MEM_32bits()) BIT_flushBits(&blockStream);
+    if (longOffsets) {
+        U32 const ofBits = ofCodeTable[nbSeq-1];
+        int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+        if (extraBits) {
+            BIT_addBits(&blockStream, sequences[nbSeq-1].offset, extraBits);
+            BIT_flushBits(&blockStream);
+        }
+        BIT_addBits(&blockStream, sequences[nbSeq-1].offset >> extraBits,
+                    ofBits - extraBits);
+    } else {
+        BIT_addBits(&blockStream, sequences[nbSeq-1].offset, ofCodeTable[nbSeq-1]);
+    }
+    BIT_flushBits(&blockStream);
+
+    {   size_t n;
+        for (n=nbSeq-2 ; n<nbSeq ; n--) {      /* intentional underflow */
+            BYTE const llCode = llCodeTable[n];
+            BYTE const ofCode = ofCodeTable[n];
+            BYTE const mlCode = mlCodeTable[n];
+            U32  const llBits = LL_bits[llCode];
+            U32  const ofBits = ofCode;                                     /* 32b*/  /* 64b*/
+            U32  const mlBits = ML_bits[mlCode];
+                                                                            /* (7)*/  /* (7)*/
+            FSE_encodeSymbol(&blockStream, &stateOffsetBits, ofCode);       /* 15 */  /* 15 */
+            FSE_encodeSymbol(&blockStream, &stateMatchLength, mlCode);      /* 24 */  /* 24 */
+            if (MEM_32bits()) BIT_flushBits(&blockStream);                  /* (7)*/
+            FSE_encodeSymbol(&blockStream, &stateLitLength, llCode);        /* 16 */  /* 33 */
+            if (MEM_32bits() || (ofBits+mlBits+llBits >= 64-7-(LLFSELog+MLFSELog+OffFSELog)))
+                BIT_flushBits(&blockStream);                                /* (7)*/
+            BIT_addBits(&blockStream, sequences[n].litLength, llBits);
+            if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(&blockStream);
+            BIT_addBits(&blockStream, sequences[n].matchLength, mlBits);
+            if (MEM_32bits() || (ofBits+mlBits+llBits > 56)) BIT_flushBits(&blockStream);
+            if (longOffsets) {
+                int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
+                if (extraBits) {
+                    BIT_addBits(&blockStream, sequences[n].offset, extraBits);
+                    BIT_flushBits(&blockStream);                            /* (7)*/
+                }
+                BIT_addBits(&blockStream, sequences[n].offset >> extraBits,
+                            ofBits - extraBits);                            /* 31 */
+            } else {
+                BIT_addBits(&blockStream, sequences[n].offset, ofBits);     /* 31 */
+            }
+            BIT_flushBits(&blockStream);                                    /* (7)*/
+    }   }
+
+    FSE_flushCState(&blockStream, &stateMatchLength);
+    FSE_flushCState(&blockStream, &stateOffsetBits);
+    FSE_flushCState(&blockStream, &stateLitLength);
+
+    {   size_t const streamSize = BIT_closeCStream(&blockStream);
+        if (streamSize==0) return ERROR(dstSize_tooSmall);   /* not enough space */
+        return streamSize;
+    }
+}
+
+MEM_STATIC size_t ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
+                              ZSTD_entropyCTables_t* entropy,
+                              ZSTD_compressionParameters const* cParams,
+                              void* dst, size_t dstCapacity)
+{
+    const int longOffsets = cParams->windowLog > STREAM_ACCUMULATOR_MIN;
     U32 count[MaxSeq+1];
-    S16 norm[MaxSeq+1];
-    FSE_CTable* CTable_LitLength = zc->litlengthCTable;
-    FSE_CTable* CTable_OffsetBits = zc->offcodeCTable;
-    FSE_CTable* CTable_MatchLength = zc->matchlengthCTable;
+    FSE_CTable* CTable_LitLength = entropy->litlengthCTable;
+    FSE_CTable* CTable_OffsetBits = entropy->offcodeCTable;
+    FSE_CTable* CTable_MatchLength = entropy->matchlengthCTable;
     U32 LLtype, Offtype, MLtype;   /* compressed, raw or rle */
     const seqDef* const sequences = seqStorePtr->sequencesStart;
     const BYTE* const ofCodeTable = seqStorePtr->ofCode;
@@ -728,13 +1408,16 @@ MEM_STATIC size_t ZSTD_compressSequences (ZSTD_CCtx* zc,
     BYTE* op = ostart;
     size_t const nbSeq = seqStorePtr->sequences - seqStorePtr->sequencesStart;
     BYTE* seqHead;
-    BYTE scratchBuffer[1<<MAX(MLFSELog,LLFSELog)];
+
+    ZSTD_STATIC_ASSERT(sizeof(entropy->workspace) >= (1<<MAX(MLFSELog,LLFSELog)));
 
     /* Compress literals */
     {   const BYTE* const literals = seqStorePtr->litStart;
         size_t const litSize = seqStorePtr->lit - literals;
-        size_t const cSize = ZSTD_compressLiterals(zc, op, dstCapacity, literals, litSize);
-        if (ZSTD_isError(cSize)) return cSize;
+        size_t const cSize = ZSTD_compressLiterals(
+                entropy, cParams->strategy, op, dstCapacity, literals, litSize);
+        if (ZSTD_isError(cSize))
+          return cSize;
         op += cSize;
     }
 
@@ -743,1739 +1426,157 @@ MEM_STATIC size_t ZSTD_compressSequences (ZSTD_CCtx* zc,
     if (nbSeq < 0x7F) *op++ = (BYTE)nbSeq;
     else if (nbSeq < LONGNBSEQ) op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
     else op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
-    if (nbSeq==0) goto _check_compressibility;
+    if (nbSeq==0) return op - ostart;
 
     /* seqHead : flags for FSE encoding type */
     seqHead = op++;
 
-#define MIN_SEQ_FOR_DYNAMIC_FSE   64
-#define MAX_SEQ_FOR_STATIC_FSE  1000
-
     /* convert length/distances into codes */
     ZSTD_seqToCodes(seqStorePtr);
-
     /* CTable for Literal Lengths */
     {   U32 max = MaxLL;
-        size_t const mostFrequent = FSE_countFast_wksp(count, &max, llCodeTable, nbSeq, zc->entropyScratchSpace);
-        if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
-            *op++ = llCodeTable[0];
-            FSE_buildCTable_rle(CTable_LitLength, (BYTE)max);
-            LLtype = set_rle;
-        } else if ((zc->fseCTables_ready) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
-            LLtype = set_repeat;
-        } else if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (LL_defaultNormLog-1)))) {
-            FSE_buildCTable_wksp(CTable_LitLength, LL_defaultNorm, MaxLL, LL_defaultNormLog, scratchBuffer, sizeof(scratchBuffer));
-            LLtype = set_basic;
-        } else {
-            size_t nbSeq_1 = nbSeq;
-            const U32 tableLog = FSE_optimalTableLog(LLFSELog, nbSeq, max);
-            if (count[llCodeTable[nbSeq-1]]>1) { count[llCodeTable[nbSeq-1]]--; nbSeq_1--; }
-            FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max);
-            { size_t const NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog);   /* overflow protected */
-              if (FSE_isError(NCountSize)) return NCountSize;
-              op += NCountSize; }
-            FSE_buildCTable_wksp(CTable_LitLength, norm, max, tableLog, scratchBuffer, sizeof(scratchBuffer));
-            LLtype = set_compressed;
+        size_t const mostFrequent = FSE_countFast_wksp(count, &max, llCodeTable, nbSeq, entropy->workspace);
+        LLtype = ZSTD_selectEncodingType(&entropy->litlength_repeatMode, mostFrequent, nbSeq, LL_defaultNormLog);
+        {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
+                    count, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
+                    entropy->workspace, sizeof(entropy->workspace));
+            if (ZSTD_isError(countSize)) return countSize;
+            op += countSize;
     }   }
-
     /* CTable for Offsets */
     {   U32 max = MaxOff;
-        size_t const mostFrequent = FSE_countFast_wksp(count, &max, ofCodeTable, nbSeq, zc->entropyScratchSpace);
-        if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
-            *op++ = ofCodeTable[0];
-            FSE_buildCTable_rle(CTable_OffsetBits, (BYTE)max);
-            Offtype = set_rle;
-        } else if ((zc->fseCTables_ready) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
-            Offtype = set_repeat;
-        } else if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (OF_defaultNormLog-1)))) {
-            FSE_buildCTable_wksp(CTable_OffsetBits, OF_defaultNorm, MaxOff, OF_defaultNormLog, scratchBuffer, sizeof(scratchBuffer));
-            Offtype = set_basic;
-        } else {
-            size_t nbSeq_1 = nbSeq;
-            const U32 tableLog = FSE_optimalTableLog(OffFSELog, nbSeq, max);
-            if (count[ofCodeTable[nbSeq-1]]>1) { count[ofCodeTable[nbSeq-1]]--; nbSeq_1--; }
-            FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max);
-            { size_t const NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog);   /* overflow protected */
-              if (FSE_isError(NCountSize)) return NCountSize;
-              op += NCountSize; }
-            FSE_buildCTable_wksp(CTable_OffsetBits, norm, max, tableLog, scratchBuffer, sizeof(scratchBuffer));
-            Offtype = set_compressed;
+        size_t const mostFrequent = FSE_countFast_wksp(count, &max, ofCodeTable, nbSeq, entropy->workspace);
+        Offtype = ZSTD_selectEncodingType(&entropy->offcode_repeatMode, mostFrequent, nbSeq, OF_defaultNormLog);
+        {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
+                    count, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, MaxOff,
+                    entropy->workspace, sizeof(entropy->workspace));
+            if (ZSTD_isError(countSize)) return countSize;
+            op += countSize;
     }   }
-
     /* CTable for MatchLengths */
     {   U32 max = MaxML;
-        size_t const mostFrequent = FSE_countFast_wksp(count, &max, mlCodeTable, nbSeq, zc->entropyScratchSpace);
-        if ((mostFrequent == nbSeq) && (nbSeq > 2)) {
-            *op++ = *mlCodeTable;
-            FSE_buildCTable_rle(CTable_MatchLength, (BYTE)max);
-            MLtype = set_rle;
-        } else if ((zc->fseCTables_ready) && (nbSeq < MAX_SEQ_FOR_STATIC_FSE)) {
-            MLtype = set_repeat;
-        } else if ((nbSeq < MIN_SEQ_FOR_DYNAMIC_FSE) || (mostFrequent < (nbSeq >> (ML_defaultNormLog-1)))) {
-            FSE_buildCTable_wksp(CTable_MatchLength, ML_defaultNorm, MaxML, ML_defaultNormLog, scratchBuffer, sizeof(scratchBuffer));
-            MLtype = set_basic;
-        } else {
-            size_t nbSeq_1 = nbSeq;
-            const U32 tableLog = FSE_optimalTableLog(MLFSELog, nbSeq, max);
-            if (count[mlCodeTable[nbSeq-1]]>1) { count[mlCodeTable[nbSeq-1]]--; nbSeq_1--; }
-            FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max);
-            { size_t const NCountSize = FSE_writeNCount(op, oend-op, norm, max, tableLog);   /* overflow protected */
-              if (FSE_isError(NCountSize)) return NCountSize;
-              op += NCountSize; }
-            FSE_buildCTable_wksp(CTable_MatchLength, norm, max, tableLog, scratchBuffer, sizeof(scratchBuffer));
-            MLtype = set_compressed;
+        size_t const mostFrequent = FSE_countFast_wksp(count, &max, mlCodeTable, nbSeq, entropy->workspace);
+        MLtype = ZSTD_selectEncodingType(&entropy->matchlength_repeatMode, mostFrequent, nbSeq, ML_defaultNormLog);
+        {   size_t const countSize = ZSTD_buildCTable(op, oend - op, CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
+                    count, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
+                    entropy->workspace, sizeof(entropy->workspace));
+            if (ZSTD_isError(countSize)) return countSize;
+            op += countSize;
     }   }
 
     *seqHead = (BYTE)((LLtype<<6) + (Offtype<<4) + (MLtype<<2));
-    zc->fseCTables_ready = 0;
-
-    /* Encoding Sequences */
-    {   BIT_CStream_t blockStream;
-        FSE_CState_t  stateMatchLength;
-        FSE_CState_t  stateOffsetBits;
-        FSE_CState_t  stateLitLength;
-
-        CHECK_E(BIT_initCStream(&blockStream, op, oend-op), dstSize_tooSmall); /* not enough space remaining */
-
-        /* first symbols */
-        FSE_initCState2(&stateMatchLength, CTable_MatchLength, mlCodeTable[nbSeq-1]);
-        FSE_initCState2(&stateOffsetBits,  CTable_OffsetBits,  ofCodeTable[nbSeq-1]);
-        FSE_initCState2(&stateLitLength,   CTable_LitLength,   llCodeTable[nbSeq-1]);
-        BIT_addBits(&blockStream, sequences[nbSeq-1].litLength, LL_bits[llCodeTable[nbSeq-1]]);
-        if (MEM_32bits()) BIT_flushBits(&blockStream);
-        BIT_addBits(&blockStream, sequences[nbSeq-1].matchLength, ML_bits[mlCodeTable[nbSeq-1]]);
-        if (MEM_32bits()) BIT_flushBits(&blockStream);
-        if (longOffsets) {
-            U32 const ofBits = ofCodeTable[nbSeq-1];
-            int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
-            if (extraBits) {
-                BIT_addBits(&blockStream, sequences[nbSeq-1].offset, extraBits);
-                BIT_flushBits(&blockStream);
-            }
-            BIT_addBits(&blockStream, sequences[nbSeq-1].offset >> extraBits,
-                        ofBits - extraBits);
-        } else {
-            BIT_addBits(&blockStream, sequences[nbSeq-1].offset, ofCodeTable[nbSeq-1]);
-        }
-        BIT_flushBits(&blockStream);
-
-        {   size_t n;
-            for (n=nbSeq-2 ; n<nbSeq ; n--) {      /* intentional underflow */
-                BYTE const llCode = llCodeTable[n];
-                BYTE const ofCode = ofCodeTable[n];
-                BYTE const mlCode = mlCodeTable[n];
-                U32  const llBits = LL_bits[llCode];
-                U32  const ofBits = ofCode;                                     /* 32b*/  /* 64b*/
-                U32  const mlBits = ML_bits[mlCode];
-                                                                                /* (7)*/  /* (7)*/
-                FSE_encodeSymbol(&blockStream, &stateOffsetBits, ofCode);       /* 15 */  /* 15 */
-                FSE_encodeSymbol(&blockStream, &stateMatchLength, mlCode);      /* 24 */  /* 24 */
-                if (MEM_32bits()) BIT_flushBits(&blockStream);                  /* (7)*/
-                FSE_encodeSymbol(&blockStream, &stateLitLength, llCode);        /* 16 */  /* 33 */
-                if (MEM_32bits() || (ofBits+mlBits+llBits >= 64-7-(LLFSELog+MLFSELog+OffFSELog)))
-                    BIT_flushBits(&blockStream);                                /* (7)*/
-                BIT_addBits(&blockStream, sequences[n].litLength, llBits);
-                if (MEM_32bits() && ((llBits+mlBits)>24)) BIT_flushBits(&blockStream);
-                BIT_addBits(&blockStream, sequences[n].matchLength, mlBits);
-                if (MEM_32bits()) BIT_flushBits(&blockStream);                  /* (7)*/
-                if (longOffsets) {
-                    int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN-1);
-                    if (extraBits) {
-                        BIT_addBits(&blockStream, sequences[n].offset, extraBits);
-                        BIT_flushBits(&blockStream);                            /* (7)*/
-                    }
-                    BIT_addBits(&blockStream, sequences[n].offset >> extraBits,
-                                ofBits - extraBits);                            /* 31 */
-                } else {
-                    BIT_addBits(&blockStream, sequences[n].offset, ofBits);     /* 31 */
-                }
-                BIT_flushBits(&blockStream);                                    /* (7)*/
-        }   }
-
-        FSE_flushCState(&blockStream, &stateMatchLength);
-        FSE_flushCState(&blockStream, &stateOffsetBits);
-        FSE_flushCState(&blockStream, &stateLitLength);
-
-        {   size_t const streamSize = BIT_closeCStream(&blockStream);
-            if (streamSize==0) return ERROR(dstSize_tooSmall);   /* not enough space */
-            op += streamSize;
-    }   }
-
-    /* check compressibility */
-_check_compressibility:
-    {   size_t const minGain = ZSTD_minGain(srcSize);
-        size_t const maxCSize = srcSize - minGain;
-        if ((size_t)(op-ostart) >= maxCSize) {
-            zc->hufCTable_repeatMode = HUF_repeat_none;
-            return 0;
-    }   }
-
-    /* confirm repcodes */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) zc->rep[i] = zc->repToConfirm[i]; }
-
-    return op - ostart;
-}
-
-#if 0 /* for debug */
-#  define STORESEQ_DEBUG
-#include <stdio.h>   /* fprintf */
-U32 g_startDebug = 0;
-const BYTE* g_start = NULL;
-#endif
-
-/*! ZSTD_storeSeq() :
-    Store a sequence (literal length, literals, offset code and match length code) into seqStore_t.
-    `offsetCode` : distance to match, or 0 == repCode.
-    `matchCode` : matchLength - MINMATCH
-*/
-MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const void* literals, U32 offsetCode, size_t matchCode)
-{
-#ifdef STORESEQ_DEBUG
-    if (g_startDebug) {
-        const U32 pos = (U32)((const BYTE*)literals - g_start);
-        if (g_start==NULL) g_start = (const BYTE*)literals;
-        if ((pos > 1895000) && (pos < 1895300))
-            DEBUGLOG(5, "Cpos %6u :%5u literals & match %3u bytes at distance %6u \n",
-                   pos, (U32)litLength, (U32)matchCode+MINMATCH, (U32)offsetCode);
-    }
-#endif
-    /* copy Literals */
-    ZSTD_wildcopy(seqStorePtr->lit, literals, litLength);
-    seqStorePtr->lit += litLength;
-
-    /* literal Length */
-    if (litLength>0xFFFF) {
-        seqStorePtr->longLengthID = 1;
-        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
-    }
-    seqStorePtr->sequences[0].litLength = (U16)litLength;
-
-    /* match offset */
-    seqStorePtr->sequences[0].offset = offsetCode + 1;
-
-    /* match Length */
-    if (matchCode>0xFFFF) {
-        seqStorePtr->longLengthID = 2;
-        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
-    }
-    seqStorePtr->sequences[0].matchLength = (U16)matchCode;
-
-    seqStorePtr->sequences++;
-}
-
-
-/*-*************************************
-*  Match length counter
-***************************************/
-static unsigned ZSTD_NbCommonBytes (register size_t val)
-{
-    if (MEM_isLittleEndian()) {
-        if (MEM_64bits()) {
-#       if defined(_MSC_VER) && defined(_WIN64)
-            unsigned long r = 0;
-            _BitScanForward64( &r, (U64)val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 3)
-            return (__builtin_ctzll((U64)val) >> 3);
-#       else
-            static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2,
-                                                     0, 3, 1, 3, 1, 4, 2, 7,
-                                                     0, 2, 3, 6, 1, 5, 3, 5,
-                                                     1, 3, 4, 4, 2, 5, 6, 7,
-                                                     7, 0, 1, 2, 3, 3, 4, 6,
-                                                     2, 6, 5, 5, 3, 4, 5, 6,
-                                                     7, 1, 2, 4, 6, 4, 4, 5,
-                                                     7, 2, 6, 5, 7, 6, 7, 7 };
-            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
-#       endif
-        } else { /* 32 bits */
-#       if defined(_MSC_VER)
-            unsigned long r=0;
-            _BitScanForward( &r, (U32)val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 3)
-            return (__builtin_ctz((U32)val) >> 3);
-#       else
-            static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0,
-                                                     3, 2, 2, 1, 3, 2, 0, 1,
-                                                     3, 3, 1, 2, 2, 2, 2, 0,
-                                                     3, 1, 2, 0, 1, 0, 1, 1 };
-            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
-#       endif
-        }
-    } else {  /* Big Endian CPU */
-        if (MEM_64bits()) {
-#       if defined(_MSC_VER) && defined(_WIN64)
-            unsigned long r = 0;
-            _BitScanReverse64( &r, val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 3)
-            return (__builtin_clzll(val) >> 3);
-#       else
-            unsigned r;
-            const unsigned n32 = sizeof(size_t)*4;   /* calculate this way due to compiler complaining in 32-bits mode */
-            if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
-            if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
-            r += (!val);
-            return r;
-#       endif
-        } else { /* 32 bits */
-#       if defined(_MSC_VER)
-            unsigned long r = 0;
-            _BitScanReverse( &r, (unsigned long)val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 3)
-            return (__builtin_clz((U32)val) >> 3);
-#       else
-            unsigned r;
-            if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
-            r += (!val);
-            return r;
-#       endif
-    }   }
-}
-
-
-static size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* const pInLimit)
-{
-    const BYTE* const pStart = pIn;
-    const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
-
-    while (pIn < pInLoopLimit) {
-        size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
-        if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
-        pIn += ZSTD_NbCommonBytes(diff);
-        return (size_t)(pIn - pStart);
-    }
-    if (MEM_64bits()) if ((pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
-    if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
-    if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
-    return (size_t)(pIn - pStart);
-}
-
-/** ZSTD_count_2segments() :
-*   can count match length with `ip` & `match` in 2 different segments.
-*   convention : on reaching mEnd, match count continue starting from iStart
-*/
-static size_t ZSTD_count_2segments(const BYTE* ip, const BYTE* match, const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
-{
-    const BYTE* const vEnd = MIN( ip + (mEnd - match), iEnd);
-    size_t const matchLength = ZSTD_count(ip, match, vEnd);
-    if (match + matchLength != mEnd) return matchLength;
-    return matchLength + ZSTD_count(ip+matchLength, iStart, iEnd);
-}
-
-
-/*-*************************************
-*  Hashes
-***************************************/
-static const U32 prime3bytes = 506832829U;
-static U32    ZSTD_hash3(U32 u, U32 h) { return ((u << (32-24)) * prime3bytes)  >> (32-h) ; }
-MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); } /* only in zstd_opt.h */
-
-static const U32 prime4bytes = 2654435761U;
-static U32    ZSTD_hash4(U32 u, U32 h) { return (u * prime4bytes) >> (32-h) ; }
-static size_t ZSTD_hash4Ptr(const void* ptr, U32 h) { return ZSTD_hash4(MEM_read32(ptr), h); }
-
-static const U64 prime5bytes = 889523592379ULL;
-static size_t ZSTD_hash5(U64 u, U32 h) { return (size_t)(((u  << (64-40)) * prime5bytes) >> (64-h)) ; }
-static size_t ZSTD_hash5Ptr(const void* p, U32 h) { return ZSTD_hash5(MEM_readLE64(p), h); }
-
-static const U64 prime6bytes = 227718039650203ULL;
-static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u  << (64-48)) * prime6bytes) >> (64-h)) ; }
-static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
-
-static const U64 prime7bytes = 58295818150454627ULL;
-static size_t ZSTD_hash7(U64 u, U32 h) { return (size_t)(((u  << (64-56)) * prime7bytes) >> (64-h)) ; }
-static size_t ZSTD_hash7Ptr(const void* p, U32 h) { return ZSTD_hash7(MEM_readLE64(p), h); }
-
-static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
-static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
-static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
-
-static size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
-{
-    switch(mls)
-    {
-    default:
-    case 4: return ZSTD_hash4Ptr(p, hBits);
-    case 5: return ZSTD_hash5Ptr(p, hBits);
-    case 6: return ZSTD_hash6Ptr(p, hBits);
-    case 7: return ZSTD_hash7Ptr(p, hBits);
-    case 8: return ZSTD_hash8Ptr(p, hBits);
-    }
-}
-
-
-/*-*************************************
-*  Fast Scan
-***************************************/
-static void ZSTD_fillHashTable (ZSTD_CCtx* zc, const void* end, const U32 mls)
-{
-    U32* const hashTable = zc->hashTable;
-    U32  const hBits = zc->params.cParams.hashLog;
-    const BYTE* const base = zc->base;
-    const BYTE* ip = base + zc->nextToUpdate;
-    const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
-    const size_t fastHashFillStep = 3;
-
-    while(ip <= iend) {
-        hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip - base);
-        ip += fastHashFillStep;
-    }
-}
-
-
-FORCE_INLINE
-void ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
-                               const void* src, size_t srcSize,
-                               const U32 mls)
-{
-    U32* const hashTable = cctx->hashTable;
-    U32  const hBits = cctx->params.cParams.hashLog;
-    seqStore_t* seqStorePtr = &(cctx->seqStore);
-    const BYTE* const base = cctx->base;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32   lowestIndex = cctx->dictLimit;
-    const BYTE* const lowest = base + lowestIndex;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - HASH_READ_SIZE;
-    U32 offset_1=cctx->rep[0], offset_2=cctx->rep[1];
-    U32 offsetSaved = 0;
-
-    /* init */
-    ip += (ip==lowest);
-    {   U32 const maxRep = (U32)(ip-lowest);
-        if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
-        if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
-    }
-
-    /* Main Search Loop */
-    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
-        size_t mLength;
-        size_t const h = ZSTD_hashPtr(ip, hBits, mls);
-        U32 const current = (U32)(ip-base);
-        U32 const matchIndex = hashTable[h];
-        const BYTE* match = base + matchIndex;
-        hashTable[h] = current;   /* update hash table */
-
-        if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
-            mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
-            ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
-        } else {
-            U32 offset;
-            if ( (matchIndex <= lowestIndex) || (MEM_read32(match) != MEM_read32(ip)) ) {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
-                continue;
-            }
-            mLength = ZSTD_count(ip+4, match+4, iend) + 4;
-            offset = (U32)(ip-match);
-            while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
-            offset_2 = offset_1;
-            offset_1 = offset;
-
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
-        }
-
-        /* match found */
-        ip += mLength;
-        anchor = ip;
-
-        if (ip <= ilimit) {
-            /* Fill Table */
-            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;  /* here because current+2 could be > iend-8 */
-            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
-            /* check immediate repcode */
-            while ( (ip <= ilimit)
-                 && ( (offset_2>0)
-                 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
-                /* store sequence */
-                size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
-                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; }  /* swap offset_2 <=> offset_1 */
-                hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip-base);
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
-                ip += rLength;
-                anchor = ip;
-                continue;   /* faster when present ... (?) */
-    }   }   }
-
-    /* save reps for next block */
-    cctx->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
-    cctx->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
-}
-
-
-static void ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
-                       const void* src, size_t srcSize)
-{
-    const U32 mls = ctx->params.cParams.searchLength;
-    switch(mls)
-    {
-    default: /* includes case 3 */
-    case 4 :
-        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 4); return;
-    case 5 :
-        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 5); return;
-    case 6 :
-        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 6); return;
-    case 7 :
-        ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 7); return;
-    }
-}
-
-
-static void ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
-{
-    U32* hashTable = ctx->hashTable;
-    const U32 hBits = ctx->params.cParams.hashLog;
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const base = ctx->base;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32   lowestIndex = ctx->lowLimit;
-    const BYTE* const dictStart = dictBase + lowestIndex;
-    const U32   dictLimit = ctx->dictLimit;
-    const BYTE* const lowPrefixPtr = base + dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    U32 offset_1=ctx->rep[0], offset_2=ctx->rep[1];
-
-    /* Search Loop */
-    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
-        const size_t h = ZSTD_hashPtr(ip, hBits, mls);
-        const U32 matchIndex = hashTable[h];
-        const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
-        const BYTE* match = matchBase + matchIndex;
-        const U32 current = (U32)(ip-base);
-        const U32 repIndex = current + 1 - offset_1;   /* offset_1 expected <= current +1 */
-        const BYTE* repBase = repIndex < dictLimit ? dictBase : base;
-        const BYTE* repMatch = repBase + repIndex;
-        size_t mLength;
-        hashTable[h] = current;   /* update hash table */
-
-        if ( (((U32)((dictLimit-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > lowestIndex))
-           && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
-            const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
-            mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
-            ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
-        } else {
-            if ( (matchIndex < lowestIndex) ||
-                 (MEM_read32(match) != MEM_read32(ip)) ) {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
-                continue;
-            }
-            {   const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
-                const BYTE* lowMatchPtr = matchIndex < dictLimit ? dictStart : lowPrefixPtr;
-                U32 offset;
-                mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, lowPrefixPtr) + 4;
-                while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
-                offset = current - matchIndex;
-                offset_2 = offset_1;
-                offset_1 = offset;
-                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
-        }   }
-
-        /* found a match : store it */
-        ip += mLength;
-        anchor = ip;
-
-        if (ip <= ilimit) {
-            /* Fill Table */
-            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;
-            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
-            /* check immediate repcode */
-            while (ip <= ilimit) {
-                U32 const current2 = (U32)(ip-base);
-                U32 const repIndex2 = current2 - offset_2;
-                const BYTE* repMatch2 = repIndex2 < dictLimit ? dictBase + repIndex2 : base + repIndex2;
-                if ( (((U32)((dictLimit-1) - repIndex2) >= 3) & (repIndex2 > lowestIndex))  /* intentional overflow */
-                   && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
-                    const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
-                    size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
-                    U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
-                    hashTable[ZSTD_hashPtr(ip, hBits, mls)] = current2;
-                    ip += repLength2;
-                    anchor = ip;
-                    continue;
-                }
-                break;
-    }   }   }
-
-    /* save reps for next block */
-    ctx->repToConfirm[0] = offset_1; ctx->repToConfirm[1] = offset_2;
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
-}
-
-
-static void ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
-                         const void* src, size_t srcSize)
-{
-    U32 const mls = ctx->params.cParams.searchLength;
-    switch(mls)
-    {
-    default: /* includes case 3 */
-    case 4 :
-        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 4); return;
-    case 5 :
-        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 5); return;
-    case 6 :
-        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 6); return;
-    case 7 :
-        ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 7); return;
-    }
-}
-
 
-/*-*************************************
-*  Double Fast
-***************************************/
-static void ZSTD_fillDoubleHashTable (ZSTD_CCtx* cctx, const void* end, const U32 mls)
-{
-    U32* const hashLarge = cctx->hashTable;
-    U32  const hBitsL = cctx->params.cParams.hashLog;
-    U32* const hashSmall = cctx->chainTable;
-    U32  const hBitsS = cctx->params.cParams.chainLog;
-    const BYTE* const base = cctx->base;
-    const BYTE* ip = base + cctx->nextToUpdate;
-    const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
-    const size_t fastHashFillStep = 3;
-
-    while(ip <= iend) {
-        hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip - base);
-        hashLarge[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip - base);
-        ip += fastHashFillStep;
-    }
-}
-
-
-FORCE_INLINE
-void ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
-{
-    U32* const hashLong = cctx->hashTable;
-    const U32 hBitsL = cctx->params.cParams.hashLog;
-    U32* const hashSmall = cctx->chainTable;
-    const U32 hBitsS = cctx->params.cParams.chainLog;
-    seqStore_t* seqStorePtr = &(cctx->seqStore);
-    const BYTE* const base = cctx->base;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32 lowestIndex = cctx->dictLimit;
-    const BYTE* const lowest = base + lowestIndex;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - HASH_READ_SIZE;
-    U32 offset_1=cctx->rep[0], offset_2=cctx->rep[1];
-    U32 offsetSaved = 0;
-
-    /* init */
-    ip += (ip==lowest);
-    {   U32 const maxRep = (U32)(ip-lowest);
-        if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
-        if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
-    }
-
-    /* Main Search Loop */
-    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
-        size_t mLength;
-        size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
-        size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
-        U32 const current = (U32)(ip-base);
-        U32 const matchIndexL = hashLong[h2];
-        U32 const matchIndexS = hashSmall[h];
-        const BYTE* matchLong = base + matchIndexL;
-        const BYTE* match = base + matchIndexS;
-        hashLong[h2] = hashSmall[h] = current;   /* update hash tables */
-
-        assert(offset_1 <= current);   /* supposed guaranteed by construction */
-        if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
-            /* favor repcode */
-            mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
-            ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
-        } else {
-            U32 offset;
-            if ( (matchIndexL > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip)) ) {
-                mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
-                offset = (U32)(ip-matchLong);
-                while (((ip>anchor) & (matchLong>lowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
-            } else if ( (matchIndexS > lowestIndex) && (MEM_read32(match) == MEM_read32(ip)) ) {
-                size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
-                U32 const matchIndexL3 = hashLong[hl3];
-                const BYTE* matchL3 = base + matchIndexL3;
-                hashLong[hl3] = current + 1;
-                if ( (matchIndexL3 > lowestIndex) && (MEM_read64(matchL3) == MEM_read64(ip+1)) ) {
-                    mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
-                    ip++;
-                    offset = (U32)(ip-matchL3);
-                    while (((ip>anchor) & (matchL3>lowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
-                } else {
-                    mLength = ZSTD_count(ip+4, match+4, iend) + 4;
-                    offset = (U32)(ip-match);
-                    while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
-                }
-            } else {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
-                continue;
-            }
-
-            offset_2 = offset_1;
-            offset_1 = offset;
-
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
-        }
-
-        /* match found */
-        ip += mLength;
-        anchor = ip;
-
-        if (ip <= ilimit) {
-            /* Fill Table */
-            hashLong[ZSTD_hashPtr(base+current+2, hBitsL, 8)] =
-                hashSmall[ZSTD_hashPtr(base+current+2, hBitsS, mls)] = current+2;  /* here because current+2 could be > iend-8 */
-            hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] =
-                hashSmall[ZSTD_hashPtr(ip-2, hBitsS, mls)] = (U32)(ip-2-base);
-
-            /* check immediate repcode */
-            while ( (ip <= ilimit)
-                 && ( (offset_2>0)
-                 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
-                /* store sequence */
-                size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
-                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
-                hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
-                hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
-                ip += rLength;
-                anchor = ip;
-                continue;   /* faster when present ... (?) */
-    }   }   }
-
-    /* save reps for next block */
-    cctx->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
-    cctx->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
-}
-
-
-static void ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    const U32 mls = ctx->params.cParams.searchLength;
-    switch(mls)
-    {
-    default: /* includes case 3 */
-    case 4 :
-        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 4); return;
-    case 5 :
-        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 5); return;
-    case 6 :
-        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 6); return;
-    case 7 :
-        ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 7); return;
-    }
-}
-
-
-static void ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
-{
-    U32* const hashLong = ctx->hashTable;
-    U32  const hBitsL = ctx->params.cParams.hashLog;
-    U32* const hashSmall = ctx->chainTable;
-    U32  const hBitsS = ctx->params.cParams.chainLog;
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const base = ctx->base;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32   lowestIndex = ctx->lowLimit;
-    const BYTE* const dictStart = dictBase + lowestIndex;
-    const U32   dictLimit = ctx->dictLimit;
-    const BYTE* const lowPrefixPtr = base + dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    U32 offset_1=ctx->rep[0], offset_2=ctx->rep[1];
-
-    /* Search Loop */
-    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
-        const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
-        const U32 matchIndex = hashSmall[hSmall];
-        const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
-        const BYTE* match = matchBase + matchIndex;
-
-        const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
-        const U32 matchLongIndex = hashLong[hLong];
-        const BYTE* matchLongBase = matchLongIndex < dictLimit ? dictBase : base;
-        const BYTE* matchLong = matchLongBase + matchLongIndex;
-
-        const U32 current = (U32)(ip-base);
-        const U32 repIndex = current + 1 - offset_1;   /* offset_1 expected <= current +1 */
-        const BYTE* repBase = repIndex < dictLimit ? dictBase : base;
-        const BYTE* repMatch = repBase + repIndex;
-        size_t mLength;
-        hashSmall[hSmall] = hashLong[hLong] = current;   /* update hash table */
-
-        if ( (((U32)((dictLimit-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > lowestIndex))
-           && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
-            const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
-            mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
-            ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
-        } else {
-            if ((matchLongIndex > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
-                const BYTE* matchEnd = matchLongIndex < dictLimit ? dictEnd : iend;
-                const BYTE* lowMatchPtr = matchLongIndex < dictLimit ? dictStart : lowPrefixPtr;
-                U32 offset;
-                mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, lowPrefixPtr) + 8;
-                offset = current - matchLongIndex;
-                while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; }   /* catch up */
-                offset_2 = offset_1;
-                offset_1 = offset;
-                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
-
-            } else if ((matchIndex > lowestIndex) && (MEM_read32(match) == MEM_read32(ip))) {
-                size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
-                U32 const matchIndex3 = hashLong[h3];
-                const BYTE* const match3Base = matchIndex3 < dictLimit ? dictBase : base;
-                const BYTE* match3 = match3Base + matchIndex3;
-                U32 offset;
-                hashLong[h3] = current + 1;
-                if ( (matchIndex3 > lowestIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
-                    const BYTE* matchEnd = matchIndex3 < dictLimit ? dictEnd : iend;
-                    const BYTE* lowMatchPtr = matchIndex3 < dictLimit ? dictStart : lowPrefixPtr;
-                    mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, lowPrefixPtr) + 8;
-                    ip++;
-                    offset = current+1 - matchIndex3;
-                    while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
-                } else {
-                    const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
-                    const BYTE* lowMatchPtr = matchIndex < dictLimit ? dictStart : lowPrefixPtr;
-                    mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, lowPrefixPtr) + 4;
-                    offset = current - matchIndex;
-                    while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
-                }
-                offset_2 = offset_1;
-                offset_1 = offset;
-                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
-
-            } else {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
-                continue;
-        }   }
-
-        /* found a match : store it */
-        ip += mLength;
-        anchor = ip;
-
-        if (ip <= ilimit) {
-            /* Fill Table */
-            hashSmall[ZSTD_hashPtr(base+current+2, hBitsS, mls)] = current+2;
-            hashLong[ZSTD_hashPtr(base+current+2, hBitsL, 8)] = current+2;
-            hashSmall[ZSTD_hashPtr(ip-2, hBitsS, mls)] = (U32)(ip-2-base);
-            hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
-            /* check immediate repcode */
-            while (ip <= ilimit) {
-                U32 const current2 = (U32)(ip-base);
-                U32 const repIndex2 = current2 - offset_2;
-                const BYTE* repMatch2 = repIndex2 < dictLimit ? dictBase + repIndex2 : base + repIndex2;
-                if ( (((U32)((dictLimit-1) - repIndex2) >= 3) & (repIndex2 > lowestIndex))  /* intentional overflow */
-                   && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
-                    const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
-                    size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
-                    U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
-                    hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
-                    hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
-                    ip += repLength2;
-                    anchor = ip;
-                    continue;
-                }
-                break;
-    }   }   }
-
-    /* save reps for next block */
-    ctx->repToConfirm[0] = offset_1; ctx->repToConfirm[1] = offset_2;
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
-}
-
-
-static void ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx,
-                         const void* src, size_t srcSize)
-{
-    U32 const mls = ctx->params.cParams.searchLength;
-    switch(mls)
-    {
-    default: /* includes case 3 */
-    case 4 :
-        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 4); return;
-    case 5 :
-        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 5); return;
-    case 6 :
-        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 6); return;
-    case 7 :
-        ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 7); return;
-    }
-}
-
-
-/*-*************************************
-*  Binary Tree search
-***************************************/
-/** ZSTD_insertBt1() : add one or multiple positions to tree.
-*   ip : assumed <= iend-8 .
-*   @return : nb of positions added */
-static U32 ZSTD_insertBt1(ZSTD_CCtx* zc, const BYTE* const ip, const U32 mls, const BYTE* const iend, U32 nbCompares,
-                          U32 extDict)
-{
-    U32*   const hashTable = zc->hashTable;
-    U32    const hashLog = zc->params.cParams.hashLog;
-    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
-    U32*   const bt = zc->chainTable;
-    U32    const btLog  = zc->params.cParams.chainLog - 1;
-    U32    const btMask = (1 << btLog) - 1;
-    U32 matchIndex = hashTable[h];
-    size_t commonLengthSmaller=0, commonLengthLarger=0;
-    const BYTE* const base = zc->base;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const BYTE* match;
-    const U32 current = (U32)(ip-base);
-    const U32 btLow = btMask >= current ? 0 : current - btMask;
-    U32* smallerPtr = bt + 2*(current&btMask);
-    U32* largerPtr  = smallerPtr + 1;
-    U32 dummy32;   /* to be nullified at the end */
-    U32 const windowLow = zc->lowLimit;
-    U32 matchEndIdx = current+8;
-    size_t bestLength = 8;
-#ifdef ZSTD_C_PREDICT
-    U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
-    U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
-    predictedSmall += (predictedSmall>0);
-    predictedLarge += (predictedLarge>0);
-#endif /* ZSTD_C_PREDICT */
-
-    hashTable[h] = current;   /* Update Hash Table */
-
-    while (nbCompares-- && (matchIndex > windowLow)) {
-        U32* const nextPtr = bt + 2*(matchIndex & btMask);
-        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
-
-#ifdef ZSTD_C_PREDICT   /* note : can create issues when hlog small <= 11 */
-        const U32* predictPtr = bt + 2*((matchIndex-1) & btMask);   /* written this way, as bt is a roll buffer */
-        if (matchIndex == predictedSmall) {
-            /* no need to check length, result known */
-            *smallerPtr = matchIndex;
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-            predictedSmall = predictPtr[1] + (predictPtr[1]>0);
-            continue;
-        }
-        if (matchIndex == predictedLarge) {
-            *largerPtr = matchIndex;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-            predictedLarge = predictPtr[0] + (predictPtr[0]>0);
-            continue;
-        }
-#endif
-        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
-            match = base + matchIndex;
-            if (match[matchLength] == ip[matchLength])
-                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
-        } else {
-            match = dictBase + matchIndex;
-            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
-            if (matchIndex+matchLength >= dictLimit)
-                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
-        }
-
-        if (matchLength > bestLength) {
-            bestLength = matchLength;
-            if (matchLength > matchEndIdx - matchIndex)
-                matchEndIdx = matchIndex + (U32)matchLength;
-        }
-
-        if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
-            break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt the tree */
-
-        if (match[matchLength] < ip[matchLength]) {  /* necessarily within correct buffer */
-            /* match is smaller than current */
-            *smallerPtr = matchIndex;             /* update smaller idx */
-            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-        } else {
-            /* match is larger than current */
-            *largerPtr = matchIndex;
-            commonLengthLarger = matchLength;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-    }   }
-
-    *smallerPtr = *largerPtr = 0;
-    if (bestLength > 384) return MIN(192, (U32)(bestLength - 384));   /* speed optimization */
-    if (matchEndIdx > current + 8) return matchEndIdx - current - 8;
-    return 1;
-}
-
-
-static size_t ZSTD_insertBtAndFindBestMatch (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iend,
-                        size_t* offsetPtr,
-                        U32 nbCompares, const U32 mls,
-                        U32 extDict)
-{
-    U32*   const hashTable = zc->hashTable;
-    U32    const hashLog = zc->params.cParams.hashLog;
-    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
-    U32*   const bt = zc->chainTable;
-    U32    const btLog  = zc->params.cParams.chainLog - 1;
-    U32    const btMask = (1 << btLog) - 1;
-    U32 matchIndex  = hashTable[h];
-    size_t commonLengthSmaller=0, commonLengthLarger=0;
-    const BYTE* const base = zc->base;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const U32 current = (U32)(ip-base);
-    const U32 btLow = btMask >= current ? 0 : current - btMask;
-    const U32 windowLow = zc->lowLimit;
-    U32* smallerPtr = bt + 2*(current&btMask);
-    U32* largerPtr  = bt + 2*(current&btMask) + 1;
-    U32 matchEndIdx = current+8;
-    U32 dummy32;   /* to be nullified at the end */
-    size_t bestLength = 0;
-
-    hashTable[h] = current;   /* Update Hash Table */
-
-    while (nbCompares-- && (matchIndex > windowLow)) {
-        U32* const nextPtr = bt + 2*(matchIndex & btMask);
-        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
-        const BYTE* match;
-
-        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
-            match = base + matchIndex;
-            if (match[matchLength] == ip[matchLength])
-                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
-        } else {
-            match = dictBase + matchIndex;
-            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
-            if (matchIndex+matchLength >= dictLimit)
-                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
-        }
-
-        if (matchLength > bestLength) {
-            if (matchLength > matchEndIdx - matchIndex)
-                matchEndIdx = matchIndex + (U32)matchLength;
-            if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
-                bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
-            if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
-                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
-        }
-
-        if (match[matchLength] < ip[matchLength]) {
-            /* match is smaller than current */
-            *smallerPtr = matchIndex;             /* update smaller idx */
-            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-        } else {
-            /* match is larger than current */
-            *largerPtr = matchIndex;
-            commonLengthLarger = matchLength;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-    }   }
-
-    *smallerPtr = *largerPtr = 0;
-
-    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
-    return bestLength;
-}
-
-
-static void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
-{
-    const BYTE* const base = zc->base;
-    const U32 target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
-
-    while(idx < target)
-        idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 0);
-}
-
-/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
-static size_t ZSTD_BtFindBestMatch (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 mls)
-{
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 0);
-}
-
-
-static size_t ZSTD_BtFindBestMatch_selectMLS (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
-{
-    switch(matchLengthSearch)
-    {
-    default : /* includes case 3 */
-    case 4 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
-    case 5 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
-    case 7 :
-    case 6 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
-    }
-}
-
-
-static void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
-{
-    const BYTE* const base = zc->base;
-    const U32 target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
-
-    while (idx < target) idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 1);
-}
-
-
-/** Tree updater, providing best match */
-static size_t ZSTD_BtFindBestMatch_extDict (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 mls)
-{
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 1);
-}
-
-
-static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
-{
-    switch(matchLengthSearch)
-    {
-    default : /* includes case 3 */
-    case 4 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
-    case 5 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
-    case 7 :
-    case 6 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
-    }
-}
-
-
-
-/* *********************************
-*  Hash Chain
-***********************************/
-#define NEXT_IN_CHAIN(d, mask)   chainTable[(d) & mask]
-
-/* Update chains up to ip (excluded)
-   Assumption : always within prefix (i.e. not within extDict) */
-FORCE_INLINE
-U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
-{
-    U32* const hashTable  = zc->hashTable;
-    const U32 hashLog = zc->params.cParams.hashLog;
-    U32* const chainTable = zc->chainTable;
-    const U32 chainMask = (1 << zc->params.cParams.chainLog) - 1;
-    const BYTE* const base = zc->base;
-    const U32 target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
-
-    while(idx < target) { /* catch up */
-        size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
-        NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
-        hashTable[h] = idx;
-        idx++;
-    }
-
-    zc->nextToUpdate = target;
-    return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
-}
-
-
-
-FORCE_INLINE /* inlining is important to hardwire a hot branch (template emulation) */
-size_t ZSTD_HcFindBestMatch_generic (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 mls, const U32 extDict)
-{
-    U32* const chainTable = zc->chainTable;
-    const U32 chainSize = (1 << zc->params.cParams.chainLog);
-    const U32 chainMask = chainSize-1;
-    const BYTE* const base = zc->base;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const U32 lowLimit = zc->lowLimit;
-    const U32 current = (U32)(ip-base);
-    const U32 minChain = current > chainSize ? current - chainSize : 0;
-    int nbAttempts=maxNbAttempts;
-    size_t ml=4-1;
-
-    /* HC4 match finder */
-    U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls);
-
-    for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
-        const BYTE* match;
-        size_t currentMl=0;
-        if ((!extDict) || matchIndex >= dictLimit) {
-            match = base + matchIndex;
-            if (match[ml] == ip[ml])   /* potentially better */
-                currentMl = ZSTD_count(ip, match, iLimit);
-        } else {
-            match = dictBase + matchIndex;
-            if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
-                currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
-        }
-
-        /* save best solution */
-        if (currentMl > ml) {
-            ml = currentMl;
-            *offsetPtr = current - matchIndex + ZSTD_REP_MOVE;
-            if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
-        }
-
-        if (matchIndex <= minChain) break;
-        matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
-    }
-
-    return ml;
-}
-
-
-FORCE_INLINE size_t ZSTD_HcFindBestMatch_selectMLS (
-                        ZSTD_CCtx* zc,
-                        const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
-{
-    switch(matchLengthSearch)
-    {
-    default : /* includes case 3 */
-    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 0);
-    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 0);
-    case 7 :
-    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 0);
+    {   size_t const streamSize = ZSTD_encodeSequences(op, oend - op,
+                CTable_MatchLength, mlCodeTable,
+                CTable_OffsetBits, ofCodeTable,
+                CTable_LitLength, llCodeTable,
+                sequences, nbSeq, longOffsets);
+        if (ZSTD_isError(streamSize)) return streamSize;
+        op += streamSize;
     }
-}
-
-
-FORCE_INLINE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
-                        ZSTD_CCtx* zc,
-                        const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
-{
-    switch(matchLengthSearch)
-    {
-    default : /* includes case 3 */
-    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 1);
-    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 1);
-    case 7 :
-    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 1);
-    }
-}
-
-
-/* *******************************
-*  Common parser - lazy strategy
-*********************************/
-FORCE_INLINE
-void ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
-                                     const void* src, size_t srcSize,
-                                     const U32 searchMethod, const U32 depth)
-{
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base + ctx->dictLimit;
-
-    U32 const maxSearches = 1 << ctx->params.cParams.searchLog;
-    U32 const mls = ctx->params.cParams.searchLength;
-
-    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
-                        size_t* offsetPtr,
-                        U32 maxNbAttempts, U32 matchLengthSearch);
-    searchMax_f const searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS;
-    U32 offset_1 = ctx->rep[0], offset_2 = ctx->rep[1], savedOffset=0;
-
-    /* init */
-    ip += (ip==base);
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
-    {   U32 const maxRep = (U32)(ip-base);
-        if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
-        if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
-    }
-
-    /* Match Loop */
-    while (ip < ilimit) {
-        size_t matchLength=0;
-        size_t offset=0;
-        const BYTE* start=ip+1;
-
-        /* check repCode */
-        if ((offset_1>0) & (MEM_read32(ip+1) == MEM_read32(ip+1 - offset_1))) {
-            /* repcode : we take it */
-            matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
-            if (depth==0) goto _storeSequence;
-        }
 
-        /* first search (depth 0) */
-        {   size_t offsetFound = 99999999;
-            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
-            if (ml2 > matchLength)
-                matchLength = ml2, start = ip, offset=offsetFound;
-        }
-
-        if (matchLength < 4) {
-            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
-            continue;
-        }
-
-        /* let's try to find a better solution */
-        if (depth>=1)
-        while (ip<ilimit) {
-            ip ++;
-            if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
-                size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
-                int const gain2 = (int)(mlRep * 3);
-                int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
-                if ((mlRep >= 4) && (gain2 > gain1))
-                    matchLength = mlRep, offset = 0, start = ip;
-            }
-            {   size_t offset2=99999999;
-                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
-                int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
-                int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
-                if ((ml2 >= 4) && (gain2 > gain1)) {
-                    matchLength = ml2, offset = offset2, start = ip;
-                    continue;   /* search a better one */
-            }   }
-
-            /* let's find an even better one */
-            if ((depth==2) && (ip<ilimit)) {
-                ip ++;
-                if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
-                    size_t const ml2 = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
-                    int const gain2 = (int)(ml2 * 4);
-                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
-                    if ((ml2 >= 4) && (gain2 > gain1))
-                        matchLength = ml2, offset = 0, start = ip;
-                }
-                {   size_t offset2=99999999;
-                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
-                    int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
-                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
-                    if ((ml2 >= 4) && (gain2 > gain1)) {
-                        matchLength = ml2, offset = offset2, start = ip;
-                        continue;
-            }   }   }
-            break;  /* nothing found : store previous solution */
-        }
-
-        /* catch up */
-        if (offset) {
-            while ( (start > anchor)
-                 && (start > base+offset-ZSTD_REP_MOVE)
-                 && (start[-1] == start[-1-offset+ZSTD_REP_MOVE]) )  /* only search for offset within prefix */
-                { start--; matchLength++; }
-            offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
-        }
-
-        /* store sequence */
-_storeSequence:
-        {   size_t const litLength = start - anchor;
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
-            anchor = ip = start + matchLength;
-        }
-
-        /* check immediate repcode */
-        while ( (ip <= ilimit)
-             && ((offset_2>0)
-             & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
-            /* store sequence */
-            matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
-            offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
-            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
-            ip += matchLength;
-            anchor = ip;
-            continue;   /* faster when present ... (?) */
-    }   }
-
-    /* Save reps for next block */
-    ctx->repToConfirm[0] = offset_1 ? offset_1 : savedOffset;
-    ctx->repToConfirm[1] = offset_2 ? offset_2 : savedOffset;
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
-}
-
-
-static void ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 1, 2);
-}
-
-static void ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 2);
-}
-
-static void ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 1);
-}
-
-static void ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 0);
+    return op - ostart;
 }
 
-
-FORCE_INLINE
-void ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
-                                     const void* src, size_t srcSize,
-                                     const U32 searchMethod, const U32 depth)
+MEM_STATIC size_t ZSTD_compressSequences(seqStore_t* seqStorePtr,
+                              ZSTD_entropyCTables_t* entropy,
+                              ZSTD_compressionParameters const* cParams,
+                              void* dst, size_t dstCapacity,
+                              size_t srcSize)
 {
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base;
-    const U32 dictLimit = ctx->dictLimit;
-    const U32 lowestIndex = ctx->lowLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const dictEnd  = dictBase + dictLimit;
-    const BYTE* const dictStart  = dictBase + ctx->lowLimit;
-
-    const U32 maxSearches = 1 << ctx->params.cParams.searchLog;
-    const U32 mls = ctx->params.cParams.searchLength;
-
-    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
-                        size_t* offsetPtr,
-                        U32 maxNbAttempts, U32 matchLengthSearch);
-    searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS;
-
-    U32 offset_1 = ctx->rep[0], offset_2 = ctx->rep[1];
-
-    /* init */
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
-    ip += (ip == prefixStart);
-
-    /* Match Loop */
-    while (ip < ilimit) {
-        size_t matchLength=0;
-        size_t offset=0;
-        const BYTE* start=ip+1;
-        U32 current = (U32)(ip-base);
-
-        /* check repCode */
-        {   const U32 repIndex = (U32)(current+1 - offset_1);
-            const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-            const BYTE* const repMatch = repBase + repIndex;
-            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))   /* intentional overflow */
-            if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
-                /* repcode detected we should take it */
-                const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4;
-                if (depth==0) goto _storeSequence;
-        }   }
-
-        /* first search (depth 0) */
-        {   size_t offsetFound = 99999999;
-            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
-            if (ml2 > matchLength)
-                matchLength = ml2, start = ip, offset=offsetFound;
-        }
-
-         if (matchLength < 4) {
-            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
-            continue;
-        }
-
-        /* let's try to find a better solution */
-        if (depth>=1)
-        while (ip<ilimit) {
-            ip ++;
-            current++;
-            /* check repCode */
-            if (offset) {
-                const U32 repIndex = (U32)(current - offset_1);
-                const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-                const BYTE* const repMatch = repBase + repIndex;
-                if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
-                if (MEM_read32(ip) == MEM_read32(repMatch)) {
-                    /* repcode detected */
-                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                    size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
-                    int const gain2 = (int)(repLength * 3);
-                    int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
-                    if ((repLength >= 4) && (gain2 > gain1))
-                        matchLength = repLength, offset = 0, start = ip;
-            }   }
-
-            /* search match, depth 1 */
-            {   size_t offset2=99999999;
-                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
-                int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
-                int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
-                if ((ml2 >= 4) && (gain2 > gain1)) {
-                    matchLength = ml2, offset = offset2, start = ip;
-                    continue;   /* search a better one */
-            }   }
-
-            /* let's find an even better one */
-            if ((depth==2) && (ip<ilimit)) {
-                ip ++;
-                current++;
-                /* check repCode */
-                if (offset) {
-                    const U32 repIndex = (U32)(current - offset_1);
-                    const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-                    const BYTE* const repMatch = repBase + repIndex;
-                    if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
-                    if (MEM_read32(ip) == MEM_read32(repMatch)) {
-                        /* repcode detected */
-                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                        size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
-                        int const gain2 = (int)(repLength * 4);
-                        int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
-                        if ((repLength >= 4) && (gain2 > gain1))
-                            matchLength = repLength, offset = 0, start = ip;
-                }   }
-
-                /* search match, depth 2 */
-                {   size_t offset2=99999999;
-                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
-                    int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
-                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
-                    if ((ml2 >= 4) && (gain2 > gain1)) {
-                        matchLength = ml2, offset = offset2, start = ip;
-                        continue;
-            }   }   }
-            break;  /* nothing found : store previous solution */
-        }
-
-        /* catch up */
-        if (offset) {
-            U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
-            const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
-            const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
-            while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; }  /* catch up */
-            offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
-        }
-
-        /* store sequence */
-_storeSequence:
-        {   size_t const litLength = start - anchor;
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
-            anchor = ip = start + matchLength;
-        }
-
-        /* check immediate repcode */
-        while (ip <= ilimit) {
-            const U32 repIndex = (U32)((ip-base) - offset_2);
-            const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-            const BYTE* const repMatch = repBase + repIndex;
-            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
-            if (MEM_read32(ip) == MEM_read32(repMatch)) {
-                /* repcode detected we should take it */
-                const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
-                offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset history */
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
-                ip += matchLength;
-                anchor = ip;
-                continue;   /* faster when present ... (?) */
-            }
-            break;
-    }   }
-
-    /* Save reps for next block */
-    ctx->repToConfirm[0] = offset_1; ctx->repToConfirm[1] = offset_2;
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
+    size_t const cSize = ZSTD_compressSequences_internal(seqStorePtr, entropy, cParams,
+                                                         dst, dstCapacity);
+    size_t const minGain = ZSTD_minGain(srcSize);
+    size_t const maxCSize = srcSize - minGain;
+    /* If the srcSize <= dstCapacity, then there is enough space to write a
+     * raw uncompressed block. Since we ran out of space, the block must not
+     * be compressible, so fall back to a raw uncompressed block.
+     */
+    int const uncompressibleError = cSize == ERROR(dstSize_tooSmall) && srcSize <= dstCapacity;
+
+    if (ZSTD_isError(cSize) && !uncompressibleError)
+        return cSize;
+    /* Check compressibility */
+    if (cSize >= maxCSize || uncompressibleError) {
+        entropy->hufCTable_repeatMode = HUF_repeat_none;
+        entropy->offcode_repeatMode = FSE_repeat_none;
+        entropy->matchlength_repeatMode = FSE_repeat_none;
+        entropy->litlength_repeatMode = FSE_repeat_none;
+        return 0;
     }
-}
-
-
-void ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 0);
-}
-
-static void ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 1);
-}
-
-static void ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 2);
-}
-
-static void ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-    ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 1, 2);
-}
-
-
-/* The optimal parser */
-#include "zstd_opt.h"
-
-static void ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-#ifdef ZSTD_OPT_H_91842398743
-    ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0);
-#else
-    (void)ctx; (void)src; (void)srcSize;
-    return;
-#endif
-}
+    assert(!ZSTD_isError(cSize));
 
-static void ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-#ifdef ZSTD_OPT_H_91842398743
-    ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 1);
-#else
-    (void)ctx; (void)src; (void)srcSize;
-    return;
-#endif
-}
-
-static void ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-#ifdef ZSTD_OPT_H_91842398743
-    ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 0);
-#else
-    (void)ctx; (void)src; (void)srcSize;
-    return;
-#endif
-}
-
-static void ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
-{
-#ifdef ZSTD_OPT_H_91842398743
-    ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 1);
-#else
-    (void)ctx; (void)src; (void)srcSize;
-    return;
-#endif
+    /* confirm repcodes */
+    { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->rep[i] = seqStorePtr->repToConfirm[i]; }
+    return cSize;
 }
 
-
-typedef void (*ZSTD_blockCompressor) (ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-
-static ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict)
+/* ZSTD_selectBlockCompressor() :
+ * Not static, but internal use only (used by long distance matcher)
+ * assumption : strat is a valid strategy */
+typedef size_t (*ZSTD_blockCompressor) (ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict)
 {
-    static const ZSTD_blockCompressor blockCompressor[2][8] = {
-        { ZSTD_compressBlock_fast, ZSTD_compressBlock_doubleFast, ZSTD_compressBlock_greedy,
+    static const ZSTD_blockCompressor blockCompressor[2][(unsigned)ZSTD_btultra+1] = {
+        { ZSTD_compressBlock_fast  /* default for 0 */,
+          ZSTD_compressBlock_fast, ZSTD_compressBlock_doubleFast, ZSTD_compressBlock_greedy,
           ZSTD_compressBlock_lazy, ZSTD_compressBlock_lazy2, ZSTD_compressBlock_btlazy2,
           ZSTD_compressBlock_btopt, ZSTD_compressBlock_btultra },
-        { ZSTD_compressBlock_fast_extDict, ZSTD_compressBlock_doubleFast_extDict, ZSTD_compressBlock_greedy_extDict,
+        { ZSTD_compressBlock_fast_extDict  /* default for 0 */,
+          ZSTD_compressBlock_fast_extDict, ZSTD_compressBlock_doubleFast_extDict, ZSTD_compressBlock_greedy_extDict,
           ZSTD_compressBlock_lazy_extDict,ZSTD_compressBlock_lazy2_extDict, ZSTD_compressBlock_btlazy2_extDict,
           ZSTD_compressBlock_btopt_extDict, ZSTD_compressBlock_btultra_extDict }
     };
+    ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);
+    assert((U32)strat >= (U32)ZSTD_fast);
+    assert((U32)strat <= (U32)ZSTD_btultra);
 
-    return blockCompressor[extDict][(U32)strat];
+    return blockCompressor[extDict!=0][(U32)strat];
 }
 
+static void ZSTD_storeLastLiterals(seqStore_t* seqStorePtr,
+                                   const BYTE* anchor, size_t lastLLSize)
+{
+    memcpy(seqStorePtr->lit, anchor, lastLLSize);
+    seqStorePtr->lit += lastLLSize;
+}
 
 static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
 {
-    ZSTD_blockCompressor const blockCompressor = ZSTD_selectBlockCompressor(zc->params.cParams.strategy, zc->lowLimit < zc->dictLimit);
     const BYTE* const base = zc->base;
     const BYTE* const istart = (const BYTE*)src;
     const U32 current = (U32)(istart-base);
+    size_t lastLLSize;
+    const BYTE* anchor;
+    U32 const extDict = zc->lowLimit < zc->dictLimit;
+    const ZSTD_blockCompressor blockCompressor =
+        zc->appliedParams.ldmParams.enableLdm
+            ? (extDict ? ZSTD_compressBlock_ldm_extDict : ZSTD_compressBlock_ldm)
+            : ZSTD_selectBlockCompressor(zc->appliedParams.cParams.strategy, extDict);
+
     if (srcSize < MIN_CBLOCK_SIZE+ZSTD_blockHeaderSize+1) return 0;   /* don't even attempt compression below a certain srcSize */
     ZSTD_resetSeqStore(&(zc->seqStore));
     if (current > zc->nextToUpdate + 384)
         zc->nextToUpdate = current - MIN(192, (U32)(current - zc->nextToUpdate - 384));   /* limited update after finding a very long match */
-    blockCompressor(zc, src, srcSize);
-    return ZSTD_compressSequences(zc, dst, dstCapacity, srcSize);
+
+    lastLLSize = blockCompressor(zc, src, srcSize);
+
+    /* Last literals */
+    anchor = (const BYTE*)src + srcSize - lastLLSize;
+    ZSTD_storeLastLiterals(&zc->seqStore, anchor, lastLLSize);
+
+    return ZSTD_compressSequences(&zc->seqStore, zc->entropy, &zc->appliedParams.cParams, dst, dstCapacity, srcSize);
 }
 
 
-/*! ZSTD_compress_generic() :
+/*! ZSTD_compress_frameChunk() :
 *   Compress a chunk of data into one or multiple blocks.
 *   All blocks will be terminated, all input will be consumed.
 *   Function will issue an error if there is not enough `dstCapacity` to hold the compressed content.
 *   Frame is supposed already started (header already produced)
 *   @return : compressed size, or an error code
 */
-static size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
+static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
                                      void* dst, size_t dstCapacity,
                                const void* src, size_t srcSize,
                                      U32 lastFrameChunk)
@@ -2485,9 +1586,9 @@ static size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
     const BYTE* ip = (const BYTE*)src;
     BYTE* const ostart = (BYTE*)dst;
     BYTE* op = ostart;
-    U32 const maxDist = 1 << cctx->params.cParams.windowLog;
+    U32 const maxDist = 1 << cctx->appliedParams.cParams.windowLog;
 
-    if (cctx->params.fParams.checksumFlag && srcSize)
+    if (cctx->appliedParams.fParams.checksumFlag && srcSize)
         XXH64_update(&cctx->xxhState, src, srcSize);
 
     while (remaining) {
@@ -2500,9 +1601,9 @@ static size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
 
         /* preemptive overflow correction */
         if (cctx->lowLimit > (3U<<29)) {
-            U32 const cycleMask = (1 << ZSTD_cycleLog(cctx->params.cParams.hashLog, cctx->params.cParams.strategy)) - 1;
+            U32 const cycleMask = (1 << ZSTD_cycleLog(cctx->appliedParams.cParams.hashLog, cctx->appliedParams.cParams.strategy)) - 1;
             U32 const current = (U32)(ip - cctx->base);
-            U32 const newCurrent = (current & cycleMask) + (1 << cctx->params.cParams.windowLog);
+            U32 const newCurrent = (current & cycleMask) + (1 << cctx->appliedParams.cParams.windowLog);
             U32 const correction = current - newCurrent;
             ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_64 <= 30);
             ZSTD_reduceIndex(cctx, correction);
@@ -2548,7 +1649,7 @@ static size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
 
 
 static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
-                                    ZSTD_parameters params, U64 pledgedSrcSize, U32 dictID)
+                                    ZSTD_CCtx_params params, U64 pledgedSrcSize, U32 dictID)
 {   BYTE* const op = (BYTE*)dst;
     U32   const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536);   /* 0-3 */
     U32   const dictIDSizeCode = params.fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength;   /* 0-3 */
@@ -2557,13 +1658,12 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
     U32   const singleSegment = params.fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);
     BYTE  const windowLogByte = (BYTE)((params.cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
     U32   const fcsCode = params.fParams.contentSizeFlag ?
-                     (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) :   /* 0-3 */
-                      0;
+                     (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) : 0;  /* 0-3 */
     BYTE  const frameHeaderDecriptionByte = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) );
     size_t pos;
 
     if (dstCapacity < ZSTD_frameHeaderSize_max) return ERROR(dstSize_tooSmall);
-    DEBUGLOG(5, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u \n",
+    DEBUGLOG(5, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",
                 !params.fParams.noDictIDFlag, dictID,  dictIDSizeCode);
 
     MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
@@ -2571,7 +1671,7 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
     if (!singleSegment) op[pos++] = windowLogByte;
     switch(dictIDSizeCode)
     {
-        default:   /* impossible */
+        default:  assert(0); /* impossible */
         case 0 : break;
         case 1 : op[pos] = (BYTE)(dictID); pos++; break;
         case 2 : MEM_writeLE16(op+pos, (U16)dictID); pos+=2; break;
@@ -2579,7 +1679,7 @@ static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
     }
     switch(fcsCode)
     {
-        default:   /* impossible */
+        default:  assert(0); /* impossible */
         case 0 : if (singleSegment) op[pos++] = (BYTE)(pledgedSrcSize); break;
         case 1 : MEM_writeLE16(op+pos, (U16)(pledgedSrcSize-256)); pos+=2; break;
         case 2 : MEM_writeLE32(op+pos, (U32)(pledgedSrcSize)); pos+=4; break;
@@ -2597,10 +1697,13 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
     const BYTE* const ip = (const BYTE*) src;
     size_t fhSize = 0;
 
+    DEBUGLOG(5, "ZSTD_compressContinue_internal");
+    DEBUGLOG(5, "stage: %u", cctx->stage);
     if (cctx->stage==ZSTDcs_created) return ERROR(stage_wrong);   /* missing init (ZSTD_compressBegin) */
 
     if (frame && (cctx->stage==ZSTDcs_init)) {
-        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->params, cctx->frameContentSize, cctx->dictID);
+        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->appliedParams,
+                                       cctx->pledgedSrcSizePlusOne-1, cctx->dictID);
         if (ZSTD_isError(fhSize)) return fhSize;
         dstCapacity -= fhSize;
         dst = (char*)dst + fhSize;
@@ -2630,7 +1733,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
 
     if (srcSize) {
         size_t const cSize = frame ?
-                             ZSTD_compress_generic (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
+                             ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
                              ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize);
         if (ZSTD_isError(cSize)) return cSize;
         cctx->consumedSrcSize += srcSize;
@@ -2639,7 +1742,6 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
         return fhSize;
 }
 
-
 size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,
                               void* dst, size_t dstCapacity,
                         const void* src, size_t srcSize)
@@ -2648,14 +1750,16 @@ size_t ZSTD_compressContinue (ZSTD_CCtx* cctx,
 }
 
 
-size_t ZSTD_getBlockSizeMax(ZSTD_CCtx* cctx)
+size_t ZSTD_getBlockSize(const ZSTD_CCtx* cctx)
 {
-    return MIN (ZSTD_BLOCKSIZE_ABSOLUTEMAX, 1 << cctx->params.cParams.windowLog);
+    ZSTD_compressionParameters const cParams =
+            ZSTD_getCParamsFromCCtxParams(cctx->appliedParams, 0, 0);
+    return MIN (ZSTD_BLOCKSIZE_MAX, 1 << cParams.windowLog);
 }
 
 size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
 {
-    size_t const blockSizeMax = ZSTD_getBlockSizeMax(cctx);
+    size_t const blockSizeMax = ZSTD_getBlockSize(cctx);
     if (srcSize > blockSizeMax) return ERROR(srcSize_wrong);
     return ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize, 0 /* frame mode */, 0 /* last chunk */);
 }
@@ -2674,37 +1778,36 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_CCtx* zc, const void* src, size_t
     zc->dictBase = zc->base;
     zc->base += ip - zc->nextSrc;
     zc->nextToUpdate = zc->dictLimit;
-    zc->loadedDictEnd = zc->forceWindow ? 0 : (U32)(iend - zc->base);
+    zc->loadedDictEnd = zc->appliedParams.forceWindow ? 0 : (U32)(iend - zc->base);
 
     zc->nextSrc = iend;
     if (srcSize <= HASH_READ_SIZE) return 0;
 
-    switch(zc->params.cParams.strategy)
+    switch(zc->appliedParams.cParams.strategy)
     {
     case ZSTD_fast:
-        ZSTD_fillHashTable (zc, iend, zc->params.cParams.searchLength);
+        ZSTD_fillHashTable (zc, iend, zc->appliedParams.cParams.searchLength);
         break;
-
     case ZSTD_dfast:
-        ZSTD_fillDoubleHashTable (zc, iend, zc->params.cParams.searchLength);
+        ZSTD_fillDoubleHashTable (zc, iend, zc->appliedParams.cParams.searchLength);
         break;
 
     case ZSTD_greedy:
     case ZSTD_lazy:
     case ZSTD_lazy2:
         if (srcSize >= HASH_READ_SIZE)
-            ZSTD_insertAndFindFirstIndex(zc, iend-HASH_READ_SIZE, zc->params.cParams.searchLength);
+            ZSTD_insertAndFindFirstIndex(zc, iend-HASH_READ_SIZE, zc->appliedParams.cParams.searchLength);
         break;
 
     case ZSTD_btlazy2:
     case ZSTD_btopt:
     case ZSTD_btultra:
         if (srcSize >= HASH_READ_SIZE)
-            ZSTD_updateTree(zc, iend-HASH_READ_SIZE, iend, 1 << zc->params.cParams.searchLog, zc->params.cParams.searchLength);
+            ZSTD_updateTree(zc, iend-HASH_READ_SIZE, iend, 1 << zc->appliedParams.cParams.searchLog, zc->appliedParams.cParams.searchLength);
         break;
 
     default:
-        return ERROR(GENERIC);   /* strategy doesn't exist; impossible */
+        assert(0);  /* not possible : not a valid strategy id */
     }
 
     zc->nextToUpdate = (U32)(iend - zc->base);
@@ -2741,13 +1844,14 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_CCtx* cctx, const void* dict, size_t
     const BYTE* const dictEnd = dictPtr + dictSize;
     short offcodeNCount[MaxOff+1];
     unsigned offcodeMaxValue = MaxOff;
-    BYTE scratchBuffer[1<<MAX(MLFSELog,LLFSELog)];
+
+    ZSTD_STATIC_ASSERT(sizeof(cctx->entropy->workspace) >= (1<<MAX(MLFSELog,LLFSELog)));
 
     dictPtr += 4;   /* skip magic number */
-    cctx->dictID = cctx->params.fParams.noDictIDFlag ? 0 :  MEM_readLE32(dictPtr);
+    cctx->dictID = cctx->appliedParams.fParams.noDictIDFlag ? 0 :  MEM_readLE32(dictPtr);
     dictPtr += 4;
 
-    {   size_t const hufHeaderSize = HUF_readCTable(cctx->hufCTable, 255, dictPtr, dictEnd-dictPtr);
+    {   size_t const hufHeaderSize = HUF_readCTable((HUF_CElt*)cctx->entropy->hufCTable, 255, dictPtr, dictEnd-dictPtr);
         if (HUF_isError(hufHeaderSize)) return ERROR(dictionary_corrupted);
         dictPtr += hufHeaderSize;
     }
@@ -2757,7 +1861,7 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_CCtx* cctx, const void* dict, size_t
         if (FSE_isError(offcodeHeaderSize)) return ERROR(dictionary_corrupted);
         if (offcodeLog > OffFSELog) return ERROR(dictionary_corrupted);
         /* Defer checking offcodeMaxValue because we need to know the size of the dictionary content */
-        CHECK_E( FSE_buildCTable_wksp(cctx->offcodeCTable, offcodeNCount, offcodeMaxValue, offcodeLog, scratchBuffer, sizeof(scratchBuffer)),
+        CHECK_E( FSE_buildCTable_wksp(cctx->entropy->offcodeCTable, offcodeNCount, offcodeMaxValue, offcodeLog, cctx->entropy->workspace, sizeof(cctx->entropy->workspace)),
                  dictionary_corrupted);
         dictPtr += offcodeHeaderSize;
     }
@@ -2769,7 +1873,7 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_CCtx* cctx, const void* dict, size_t
         if (matchlengthLog > MLFSELog) return ERROR(dictionary_corrupted);
         /* Every match length code must have non-zero probability */
         CHECK_F( ZSTD_checkDictNCount(matchlengthNCount, matchlengthMaxValue, MaxML));
-        CHECK_E( FSE_buildCTable_wksp(cctx->matchlengthCTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog, scratchBuffer, sizeof(scratchBuffer)),
+        CHECK_E( FSE_buildCTable_wksp(cctx->entropy->matchlengthCTable, matchlengthNCount, matchlengthMaxValue, matchlengthLog, cctx->entropy->workspace, sizeof(cctx->entropy->workspace)),
                  dictionary_corrupted);
         dictPtr += matchlengthHeaderSize;
     }
@@ -2781,15 +1885,15 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_CCtx* cctx, const void* dict, size_t
         if (litlengthLog > LLFSELog) return ERROR(dictionary_corrupted);
         /* Every literal length code must have non-zero probability */
         CHECK_F( ZSTD_checkDictNCount(litlengthNCount, litlengthMaxValue, MaxLL));
-        CHECK_E( FSE_buildCTable_wksp(cctx->litlengthCTable, litlengthNCount, litlengthMaxValue, litlengthLog, scratchBuffer, sizeof(scratchBuffer)),
+        CHECK_E( FSE_buildCTable_wksp(cctx->entropy->litlengthCTable, litlengthNCount, litlengthMaxValue, litlengthLog, cctx->entropy->workspace, sizeof(cctx->entropy->workspace)),
                  dictionary_corrupted);
         dictPtr += litlengthHeaderSize;
     }
 
     if (dictPtr+12 > dictEnd) return ERROR(dictionary_corrupted);
-    cctx->rep[0] = MEM_readLE32(dictPtr+0);
-    cctx->rep[1] = MEM_readLE32(dictPtr+4);
-    cctx->rep[2] = MEM_readLE32(dictPtr+8);
+    cctx->seqStore.rep[0] = MEM_readLE32(dictPtr+0);
+    cctx->seqStore.rep[1] = MEM_readLE32(dictPtr+4);
+    cctx->seqStore.rep[2] = MEM_readLE32(dictPtr+8);
     dictPtr += 12;
 
     {   size_t const dictContentSize = (size_t)(dictEnd - dictPtr);
@@ -2803,42 +1907,85 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_CCtx* cctx, const void* dict, size_t
         /* All repCodes must be <= dictContentSize and != 0*/
         {   U32 u;
             for (u=0; u<3; u++) {
-                if (cctx->rep[u] == 0) return ERROR(dictionary_corrupted);
-                if (cctx->rep[u] > dictContentSize) return ERROR(dictionary_corrupted);
+                if (cctx->seqStore.rep[u] == 0) return ERROR(dictionary_corrupted);
+                if (cctx->seqStore.rep[u] > dictContentSize) return ERROR(dictionary_corrupted);
         }   }
 
-        cctx->fseCTables_ready = 1;
-        cctx->hufCTable_repeatMode = HUF_repeat_valid;
+        cctx->entropy->hufCTable_repeatMode = HUF_repeat_valid;
+        cctx->entropy->offcode_repeatMode = FSE_repeat_valid;
+        cctx->entropy->matchlength_repeatMode = FSE_repeat_valid;
+        cctx->entropy->litlength_repeatMode = FSE_repeat_valid;
         return ZSTD_loadDictionaryContent(cctx, dictPtr, dictContentSize);
     }
 }
 
 /** ZSTD_compress_insertDictionary() :
 *   @return : 0, or an error code */
-static size_t ZSTD_compress_insertDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize)
+static size_t ZSTD_compress_insertDictionary(ZSTD_CCtx* cctx,
+                                       const void* dict, size_t dictSize,
+                                             ZSTD_dictMode_e dictMode)
 {
+    DEBUGLOG(5, "ZSTD_compress_insertDictionary");
     if ((dict==NULL) || (dictSize<=8)) return 0;
 
-    /* dict as pure content */
-    if ((MEM_readLE32(dict) != ZSTD_DICT_MAGIC) || (cctx->forceRawDict))
+    /* dict restricted modes */
+    if (dictMode==ZSTD_dm_rawContent)
         return ZSTD_loadDictionaryContent(cctx, dict, dictSize);
 
-    /* dict as zstd dictionary */
+    if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) {
+        if (dictMode == ZSTD_dm_auto) {
+            DEBUGLOG(5, "raw content dictionary detected");
+            return ZSTD_loadDictionaryContent(cctx, dict, dictSize);
+        }
+        if (dictMode == ZSTD_dm_fullDict)
+            return ERROR(dictionary_wrong);
+        assert(0);   /* impossible */
+    }
+
+    /* dict as full zstd dictionary */
     return ZSTD_loadZstdDictionary(cctx, dict, dictSize);
 }
 
 /*! ZSTD_compressBegin_internal() :
-*   @return : 0, or an error code */
+ * @return : 0, or an error code */
 static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
                              const void* dict, size_t dictSize,
-                                   ZSTD_parameters params, U64 pledgedSrcSize)
-{
-    ZSTD_compResetPolicy_e const crp = dictSize ? ZSTDcrp_fullReset : ZSTDcrp_continue;
+                             ZSTD_dictMode_e dictMode,
+                             const ZSTD_CDict* cdict,
+                                   ZSTD_CCtx_params params, U64 pledgedSrcSize,
+                                   ZSTD_buffered_policy_e zbuff)
+{
+    DEBUGLOG(4, "ZSTD_compressBegin_internal");
+    DEBUGLOG(4, "dict ? %s", dict ? "dict" : (cdict ? "cdict" : "none"));
+    DEBUGLOG(4, "dictMode : %u", (U32)dictMode);
+    /* params are supposed to be fully validated at this point */
     assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
-    CHECK_F(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize, crp));
-    return ZSTD_compress_insertDictionary(cctx, dict, dictSize);
+    assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
+
+    if (cdict && cdict->dictContentSize>0) {
+        return ZSTD_copyCCtx_internal(cctx, cdict->refContext,
+                                      params.fParams, pledgedSrcSize,
+                                      zbuff);
+    }
+
+    CHECK_F( ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
+                                     ZSTDcrp_continue, zbuff) );
+    return ZSTD_compress_insertDictionary(cctx, dict, dictSize, dictMode);
 }
 
+size_t ZSTD_compressBegin_advanced_internal(
+                                    ZSTD_CCtx* cctx,
+                                    const void* dict, size_t dictSize,
+                                    ZSTD_dictMode_e dictMode,
+                                    ZSTD_CCtx_params params,
+                                    unsigned long long pledgedSrcSize)
+{
+    /* compression parameters verification and optimization */
+    CHECK_F( ZSTD_checkCParams(params.cParams) );
+    return ZSTD_compressBegin_internal(cctx, dict, dictSize, dictMode, NULL,
+                                       params, pledgedSrcSize,
+                                       ZSTDb_not_buffered);
+}
 
 /*! ZSTD_compressBegin_advanced() :
 *   @return : 0, or an error code */
@@ -2846,19 +1993,22 @@ size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx,
                              const void* dict, size_t dictSize,
                                    ZSTD_parameters params, unsigned long long pledgedSrcSize)
 {
-    /* compression parameters verification and optimization */
-    CHECK_F(ZSTD_checkCParams(params.cParams));
-    return ZSTD_compressBegin_internal(cctx, dict, dictSize, params, pledgedSrcSize);
+    ZSTD_CCtx_params const cctxParams =
+            ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+    return ZSTD_compressBegin_advanced_internal(cctx, dict, dictSize, ZSTD_dm_auto,
+                                                cctxParams,
+                                                pledgedSrcSize);
 }
 
-
 size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel)
 {
     ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, dictSize);
-    return ZSTD_compressBegin_internal(cctx, dict, dictSize, params, 0);
+    ZSTD_CCtx_params const cctxParams =
+            ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+    return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dm_auto, NULL,
+                                       cctxParams, 0, ZSTDb_not_buffered);
 }
 
-
 size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel)
 {
     return ZSTD_compressBegin_usingDict(cctx, NULL, 0, compressionLevel);
@@ -2874,12 +2024,12 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
     BYTE* op = ostart;
     size_t fhSize = 0;
 
-    DEBUGLOG(5, "ZSTD_writeEpilogue \n");
+    DEBUGLOG(5, "ZSTD_writeEpilogue");
     if (cctx->stage == ZSTDcs_created) return ERROR(stage_wrong);  /* init missing */
 
     /* special case : empty frame */
     if (cctx->stage == ZSTDcs_init) {
-        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->params, 0, 0);
+        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->appliedParams, 0, 0);
         if (ZSTD_isError(fhSize)) return fhSize;
         dstCapacity -= fhSize;
         op += fhSize;
@@ -2895,7 +2045,7 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
         dstCapacity -= ZSTD_blockHeaderSize;
     }
 
-    if (cctx->params.fParams.checksumFlag) {
+    if (cctx->appliedParams.fParams.checksumFlag) {
         U32 const checksum = (U32) XXH64_digest(&cctx->xxhState);
         if (dstCapacity<4) return ERROR(dstSize_tooSmall);
         MEM_writeLE32(op, checksum);
@@ -2912,15 +2062,19 @@ size_t ZSTD_compressEnd (ZSTD_CCtx* cctx,
                    const void* src, size_t srcSize)
 {
     size_t endResult;
-    size_t const cSize = ZSTD_compressContinue_internal(cctx, dst, dstCapacity, src, srcSize,
-                               1 /* frame mode */, 1 /* last chunk */);
+    size_t const cSize = ZSTD_compressContinue_internal(cctx,
+                                dst, dstCapacity, src, srcSize,
+                                1 /* frame mode */, 1 /* last chunk */);
     if (ZSTD_isError(cSize)) return cSize;
     endResult = ZSTD_writeEpilogue(cctx, (char*)dst + cSize, dstCapacity-cSize);
     if (ZSTD_isError(endResult)) return endResult;
-    if (cctx->params.fParams.contentSizeFlag) {  /* control src size */
-        if (cctx->frameContentSize != cctx->consumedSrcSize)
+    if (cctx->appliedParams.fParams.contentSizeFlag) {  /* control src size */
+        DEBUGLOG(5, "end of frame : controlling src size");
+        if (cctx->pledgedSrcSizePlusOne != cctx->consumedSrcSize+1) {
+            DEBUGLOG(5, "error : pledgedSrcSize = %u, while realSrcSize = %u",
+                (U32)cctx->pledgedSrcSizePlusOne-1, (U32)cctx->consumedSrcSize);
             return ERROR(srcSize_wrong);
-    }
+    }   }
     return cSize + endResult;
 }
 
@@ -2931,8 +2085,13 @@ static size_t ZSTD_compress_internal (ZSTD_CCtx* cctx,
                          const void* dict,size_t dictSize,
                                ZSTD_parameters params)
 {
-    CHECK_F(ZSTD_compressBegin_internal(cctx, dict, dictSize, params, srcSize));
-    return ZSTD_compressEnd(cctx, dst,  dstCapacity, src, srcSize);
+    ZSTD_CCtx_params const cctxParams =
+            ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+    return ZSTD_compress_advanced_internal(cctx,
+                                          dst, dstCapacity,
+                                          src, srcSize,
+                                          dict, dictSize,
+                                          cctxParams);
 }
 
 size_t ZSTD_compress_advanced (ZSTD_CCtx* ctx,
@@ -2945,6 +2104,19 @@ size_t ZSTD_compress_advanced (ZSTD_CCtx* ctx,
     return ZSTD_compress_internal(ctx, dst, dstCapacity, src, srcSize, dict, dictSize, params);
 }
 
+/* Internal */
+size_t ZSTD_compress_advanced_internal(
+        ZSTD_CCtx* cctx,
+        void* dst, size_t dstCapacity,
+        const void* src, size_t srcSize,
+        const void* dict,size_t dictSize,
+        ZSTD_CCtx_params params)
+{
+    CHECK_F( ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dm_auto, NULL,
+                                         params, srcSize, ZSTDb_not_buffered) );
+    return ZSTD_compressEnd(cctx, dst, dstCapacity, src, srcSize);
+}
+
 size_t ZSTD_compress_usingDict(ZSTD_CCtx* ctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize,
                                const void* dict, size_t dictSize, int compressionLevel)
 {
@@ -2963,51 +2135,83 @@ size_t ZSTD_compress(void* dst, size_t dstCapacity, const void* src, size_t srcS
     size_t result;
     ZSTD_CCtx ctxBody;
     memset(&ctxBody, 0, sizeof(ctxBody));
-    memcpy(&ctxBody.customMem, &defaultCustomMem, sizeof(ZSTD_customMem));
+    ctxBody.customMem = ZSTD_defaultCMem;
     result = ZSTD_compressCCtx(&ctxBody, dst, dstCapacity, src, srcSize, compressionLevel);
-    ZSTD_free(ctxBody.workSpace, defaultCustomMem);  /* can't free ctxBody itself, as it's on stack; free only heap content */
+    ZSTD_free(ctxBody.workSpace, ZSTD_defaultCMem);  /* can't free ctxBody itself, as it's on stack; free only heap content */
     return result;
 }
 
 
 /* =====  Dictionary API  ===== */
 
-struct ZSTD_CDict_s {
-    void* dictBuffer;
-    const void* dictContent;
-    size_t dictContentSize;
-    ZSTD_CCtx* refContext;
-};  /* typedef'd tp ZSTD_CDict within "zstd.h" */
-
-/*! ZSTD_estimateCDictSize() :
+/*! ZSTD_estimateCDictSize_advanced() :
  *  Estimate amount of memory that will be needed to create a dictionary with following arguments */
-size_t ZSTD_estimateCDictSize(ZSTD_compressionParameters cParams, size_t dictSize)
+size_t ZSTD_estimateCDictSize_advanced(
+        size_t dictSize, ZSTD_compressionParameters cParams,
+        ZSTD_dictLoadMethod_e dictLoadMethod)
+{
+    DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (U32)sizeof(ZSTD_CDict));
+    DEBUGLOG(5, "CCtx estimate : %u",
+             (U32)ZSTD_estimateCCtxSize_advanced_usingCParams(cParams));
+    return sizeof(ZSTD_CDict) + ZSTD_estimateCCtxSize_advanced_usingCParams(cParams)
+           + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+}
+
+size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)
 {
-    cParams = ZSTD_adjustCParams(cParams, 0, dictSize);
-    return sizeof(ZSTD_CDict) + dictSize + ZSTD_estimateCCtxSize(cParams);
+    ZSTD_compressionParameters const cParams = ZSTD_getCParams(compressionLevel, 0, dictSize);
+    return ZSTD_estimateCDictSize_advanced(dictSize, cParams, ZSTD_dlm_byCopy);
 }
 
 size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)
 {
     if (cdict==NULL) return 0;   /* support sizeof on NULL */
+    DEBUGLOG(5, "sizeof(*cdict) : %u", (U32)sizeof(*cdict));
+    DEBUGLOG(5, "ZSTD_sizeof_CCtx : %u", (U32)ZSTD_sizeof_CCtx(cdict->refContext));
     return ZSTD_sizeof_CCtx(cdict->refContext) + (cdict->dictBuffer ? cdict->dictContentSize : 0) + sizeof(*cdict);
 }
 
-static ZSTD_parameters ZSTD_makeParams(ZSTD_compressionParameters cParams, ZSTD_frameParameters fParams)
+static size_t ZSTD_initCDict_internal(
+                    ZSTD_CDict* cdict,
+              const void* dictBuffer, size_t dictSize,
+                    ZSTD_dictLoadMethod_e dictLoadMethod,
+                    ZSTD_dictMode_e dictMode,
+                    ZSTD_compressionParameters cParams)
 {
-    ZSTD_parameters params;
-    params.cParams = cParams;
-    params.fParams = fParams;
-    return params;
+    DEBUGLOG(5, "ZSTD_initCDict_internal, mode %u", (U32)dictMode);
+    if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) {
+        cdict->dictBuffer = NULL;
+        cdict->dictContent = dictBuffer;
+    } else {
+        void* const internalBuffer = ZSTD_malloc(dictSize, cdict->refContext->customMem);
+        cdict->dictBuffer = internalBuffer;
+        cdict->dictContent = internalBuffer;
+        if (!internalBuffer) return ERROR(memory_allocation);
+        memcpy(internalBuffer, dictBuffer, dictSize);
+    }
+    cdict->dictContentSize = dictSize;
+
+    {   ZSTD_CCtx_params cctxParams = cdict->refContext->requestedParams;
+        cctxParams.cParams = cParams;
+        CHECK_F( ZSTD_compressBegin_internal(cdict->refContext,
+                                        cdict->dictContent, dictSize, dictMode,
+                                        NULL,
+                                        cctxParams, ZSTD_CONTENTSIZE_UNKNOWN,
+                                        ZSTDb_not_buffered) );
+    }
+
+    return 0;
 }
 
-ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize, unsigned byReference,
+ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,
+                                      ZSTD_dictLoadMethod_e dictLoadMethod,
+                                      ZSTD_dictMode_e dictMode,
                                       ZSTD_compressionParameters cParams, ZSTD_customMem customMem)
 {
-    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
-    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+    DEBUGLOG(5, "ZSTD_createCDict_advanced, mode %u", (U32)dictMode);
+    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
 
-    {   ZSTD_CDict* const cdict = (ZSTD_CDict*) ZSTD_malloc(sizeof(ZSTD_CDict), customMem);
+    {   ZSTD_CDict* const cdict = (ZSTD_CDict*)ZSTD_malloc(sizeof(ZSTD_CDict), customMem);
         ZSTD_CCtx* const cctx = ZSTD_createCCtx_advanced(customMem);
 
         if (!cdict || !cctx) {
@@ -3015,46 +2219,33 @@ ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize, u
             ZSTD_freeCCtx(cctx);
             return NULL;
         }
-
-        if ((byReference) || (!dictBuffer) || (!dictSize)) {
-            cdict->dictBuffer = NULL;
-            cdict->dictContent = dictBuffer;
-        } else {
-            void* const internalBuffer = ZSTD_malloc(dictSize, customMem);
-            if (!internalBuffer) { ZSTD_free(cctx, customMem); ZSTD_free(cdict, customMem); return NULL; }
-            memcpy(internalBuffer, dictBuffer, dictSize);
-            cdict->dictBuffer = internalBuffer;
-            cdict->dictContent = internalBuffer;
+        cdict->refContext = cctx;
+        if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
+                                        dictBuffer, dictSize,
+                                        dictLoadMethod, dictMode,
+                                        cParams) )) {
+            ZSTD_freeCDict(cdict);
+            return NULL;
         }
 
-        {   ZSTD_frameParameters const fParams = { 0 /* contentSizeFlag */, 0 /* checksumFlag */, 0 /* noDictIDFlag */ };   /* dummy */
-            ZSTD_parameters const params = ZSTD_makeParams(cParams, fParams);
-            size_t const errorCode = ZSTD_compressBegin_advanced(cctx, cdict->dictContent, dictSize, params, 0);
-            if (ZSTD_isError(errorCode)) {
-                ZSTD_free(cdict->dictBuffer, customMem);
-                ZSTD_free(cdict, customMem);
-                ZSTD_freeCCtx(cctx);
-                return NULL;
-        }   }
-
-        cdict->refContext = cctx;
-        cdict->dictContentSize = dictSize;
         return cdict;
     }
 }
 
 ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel)
 {
-    ZSTD_customMem const allocator = { NULL, NULL, NULL };
     ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, 0, dictSize);
-    return ZSTD_createCDict_advanced(dict, dictSize, 0, cParams, allocator);
+    return ZSTD_createCDict_advanced(dict, dictSize,
+                                     ZSTD_dlm_byCopy, ZSTD_dm_auto,
+                                     cParams, ZSTD_defaultCMem);
 }
 
 ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel)
 {
-    ZSTD_customMem const allocator = { NULL, NULL, NULL };
     ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, 0, dictSize);
-    return ZSTD_createCDict_advanced(dict, dictSize, 1, cParams, allocator);
+    return ZSTD_createCDict_advanced(dict, dictSize,
+                                     ZSTD_dlm_byRef, ZSTD_dm_auto,
+                                     cParams, ZSTD_defaultCMem);
 }
 
 size_t ZSTD_freeCDict(ZSTD_CDict* cdict)
@@ -3068,8 +2259,56 @@ size_t ZSTD_freeCDict(ZSTD_CDict* cdict)
     }
 }
 
-static ZSTD_parameters ZSTD_getParamsFromCDict(const ZSTD_CDict* cdict) {
-    return ZSTD_getParamsFromCCtx(cdict->refContext);
+/*! ZSTD_initStaticCDict_advanced() :
+ *  Generate a digested dictionary in provided memory area.
+ *  workspace: The memory area to emplace the dictionary into.
+ *             Provided pointer must 8-bytes aligned.
+ *             It must outlive dictionary usage.
+ *  workspaceSize: Use ZSTD_estimateCDictSize()
+ *                 to determine how large workspace must be.
+ *  cParams : use ZSTD_getCParams() to transform a compression level
+ *            into its relevants cParams.
+ * @return : pointer to ZSTD_CDict*, or NULL if error (size too small)
+ *  Note : there is no corresponding "free" function.
+ *         Since workspace was allocated externally, it must be freed externally.
+ */
+ZSTD_CDict* ZSTD_initStaticCDict(void* workspace, size_t workspaceSize,
+                           const void* dict, size_t dictSize,
+                                 ZSTD_dictLoadMethod_e dictLoadMethod,
+                                 ZSTD_dictMode_e dictMode,
+                                 ZSTD_compressionParameters cParams)
+{
+    size_t const cctxSize = ZSTD_estimateCCtxSize_advanced_usingCParams(cParams);
+    size_t const neededSize = sizeof(ZSTD_CDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize)
+                            + cctxSize;
+    ZSTD_CDict* const cdict = (ZSTD_CDict*) workspace;
+    void* ptr;
+    DEBUGLOG(5, "(size_t)workspace & 7 : %u", (U32)(size_t)workspace & 7);
+    if ((size_t)workspace & 7) return NULL;  /* 8-aligned */
+    DEBUGLOG(5, "(workspaceSize < neededSize) : (%u < %u) => %u",
+        (U32)workspaceSize, (U32)neededSize, (U32)(workspaceSize < neededSize));
+    if (workspaceSize < neededSize) return NULL;
+
+    if (dictLoadMethod == ZSTD_dlm_byCopy) {
+        memcpy(cdict+1, dict, dictSize);
+        dict = cdict+1;
+        ptr = (char*)workspace + sizeof(ZSTD_CDict) + dictSize;
+    } else {
+        ptr = cdict+1;
+    }
+    cdict->refContext = ZSTD_initStaticCCtx(ptr, cctxSize);
+
+    if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
+                                              dict, dictSize,
+                                              ZSTD_dlm_byRef, dictMode,
+                                              cParams) ))
+        return NULL;
+
+    return cdict;
+}
+
+ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict) {
+    return cdict->refContext->appliedParams.cParams;
 }
 
 /* ZSTD_compressBegin_usingCDict_advanced() :
@@ -3078,16 +2317,17 @@ size_t ZSTD_compressBegin_usingCDict_advanced(
     ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict,
     ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize)
 {
-    if (cdict==NULL) return ERROR(GENERIC);  /* does not support NULL cdict */
-    DEBUGLOG(5, "ZSTD_compressBegin_usingCDict_advanced : dictIDFlag == %u \n", !fParams.noDictIDFlag);
-    if (cdict->dictContentSize)
-        CHECK_F( ZSTD_copyCCtx_internal(cctx, cdict->refContext, fParams, pledgedSrcSize) )
-    else {
-        ZSTD_parameters params = cdict->refContext->params;
+    if (cdict==NULL) return ERROR(dictionary_wrong);
+    {   ZSTD_CCtx_params params = cctx->requestedParams;
+        params.cParams = ZSTD_getCParamsFromCDict(cdict);
         params.fParams = fParams;
-        CHECK_F(ZSTD_compressBegin_internal(cctx, NULL, 0, params, pledgedSrcSize));
+        DEBUGLOG(5, "ZSTD_compressBegin_usingCDict_advanced");
+        return ZSTD_compressBegin_internal(cctx,
+                                           NULL, 0, ZSTD_dm_auto,
+                                           cdict,
+                                           params, pledgedSrcSize,
+                                           ZSTDb_not_buffered);
     }
-    return 0;
 }
 
 /* ZSTD_compressBegin_usingCDict() :
@@ -3096,7 +2336,7 @@ size_t ZSTD_compressBegin_usingCDict_advanced(
 size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict)
 {
     ZSTD_frameParameters const fParams = { 0 /*content*/, 0 /*checksum*/, 0 /*noDictID*/ };
-    DEBUGLOG(5, "ZSTD_compressBegin_usingCDict : dictIDFlag == %u \n", !fParams.noDictIDFlag);
+    DEBUGLOG(5, "ZSTD_compressBegin_usingCDict : dictIDFlag == %u", !fParams.noDictIDFlag);
     return ZSTD_compressBegin_usingCDict_advanced(cctx, cdict, fParams, 0);
 }
 
@@ -3131,12 +2371,16 @@ size_t ZSTD_compress_usingCDict(ZSTD_CCtx* cctx,
 
 ZSTD_CStream* ZSTD_createCStream(void)
 {
-    return ZSTD_createCStream_advanced(defaultCustomMem);
+    return ZSTD_createCStream_advanced(ZSTD_defaultCMem);
 }
 
-ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem)
+ZSTD_CStream* ZSTD_initStaticCStream(void *workspace, size_t workspaceSize)
 {
-    /* CStream and CCtx are now same object */
+    return ZSTD_initStaticCCtx(workspace, workspaceSize);
+}
+
+ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem)
+{   /* CStream and CCtx are now same object */
     return ZSTD_createCCtx_advanced(customMem);
 }
 
@@ -3145,35 +2389,32 @@ size_t ZSTD_freeCStream(ZSTD_CStream* zcs)
     return ZSTD_freeCCtx(zcs);   /* same object */
 }
 
-size_t ZSTD_estimateCStreamSize(ZSTD_compressionParameters cParams)
-{
-    size_t const CCtxSize = ZSTD_estimateCCtxSize(cParams);
-    size_t const blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, (size_t)1 << cParams.windowLog);
-    size_t const inBuffSize = ((size_t)1 << cParams.windowLog) + blockSize;
-    size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1;
-    size_t const streamingSize = inBuffSize + outBuffSize;
-
-    return CCtxSize + streamingSize;
-}
 
 
 /*======   Initialization   ======*/
 
-size_t ZSTD_CStreamInSize(void)  { return ZSTD_BLOCKSIZE_ABSOLUTEMAX; }
+size_t ZSTD_CStreamInSize(void)  { return ZSTD_BLOCKSIZE_MAX; }
 
 size_t ZSTD_CStreamOutSize(void)
 {
-    return ZSTD_compressBound(ZSTD_BLOCKSIZE_ABSOLUTEMAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ;
+    return ZSTD_compressBound(ZSTD_BLOCKSIZE_MAX) + ZSTD_blockHeaderSize + 4 /* 32-bits hash */ ;
 }
 
-static size_t ZSTD_resetCStream_internal(ZSTD_CStream* zcs, ZSTD_parameters params, unsigned long long pledgedSrcSize)
+static size_t ZSTD_resetCStream_internal(ZSTD_CStream* zcs,
+                    const void* dict, size_t dictSize, ZSTD_dictMode_e dictMode,
+                    const ZSTD_CDict* cdict,
+                    const ZSTD_CCtx_params params, unsigned long long pledgedSrcSize)
 {
-    if (zcs->inBuffSize==0) return ERROR(stage_wrong);   /* zcs has not been init at least once => can't reset */
-
-    DEBUGLOG(5, "ZSTD_resetCStream_internal : dictIDFlag == %u \n", !zcs->params.fParams.noDictIDFlag);
+    DEBUGLOG(4, "ZSTD_resetCStream_internal");
+    /* params are supposed to be fully validated at this point */
+    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+    assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
 
-    if (zcs->cdict) CHECK_F(ZSTD_compressBegin_usingCDict_advanced(zcs, zcs->cdict, params.fParams, pledgedSrcSize))
-    else CHECK_F(ZSTD_compressBegin_internal(zcs, NULL, 0, params, pledgedSrcSize));
+    CHECK_F( ZSTD_compressBegin_internal(zcs,
+                                        dict, dictSize, dictMode,
+                                        cdict,
+                                        params, pledgedSrcSize,
+                                        ZSTDb_buffered) );
 
     zcs->inToCompress = 0;
     zcs->inBuffPos = 0;
@@ -3181,215 +2422,261 @@ static size_t ZSTD_resetCStream_internal(ZSTD_CStream* zcs, ZSTD_parameters para
     zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;
     zcs->streamStage = zcss_load;
     zcs->frameEnded = 0;
-    zcs->pledgedSrcSize = pledgedSrcSize;
     return 0;   /* ready to go */
 }
 
 size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize)
 {
-
-    ZSTD_parameters params = zcs->params;
+    ZSTD_CCtx_params params = zcs->requestedParams;
     params.fParams.contentSizeFlag = (pledgedSrcSize > 0);
-    DEBUGLOG(5, "ZSTD_resetCStream : dictIDFlag == %u \n", !zcs->params.fParams.noDictIDFlag);
-    return ZSTD_resetCStream_internal(zcs, params, pledgedSrcSize);
+    params.cParams = ZSTD_getCParamsFromCCtxParams(params, pledgedSrcSize, 0);
+    DEBUGLOG(5, "ZSTD_resetCStream");
+    return ZSTD_resetCStream_internal(zcs, NULL, 0, ZSTD_dm_auto, zcs->cdict, params, pledgedSrcSize);
 }
 
-/* ZSTD_initCStream_internal() :
- * params are supposed validated at this stage
- * and zcs->cdict is supposed to be correct */
-static size_t ZSTD_initCStream_stage2(ZSTD_CStream* zcs,
-                                const ZSTD_parameters params,
-                                unsigned long long pledgedSrcSize)
+/*! ZSTD_initCStream_internal() :
+ *  Note : not static, but hidden (not exposed). Used by zstdmt_compress.c
+ *  Assumption 1 : params are valid
+ *  Assumption 2 : either dict, or cdict, is defined, not both */
+size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
+                    const void* dict, size_t dictSize, const ZSTD_CDict* cdict,
+                    ZSTD_CCtx_params params, unsigned long long pledgedSrcSize)
 {
     assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
-    zcs->blockSize = MIN(ZSTD_BLOCKSIZE_ABSOLUTEMAX, (size_t)1 << params.cParams.windowLog);
-
-    /* allocate buffers */
-    {   size_t const neededInBuffSize = ((size_t)1 << params.cParams.windowLog) + zcs->blockSize;
-        if (zcs->inBuffSize < neededInBuffSize) {
-            zcs->inBuffSize = 0;
-            ZSTD_free(zcs->inBuff, zcs->customMem);
-            zcs->inBuff = (char*)ZSTD_malloc(neededInBuffSize, zcs->customMem);
-            if (zcs->inBuff == NULL) return ERROR(memory_allocation);
-            zcs->inBuffSize = neededInBuffSize;
+    assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
+
+    if (dict && dictSize >= 8) {
+        DEBUGLOG(5, "loading dictionary of size %u", (U32)dictSize);
+        if (zcs->staticSize) {   /* static CCtx : never uses malloc */
+            /* incompatible with internal cdict creation */
+            return ERROR(memory_allocation);
         }
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = ZSTD_createCDict_advanced(dict, dictSize,
+                                            ZSTD_dlm_byCopy, ZSTD_dm_auto,
+                                            params.cParams, zcs->customMem);
+        zcs->cdict = zcs->cdictLocal;
+        if (zcs->cdictLocal == NULL) return ERROR(memory_allocation);
+    } else {
+        if (cdict) {
+            params.cParams = ZSTD_getCParamsFromCDict(cdict);  /* cParams are enforced from cdict */
+        }
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = NULL;
+        zcs->cdict = cdict;
     }
-    if (zcs->outBuffSize < ZSTD_compressBound(zcs->blockSize)+1) {
-        size_t const outBuffSize = ZSTD_compressBound(zcs->blockSize)+1;
-        zcs->outBuffSize = 0;
-        ZSTD_free(zcs->outBuff, zcs->customMem);
-        zcs->outBuff = (char*)ZSTD_malloc(outBuffSize, zcs->customMem);
-        if (zcs->outBuff == NULL) return ERROR(memory_allocation);
-        zcs->outBuffSize = outBuffSize;
-    }
 
-    DEBUGLOG(5, "ZSTD_initCStream_stage2 : dictIDFlag == %u \n", !params.fParams.noDictIDFlag);
-    return ZSTD_resetCStream_internal(zcs, params, pledgedSrcSize);
+    params.compressionLevel = ZSTD_CLEVEL_CUSTOM;
+    zcs->requestedParams = params;
+
+    return ZSTD_resetCStream_internal(zcs, NULL, 0, ZSTD_dm_auto, zcs->cdict, params, pledgedSrcSize);
 }
 
 /* ZSTD_initCStream_usingCDict_advanced() :
  * same as ZSTD_initCStream_usingCDict(), with control over frame parameters */
-size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, unsigned long long pledgedSrcSize, ZSTD_frameParameters fParams)
-{
-    if (!cdict) return ERROR(GENERIC);   /* cannot handle NULL cdict (does not know what to do) */
-    {   ZSTD_parameters params = ZSTD_getParamsFromCDict(cdict);
+size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,
+                                            const ZSTD_CDict* cdict,
+                                            ZSTD_frameParameters fParams,
+                                            unsigned long long pledgedSrcSize)
+{   /* cannot handle NULL cdict (does not know what to do) */
+    if (!cdict) return ERROR(dictionary_wrong);
+    {   ZSTD_CCtx_params params = zcs->requestedParams;
+        params.cParams = ZSTD_getCParamsFromCDict(cdict);
         params.fParams = fParams;
-        zcs->cdict = cdict;
-        return ZSTD_initCStream_stage2(zcs, params, pledgedSrcSize);
+        return ZSTD_initCStream_internal(zcs,
+                                NULL, 0, cdict,
+                                params, pledgedSrcSize);
     }
 }
 
 /* note : cdict must outlive compression session */
 size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)
 {
-    ZSTD_frameParameters const fParams = { 0 /* content */, 0 /* checksum */, 0 /* noDictID */ };
-    return ZSTD_initCStream_usingCDict_advanced(zcs, cdict, 0, fParams);
-}
-
-static size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
-                                const void* dict, size_t dictSize,
-                                ZSTD_parameters params, unsigned long long pledgedSrcSize)
-{
-    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
-    zcs->cdict = NULL;
-
-    if (dict && dictSize >= 8) {
-        ZSTD_freeCDict(zcs->cdictLocal);
-        zcs->cdictLocal = ZSTD_createCDict_advanced(dict, dictSize, 0 /* copy */, params.cParams, zcs->customMem);
-        if (zcs->cdictLocal == NULL) return ERROR(memory_allocation);
-        zcs->cdict = zcs->cdictLocal;
-    }
-
-    DEBUGLOG(5, "ZSTD_initCStream_internal : dictIDFlag == %u \n", !params.fParams.noDictIDFlag);
-    return ZSTD_initCStream_stage2(zcs, params, pledgedSrcSize);
+    ZSTD_frameParameters const fParams = { 0 /* contentSize */, 0 /* checksum */, 0 /* hideDictID */ };
+    return ZSTD_initCStream_usingCDict_advanced(zcs, cdict, fParams, 0);  /* note : will check that cdict != NULL */
 }
 
 size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
                                  const void* dict, size_t dictSize,
                                  ZSTD_parameters params, unsigned long long pledgedSrcSize)
 {
+    ZSTD_CCtx_params const cctxParams =
+            ZSTD_assignParamsToCCtxParams(zcs->requestedParams, params);
     CHECK_F( ZSTD_checkCParams(params.cParams) );
-    DEBUGLOG(5, "ZSTD_initCStream_advanced : pledgedSrcSize == %u \n", (U32)pledgedSrcSize);
-    DEBUGLOG(5, "wlog %u \n", params.cParams.windowLog);
-    return ZSTD_initCStream_internal(zcs, dict, dictSize, params, pledgedSrcSize);
+    return ZSTD_initCStream_internal(zcs, dict, dictSize, NULL, cctxParams, pledgedSrcSize);
 }
 
 size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel)
 {
     ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, dictSize);
-    return ZSTD_initCStream_internal(zcs, dict, dictSize, params, 0);
+    ZSTD_CCtx_params const cctxParams =
+            ZSTD_assignParamsToCCtxParams(zcs->requestedParams, params);
+    return ZSTD_initCStream_internal(zcs, dict, dictSize, NULL, cctxParams, 0);
 }
 
 size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize)
 {
-    ZSTD_parameters params = ZSTD_getParams(compressionLevel, pledgedSrcSize, 0);
-    params.fParams.contentSizeFlag = (pledgedSrcSize>0);
-    return ZSTD_initCStream_internal(zcs, NULL, 0, params, pledgedSrcSize);
+    ZSTD_CCtx_params cctxParams;
+    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, pledgedSrcSize, 0);
+    cctxParams = ZSTD_assignParamsToCCtxParams(zcs->requestedParams, params);
+    cctxParams.fParams.contentSizeFlag = (pledgedSrcSize>0);
+    return ZSTD_initCStream_internal(zcs, NULL, 0, NULL, cctxParams, pledgedSrcSize);
 }
 
 size_t ZSTD_initCStream(ZSTD_CStream* zcs, int compressionLevel)
 {
-    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, 0);
-    return ZSTD_initCStream_internal(zcs, NULL, 0, params, 0);
-}
-
-size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs)
-{
-    return ZSTD_sizeof_CCtx(zcs);   /* same object */
+    return ZSTD_initCStream_srcSize(zcs, compressionLevel, 0);
 }
 
 /*======   Compression   ======*/
 
-typedef enum { zsf_gather, zsf_flush, zsf_end } ZSTD_flush_e;
-
-MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
+MEM_STATIC size_t ZSTD_limitCopy(void* dst, size_t dstCapacity,
+                           const void* src, size_t srcSize)
 {
     size_t const length = MIN(dstCapacity, srcSize);
-    memcpy(dst, src, length);
+    if (length) memcpy(dst, src, length);
     return length;
 }
 
-static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
-                              void* dst, size_t* dstCapacityPtr,
-                        const void* src, size_t* srcSizePtr,
-                              ZSTD_flush_e const flush)
-{
+/** ZSTD_compressStream_generic():
+ *  internal function for all *compressStream*() variants and *compress_generic()
+ * @return : hint size for next input */
+size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
+                                   ZSTD_outBuffer* output,
+                                   ZSTD_inBuffer* input,
+                                   ZSTD_EndDirective const flushMode)
+{
+    const char* const istart = (const char*)input->src;
+    const char* const iend = istart + input->size;
+    const char* ip = istart + input->pos;
+    char* const ostart = (char*)output->dst;
+    char* const oend = ostart + output->size;
+    char* op = ostart + output->pos;
     U32 someMoreWork = 1;
-    const char* const istart = (const char*)src;
-    const char* const iend = istart + *srcSizePtr;
-    const char* ip = istart;
-    char* const ostart = (char*)dst;
-    char* const oend = ostart + *dstCapacityPtr;
-    char* op = ostart;
-
-    DEBUGLOG(5, "ZSTD_compressStream_generic \n");
+
+    /* check expectations */
+    DEBUGLOG(5, "ZSTD_compressStream_generic, flush=%u", (U32)flushMode);
+    assert(zcs->inBuff != NULL);
+    assert(zcs->inBuffSize>0);
+    assert(zcs->outBuff!= NULL);
+    assert(zcs->outBuffSize>0);
+    assert(output->pos <= output->size);
+    assert(input->pos <= input->size);
+
     while (someMoreWork) {
         switch(zcs->streamStage)
         {
-        case zcss_init: return ERROR(init_missing);   /* call ZBUFF_compressInit() first ! */
+        case zcss_init:
+            /* call ZSTD_initCStream() first ! */
+            return ERROR(init_missing);
 
         case zcss_load:
-            /* complete inBuffer */
+            if ( (flushMode == ZSTD_e_end)
+              && ((size_t)(oend-op) >= ZSTD_compressBound(iend-ip))  /* enough dstCapacity */
+              && (zcs->inBuffPos == 0) ) {
+                /* shortcut to compression pass directly into output buffer */
+                size_t const cSize = ZSTD_compressEnd(zcs,
+                                                op, oend-op, ip, iend-ip);
+                DEBUGLOG(4, "ZSTD_compressEnd : %u", (U32)cSize);
+                if (ZSTD_isError(cSize)) return cSize;
+                ip = iend;
+                op += cSize;
+                zcs->frameEnded = 1;
+                ZSTD_startNewCompression(zcs);
+                someMoreWork = 0; break;
+            }
+            /* complete loading into inBuffer */
             {   size_t const toLoad = zcs->inBuffTarget - zcs->inBuffPos;
-                size_t const loaded = ZSTD_limitCopy(zcs->inBuff + zcs->inBuffPos, toLoad, ip, iend-ip);
-                DEBUGLOG(5, "loading %u/%u \n", (U32)loaded, (U32)toLoad);
+                size_t const loaded = ZSTD_limitCopy(
+                                        zcs->inBuff + zcs->inBuffPos, toLoad,
+                                        ip, iend-ip);
                 zcs->inBuffPos += loaded;
                 ip += loaded;
-                if ( (zcs->inBuffPos==zcs->inToCompress) || (!flush && (toLoad != loaded)) ) {
-                    someMoreWork = 0; break;  /* not enough input to get a full block : stop there, wait for more */
-            }   }
+                if ( (flushMode == ZSTD_e_continue)
+                  && (zcs->inBuffPos < zcs->inBuffTarget) ) {
+                    /* not enough input to fill full block : stop here */
+                    someMoreWork = 0; break;
+                }
+                if ( (flushMode == ZSTD_e_flush)
+                  && (zcs->inBuffPos == zcs->inToCompress) ) {
+                    /* empty */
+                    someMoreWork = 0; break;
+                }
+            }
             /* compress current block (note : this stage cannot be stopped in the middle) */
-            DEBUGLOG(5, "stream compression stage (flush==%u)\n", flush);
+            DEBUGLOG(5, "stream compression stage (flushMode==%u)", flushMode);
             {   void* cDst;
                 size_t cSize;
                 size_t const iSize = zcs->inBuffPos - zcs->inToCompress;
                 size_t oSize = oend-op;
+                unsigned const lastBlock = (flushMode == ZSTD_e_end) && (ip==iend);
                 if (oSize >= ZSTD_compressBound(iSize))
-                    cDst = op;   /* compress directly into output buffer (avoid flush stage) */
+                    cDst = op;   /* compress into output buffer, to skip flush stage */
                 else
                     cDst = zcs->outBuff, oSize = zcs->outBuffSize;
-                cSize = (flush == zsf_end) ?
-                        ZSTD_compressEnd(zcs, cDst, oSize, zcs->inBuff + zcs->inToCompress, iSize) :
-                        ZSTD_compressContinue(zcs, cDst, oSize, zcs->inBuff + zcs->inToCompress, iSize);
+                cSize = lastBlock ?
+                        ZSTD_compressEnd(zcs, cDst, oSize,
+                                    zcs->inBuff + zcs->inToCompress, iSize) :
+                        ZSTD_compressContinue(zcs, cDst, oSize,
+                                    zcs->inBuff + zcs->inToCompress, iSize);
                 if (ZSTD_isError(cSize)) return cSize;
-                DEBUGLOG(5, "cSize = %u \n", (U32)cSize);
-                if (flush == zsf_end) zcs->frameEnded = 1;
+                zcs->frameEnded = lastBlock;
                 /* prepare next block */
                 zcs->inBuffTarget = zcs->inBuffPos + zcs->blockSize;
                 if (zcs->inBuffTarget > zcs->inBuffSize)
-                    zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;   /* note : inBuffTarget == blockSize <= inBuffSize */
-                assert(zcs->inBuffTarget <= zcs->inBuffSize);
+                    zcs->inBuffPos = 0, zcs->inBuffTarget = zcs->blockSize;
+                DEBUGLOG(5, "inBuffTarget:%u / inBuffSize:%u",
+                         (U32)zcs->inBuffTarget, (U32)zcs->inBuffSize);
+                if (!lastBlock)
+                    assert(zcs->inBuffTarget <= zcs->inBuffSize);
                 zcs->inToCompress = zcs->inBuffPos;
-                if (cDst == op) { op += cSize; break; }   /* no need to flush */
+                if (cDst == op) {  /* no need to flush */
+                    op += cSize;
+                    if (zcs->frameEnded) {
+                        DEBUGLOG(5, "Frame completed directly in outBuffer");
+                        someMoreWork = 0;
+                        ZSTD_startNewCompression(zcs);
+                    }
+                    break;
+                }
                 zcs->outBuffContentSize = cSize;
                 zcs->outBuffFlushedSize = 0;
-                zcs->streamStage = zcss_flush;   /* pass-through to flush stage */
+                zcs->streamStage = zcss_flush; /* pass-through to flush stage */
             }
 	    /* fall-through */
         case zcss_flush:
-            DEBUGLOG(5, "flush stage \n");
+            DEBUGLOG(5, "flush stage");
             {   size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
-                size_t const flushed = ZSTD_limitCopy(op, oend-op, zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
-                DEBUGLOG(5, "toFlush: %u  ; flushed: %u \n", (U32)toFlush, (U32)flushed);
+                size_t const flushed = ZSTD_limitCopy(op, oend-op,
+                            zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
+                DEBUGLOG(5, "toFlush: %u into %u ==> flushed: %u",
+                            (U32)toFlush, (U32)(oend-op), (U32)flushed);
                 op += flushed;
                 zcs->outBuffFlushedSize += flushed;
-                if (toFlush!=flushed) { someMoreWork = 0; break; }  /* dst too small to store flushed data : stop there */
+                if (toFlush!=flushed) {
+                    /* flush not fully completed, presumably because dst is too small */
+                    assert(op==oend);
+                    someMoreWork = 0;
+                    break;
+                }
                 zcs->outBuffContentSize = zcs->outBuffFlushedSize = 0;
+                if (zcs->frameEnded) {
+                    DEBUGLOG(5, "Frame completed on flush");
+                    someMoreWork = 0;
+                    ZSTD_startNewCompression(zcs);
+                    break;
+                }
                 zcs->streamStage = zcss_load;
                 break;
             }
 
-        case zcss_final:
-            someMoreWork = 0;   /* do nothing */
-            break;
-
-        default:
-            return ERROR(GENERIC);   /* impossible */
+        default: /* impossible */
+            assert(0);
         }
     }
 
-    *srcSizePtr = ip - istart;
-    *dstCapacityPtr = op - ostart;
+    input->pos = ip - istart;
+    output->pos = op - ostart;
     if (zcs->frameEnded) return 0;
     {   size_t hintInSize = zcs->inBuffTarget - zcs->inBuffPos;
         if (hintInSize==0) hintInSize = zcs->blockSize;
@@ -3399,14 +2686,86 @@ static size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
 
 size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
-    size_t sizeRead = input->size - input->pos;
-    size_t sizeWritten = output->size - output->pos;
-    size_t const result = ZSTD_compressStream_generic(zcs,
-                                                      (char*)(output->dst) + output->pos, &sizeWritten,
-                                                      (const char*)(input->src) + input->pos, &sizeRead, zsf_gather);
-    input->pos += sizeRead;
-    output->pos += sizeWritten;
-    return result;
+    /* check conditions */
+    if (output->pos > output->size) return ERROR(GENERIC);
+    if (input->pos  > input->size)  return ERROR(GENERIC);
+
+    return ZSTD_compressStream_generic(zcs, output, input, ZSTD_e_continue);
+}
+
+
+size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
+                              ZSTD_outBuffer* output,
+                              ZSTD_inBuffer* input,
+                              ZSTD_EndDirective endOp)
+{
+    /* check conditions */
+    if (output->pos > output->size) return ERROR(GENERIC);
+    if (input->pos  > input->size)  return ERROR(GENERIC);
+    assert(cctx!=NULL);
+
+    /* transparent initialization stage */
+    if (cctx->streamStage == zcss_init) {
+        ZSTD_prefixDict const prefixDict = cctx->prefixDict;
+        ZSTD_CCtx_params params = cctx->requestedParams;
+        params.cParams = ZSTD_getCParamsFromCCtxParams(
+                cctx->requestedParams, cctx->pledgedSrcSizePlusOne-1, 0 /*dictSize*/);
+        memset(&cctx->prefixDict, 0, sizeof(cctx->prefixDict));  /* single usage */
+        assert(prefixDict.dict==NULL || cctx->cdict==NULL);   /* only one can be set */
+
+#ifdef ZSTD_MULTITHREAD
+        if (params.nbThreads > 1) {
+            if (cctx->mtctx == NULL || cctx->appliedParams.nbThreads != params.nbThreads) {
+                ZSTDMT_freeCCtx(cctx->mtctx);
+                cctx->mtctx = ZSTDMT_createCCtx_advanced(params.nbThreads, cctx->customMem);
+                if (cctx->mtctx == NULL) return ERROR(memory_allocation);
+            }
+            DEBUGLOG(4, "call ZSTDMT_initCStream_internal as nbThreads=%u", params.nbThreads);
+            CHECK_F( ZSTDMT_initCStream_internal(
+                             cctx->mtctx,
+                             prefixDict.dict, prefixDict.dictSize, ZSTD_dm_rawContent,
+                             cctx->cdict, params, cctx->pledgedSrcSizePlusOne-1) );
+            cctx->streamStage = zcss_load;
+            cctx->appliedParams.nbThreads = params.nbThreads;
+        } else
+#endif
+        {
+            CHECK_F( ZSTD_resetCStream_internal(
+                             cctx, prefixDict.dict, prefixDict.dictSize,
+                             prefixDict.dictMode, cctx->cdict, params,
+                             cctx->pledgedSrcSizePlusOne-1) );
+    }   }
+
+    /* compression stage */
+#ifdef ZSTD_MULTITHREAD
+    if (cctx->appliedParams.nbThreads > 1) {
+        size_t const flushMin = ZSTDMT_compressStream_generic(cctx->mtctx, output, input, endOp);
+        DEBUGLOG(5, "ZSTDMT_compressStream_generic : %u", (U32)flushMin);
+        if ( ZSTD_isError(flushMin)
+          || (endOp == ZSTD_e_end && flushMin == 0) ) { /* compression completed */
+            ZSTD_startNewCompression(cctx);
+        }
+        return flushMin;
+    }
+#endif
+    CHECK_F( ZSTD_compressStream_generic(cctx, output, input, endOp) );
+    DEBUGLOG(5, "completed ZSTD_compress_generic");
+    return cctx->outBuffContentSize - cctx->outBuffFlushedSize; /* remaining to flush */
+}
+
+size_t ZSTD_compress_generic_simpleArgs (
+                            ZSTD_CCtx* cctx,
+                            void* dst, size_t dstCapacity, size_t* dstPos,
+                      const void* src, size_t srcSize, size_t* srcPos,
+                            ZSTD_EndDirective endOp)
+{
+    ZSTD_outBuffer output = { dst, dstCapacity, *dstPos };
+    ZSTD_inBuffer  input  = { src, srcSize, *srcPos };
+    /* ZSTD_compress_generic() will check validity of dstPos and srcPos */
+    size_t const cErr = ZSTD_compress_generic(cctx, &output, &input, endOp);
+    *dstPos = output.pos;
+    *srcPos = input.pos;
+    return cErr;
 }
 
 
@@ -3416,90 +2775,58 @@ size_t ZSTD_compressStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output, ZSTD_inBuf
 *   @return : amount of data remaining to flush */
 size_t ZSTD_flushStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
 {
-    size_t srcSize = 0;
-    size_t sizeWritten = output->size - output->pos;
-    size_t const result = ZSTD_compressStream_generic(zcs,
-                                                     (char*)(output->dst) + output->pos, &sizeWritten,
-                                                     &srcSize, &srcSize, /* use a valid src address instead of NULL */
-                                                      zsf_flush);
-    output->pos += sizeWritten;
-    if (ZSTD_isError(result)) return result;
-    return zcs->outBuffContentSize - zcs->outBuffFlushedSize;   /* remaining to flush */
+    ZSTD_inBuffer input = { NULL, 0, 0 };
+    if (output->pos > output->size) return ERROR(GENERIC);
+    CHECK_F( ZSTD_compressStream_generic(zcs, output, &input, ZSTD_e_flush) );
+    return zcs->outBuffContentSize - zcs->outBuffFlushedSize;  /* remaining to flush */
 }
 
 
 size_t ZSTD_endStream(ZSTD_CStream* zcs, ZSTD_outBuffer* output)
 {
-    BYTE* const ostart = (BYTE*)(output->dst) + output->pos;
-    BYTE* const oend = (BYTE*)(output->dst) + output->size;
-    BYTE* op = ostart;
-
-    DEBUGLOG(5, "ZSTD_endStream (dstCapacity : %u) \n", (U32)(oend-op));
-    if (zcs->streamStage != zcss_final) {
-        /* flush whatever remains */
-        size_t srcSize = 0;
-        size_t sizeWritten = output->size - output->pos;
-        size_t const notEnded = ZSTD_compressStream_generic(zcs, ostart, &sizeWritten,
-                                     &srcSize /* use a valid src address instead of NULL */, &srcSize, zsf_end);
-        size_t const remainingToFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
-        op += sizeWritten;
-        if (remainingToFlush) {
-            output->pos += sizeWritten;
-            return remainingToFlush + ZSTD_BLOCKHEADERSIZE /* final empty block */
-                   + ((zcs->params.fParams.checksumFlag > 0) * 4) /* optional 32-bits checksum */;
-        }
-        /* create epilogue */
-        zcs->streamStage = zcss_final;
-        zcs->outBuffContentSize = !notEnded ? 0 :
-            /* write epilogue, including final empty block, into outBuff */
-            ZSTD_compressEnd(zcs, zcs->outBuff, zcs->outBuffSize, NULL, 0);
-        if (ZSTD_isError(zcs->outBuffContentSize)) return zcs->outBuffContentSize;
-    }
-
-    /* flush epilogue */
-    {   size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize;
-        size_t const flushed = ZSTD_limitCopy(op, oend-op, zcs->outBuff + zcs->outBuffFlushedSize, toFlush);
-        op += flushed;
-        zcs->outBuffFlushedSize += flushed;
-        output->pos += op-ostart;
-        if (toFlush==flushed) zcs->streamStage = zcss_init;  /* end reached */
-        return toFlush - flushed;
+    ZSTD_inBuffer input = { NULL, 0, 0 };
+    if (output->pos > output->size) return ERROR(GENERIC);
+    CHECK_F( ZSTD_compressStream_generic(zcs, output, &input, ZSTD_e_end) );
+    {   size_t const lastBlockSize = zcs->frameEnded ? 0 : ZSTD_BLOCKHEADERSIZE;
+        size_t const checksumSize = zcs->frameEnded ? 0 : zcs->appliedParams.fParams.checksumFlag * 4;
+        size_t const toFlush = zcs->outBuffContentSize - zcs->outBuffFlushedSize + lastBlockSize + checksumSize;
+        DEBUGLOG(5, "ZSTD_endStream : remaining to flush : %u",
+                (unsigned)toFlush);
+        return toFlush;
     }
 }
 
 
-
 /*-=====  Pre-defined compression levels  =====-*/
 
-#define ZSTD_DEFAULT_CLEVEL 1
 #define ZSTD_MAX_CLEVEL     22
 int ZSTD_maxCLevel(void) { return ZSTD_MAX_CLEVEL; }
 
 static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEVEL+1] = {
-{   /* "default" */
+{   /* "default" - guarantees a monotonically increasing memory budget */
     /* W,  C,  H,  S,  L, TL, strat */
     { 18, 12, 12,  1,  7, 16, ZSTD_fast    },  /* level  0 - never used */
     { 19, 13, 14,  1,  7, 16, ZSTD_fast    },  /* level  1 */
     { 19, 15, 16,  1,  6, 16, ZSTD_fast    },  /* level  2 */
-    { 20, 16, 17,  1,  5, 16, ZSTD_dfast   },  /* level  3.*/
-    { 20, 18, 18,  1,  5, 16, ZSTD_dfast   },  /* level  4.*/
-    { 20, 15, 18,  3,  5, 16, ZSTD_greedy  },  /* level  5 */
-    { 21, 16, 19,  2,  5, 16, ZSTD_lazy    },  /* level  6 */
-    { 21, 17, 20,  3,  5, 16, ZSTD_lazy    },  /* level  7 */
+    { 20, 16, 17,  1,  5, 16, ZSTD_dfast   },  /* level  3 */
+    { 20, 17, 18,  1,  5, 16, ZSTD_dfast   },  /* level  4 */
+    { 20, 17, 18,  2,  5, 16, ZSTD_greedy  },  /* level  5 */
+    { 21, 17, 19,  2,  5, 16, ZSTD_lazy    },  /* level  6 */
+    { 21, 18, 19,  3,  5, 16, ZSTD_lazy    },  /* level  7 */
     { 21, 18, 20,  3,  5, 16, ZSTD_lazy2   },  /* level  8 */
-    { 21, 20, 20,  3,  5, 16, ZSTD_lazy2   },  /* level  9 */
+    { 21, 19, 20,  3,  5, 16, ZSTD_lazy2   },  /* level  9 */
     { 21, 19, 21,  4,  5, 16, ZSTD_lazy2   },  /* level 10 */
     { 22, 20, 22,  4,  5, 16, ZSTD_lazy2   },  /* level 11 */
     { 22, 20, 22,  5,  5, 16, ZSTD_lazy2   },  /* level 12 */
     { 22, 21, 22,  5,  5, 16, ZSTD_lazy2   },  /* level 13 */
     { 22, 21, 22,  6,  5, 16, ZSTD_lazy2   },  /* level 14 */
-    { 22, 21, 21,  5,  5, 16, ZSTD_btlazy2 },  /* level 15 */
+    { 22, 21, 22,  5,  5, 16, ZSTD_btlazy2 },  /* level 15 */
     { 23, 22, 22,  5,  5, 16, ZSTD_btlazy2 },  /* level 16 */
-    { 23, 21, 22,  4,  5, 24, ZSTD_btopt   },  /* level 17 */
+    { 23, 22, 22,  4,  5, 24, ZSTD_btopt   },  /* level 17 */
     { 23, 22, 22,  5,  4, 32, ZSTD_btopt   },  /* level 18 */
     { 23, 23, 22,  6,  3, 48, ZSTD_btopt   },  /* level 19 */
     { 25, 25, 23,  7,  3, 64, ZSTD_btultra },  /* level 20 */
-    { 26, 26, 23,  7,  3,256, ZSTD_btultra },  /* level 21 */
+    { 26, 26, 24,  7,  3,256, ZSTD_btultra },  /* level 21 */
     { 27, 27, 25,  9,  3,512, ZSTD_btultra },  /* level 22 */
 },
 {   /* for srcSize <= 256 KB */
@@ -3582,33 +2909,54 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
 },
 };
 
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
+/* This function just controls
+ * the monotonic memory budget increase of ZSTD_defaultCParameters[0].
+ * Run once, on first ZSTD_getCParams() usage, if ZSTD_DEBUG is enabled
+ */
+MEM_STATIC void ZSTD_check_compressionLevel_monotonicIncrease_memoryBudget(void)
+{
+    int level;
+    for (level=1; level<ZSTD_maxCLevel(); level++) {
+        ZSTD_compressionParameters const c1 = ZSTD_defaultCParameters[0][level];
+        ZSTD_compressionParameters const c2 = ZSTD_defaultCParameters[0][level+1];
+        assert(c1.windowLog <= c2.windowLog);
+#       define ZSTD_TABLECOST(h,c) ((1<<(h)) + (1<<(c)))
+        assert(ZSTD_TABLECOST(c1.hashLog, c1.chainLog) <= ZSTD_TABLECOST(c2.hashLog, c2.chainLog));
+    }
+}
+#endif
+
 /*! ZSTD_getCParams() :
 *   @return ZSTD_compressionParameters structure for a selected compression level, `srcSize` and `dictSize`.
 *   Size values are optional, provide 0 if not known or unused */
-ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSize, size_t dictSize)
+ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize)
 {
-    ZSTD_compressionParameters cp;
-    size_t const addedSize = srcSize ? 0 : 500;
-    U64 const rSize = srcSize+dictSize ? srcSize+dictSize+addedSize : (U64)-1;
+    size_t const addedSize = srcSizeHint ? 0 : 500;
+    U64 const rSize = srcSizeHint+dictSize ? srcSizeHint+dictSize+addedSize : (U64)-1;
     U32 const tableID = (rSize <= 256 KB) + (rSize <= 128 KB) + (rSize <= 16 KB);   /* intentional underflow for srcSizeHint == 0 */
-    if (compressionLevel <= 0) compressionLevel = ZSTD_DEFAULT_CLEVEL;   /* 0 == default; no negative compressionLevel yet */
-    if (compressionLevel > ZSTD_MAX_CLEVEL) compressionLevel = ZSTD_MAX_CLEVEL;
-    cp = ZSTD_defaultCParameters[tableID][compressionLevel];
-    if (MEM_32bits()) {   /* auto-correction, for 32-bits mode */
-        if (cp.windowLog > ZSTD_WINDOWLOG_MAX) cp.windowLog = ZSTD_WINDOWLOG_MAX;
-        if (cp.chainLog > ZSTD_CHAINLOG_MAX) cp.chainLog = ZSTD_CHAINLOG_MAX;
-        if (cp.hashLog > ZSTD_HASHLOG_MAX) cp.hashLog = ZSTD_HASHLOG_MAX;
+
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
+    static int g_monotonicTest = 1;
+    if (g_monotonicTest) {
+        ZSTD_check_compressionLevel_monotonicIncrease_memoryBudget();
+        g_monotonicTest=0;
     }
-    cp = ZSTD_adjustCParams(cp, srcSize, dictSize);
-    return cp;
+#endif
+
+    if (compressionLevel <= 0) compressionLevel = ZSTD_CLEVEL_DEFAULT;   /* 0 == default; no negative compressionLevel yet */
+    if (compressionLevel > ZSTD_MAX_CLEVEL) compressionLevel = ZSTD_MAX_CLEVEL;
+    { ZSTD_compressionParameters const cp = ZSTD_defaultCParameters[tableID][compressionLevel];
+      return ZSTD_adjustCParams_internal(cp, srcSizeHint, dictSize); }
+
 }
 
 /*! ZSTD_getParams() :
 *   same as ZSTD_getCParams(), but @return a `ZSTD_parameters` object (instead of `ZSTD_compressionParameters`).
 *   All fields of `ZSTD_frameParameters` are set to default (0) */
-ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSize, size_t dictSize) {
+ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long srcSizeHint, size_t dictSize) {
     ZSTD_parameters params;
-    ZSTD_compressionParameters const cParams = ZSTD_getCParams(compressionLevel, srcSize, dictSize);
+    ZSTD_compressionParameters const cParams = ZSTD_getCParams(compressionLevel, srcSizeHint, dictSize);
     memset(&params, 0, sizeof(params));
     params.cParams = cParams;
     return params;
diff --git a/contrib/zstd/zstd_compress.h b/contrib/zstd/zstd_compress.h
new file mode 100644
index 000000000..94606edc9
--- /dev/null
+++ b/contrib/zstd/zstd_compress.h
@@ -0,0 +1,307 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+#ifndef ZSTD_COMPRESS_H
+#define ZSTD_COMPRESS_H
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include "zstd_internal.h"
+#ifdef ZSTD_MULTITHREAD
+#  include "zstdmt_compress.h"
+#endif
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-*************************************
+*  Constants
+***************************************/
+static const U32 g_searchStrength = 8;
+#define HASH_READ_SIZE 8
+
+
+/*-*************************************
+*  Context memory management
+***************************************/
+typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
+typedef enum { zcss_init=0, zcss_load, zcss_flush } ZSTD_cStreamStage;
+
+typedef struct ZSTD_prefixDict_s {
+    const void* dict;
+    size_t dictSize;
+    ZSTD_dictMode_e dictMode;
+} ZSTD_prefixDict;
+
+struct ZSTD_CCtx_s {
+    const BYTE* nextSrc;    /* next block here to continue on current prefix */
+    const BYTE* base;       /* All regular indexes relative to this position */
+    const BYTE* dictBase;   /* extDict indexes relative to this position */
+    U32   dictLimit;        /* below that point, need extDict */
+    U32   lowLimit;         /* below that point, no more data */
+    U32   nextToUpdate;     /* index from which to continue dictionary update */
+    U32   nextToUpdate3;    /* index from which to continue dictionary update */
+    U32   hashLog3;         /* dispatch table : larger == faster, more memory */
+    U32   loadedDictEnd;    /* index of end of dictionary */
+    ZSTD_compressionStage_e stage;
+    U32   dictID;
+    ZSTD_CCtx_params requestedParams;
+    ZSTD_CCtx_params appliedParams;
+    void* workSpace;
+    size_t workSpaceSize;
+    size_t blockSize;
+    U64 pledgedSrcSizePlusOne;  /* this way, 0 (default) == unknown */
+    U64 consumedSrcSize;
+    XXH64_state_t xxhState;
+    ZSTD_customMem customMem;
+    size_t staticSize;
+
+    seqStore_t seqStore;    /* sequences storage ptrs */
+    optState_t optState;
+    ldmState_t ldmState;    /* long distance matching state */
+    U32* hashTable;
+    U32* hashTable3;
+    U32* chainTable;
+    ZSTD_entropyCTables_t* entropy;
+
+    /* streaming */
+    char*  inBuff;
+    size_t inBuffSize;
+    size_t inToCompress;
+    size_t inBuffPos;
+    size_t inBuffTarget;
+    char*  outBuff;
+    size_t outBuffSize;
+    size_t outBuffContentSize;
+    size_t outBuffFlushedSize;
+    ZSTD_cStreamStage streamStage;
+    U32    frameEnded;
+
+    /* Dictionary */
+    ZSTD_CDict* cdictLocal;
+    const ZSTD_CDict* cdict;
+    ZSTD_prefixDict prefixDict;   /* single-usage dictionary */
+
+    /* Multi-threading */
+#ifdef ZSTD_MULTITHREAD
+    ZSTDMT_CCtx* mtctx;
+#endif
+};
+
+
+static const BYTE LL_Code[64] = {  0,  1,  2,  3,  4,  5,  6,  7,
+                                   8,  9, 10, 11, 12, 13, 14, 15,
+                                  16, 16, 17, 17, 18, 18, 19, 19,
+                                  20, 20, 20, 20, 21, 21, 21, 21,
+                                  22, 22, 22, 22, 22, 22, 22, 22,
+                                  23, 23, 23, 23, 23, 23, 23, 23,
+                                  24, 24, 24, 24, 24, 24, 24, 24,
+                                  24, 24, 24, 24, 24, 24, 24, 24 };
+
+static const BYTE ML_Code[128] = { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+                                  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+                                  32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
+                                  38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39,
+                                  40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
+                                  41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
+                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42 };
+
+/*! ZSTD_storeSeq() :
+    Store a sequence (literal length, literals, offset code and match length code) into seqStore_t.
+    `offsetCode` : distance to match, or 0 == repCode.
+    `matchCode` : matchLength - MINMATCH
+*/
+MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const void* literals, U32 offsetCode, size_t matchCode)
+{
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG >= 6)
+    static const BYTE* g_start = NULL;
+    U32 const pos = (U32)((const BYTE*)literals - g_start);
+    if (g_start==NULL) g_start = (const BYTE*)literals;
+    if ((pos > 0) && (pos < 1000000000))
+        DEBUGLOG(6, "Cpos %6u :%5u literals & match %3u bytes at distance %6u",
+               pos, (U32)litLength, (U32)matchCode+MINMATCH, (U32)offsetCode);
+#endif
+    /* copy Literals */
+    assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + 128 KB);
+    ZSTD_wildcopy(seqStorePtr->lit, literals, litLength);
+    seqStorePtr->lit += litLength;
+
+    /* literal Length */
+    if (litLength>0xFFFF) {
+        seqStorePtr->longLengthID = 1;
+        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+    }
+    seqStorePtr->sequences[0].litLength = (U16)litLength;
+
+    /* match offset */
+    seqStorePtr->sequences[0].offset = offsetCode + 1;
+
+    /* match Length */
+    if (matchCode>0xFFFF) {
+        seqStorePtr->longLengthID = 2;
+        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+    }
+    seqStorePtr->sequences[0].matchLength = (U16)matchCode;
+
+    seqStorePtr->sequences++;
+}
+
+
+/*-*************************************
+*  Match length counter
+***************************************/
+static unsigned ZSTD_NbCommonBytes (register size_t val)
+{
+    if (MEM_isLittleEndian()) {
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanForward64( &r, (U64)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 4)
+            return (__builtin_ctzll((U64)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2,
+                                                     0, 3, 1, 3, 1, 4, 2, 7,
+                                                     0, 2, 3, 6, 1, 5, 3, 5,
+                                                     1, 3, 4, 4, 2, 5, 6, 7,
+                                                     7, 0, 1, 2, 3, 3, 4, 6,
+                                                     2, 6, 5, 5, 3, 4, 5, 6,
+                                                     7, 1, 2, 4, 6, 4, 4, 5,
+                                                     7, 2, 6, 5, 7, 6, 7, 7 };
+            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r=0;
+            _BitScanForward( &r, (U32)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctz((U32)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0,
+                                                     3, 2, 2, 1, 3, 2, 0, 1,
+                                                     3, 3, 1, 2, 2, 2, 2, 0,
+                                                     3, 1, 2, 0, 1, 0, 1, 1 };
+            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
+#       endif
+        }
+    } else {  /* Big Endian CPU */
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanReverse64( &r, val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 4)
+            return (__builtin_clzll(val) >> 3);
+#       else
+            unsigned r;
+            const unsigned n32 = sizeof(size_t)*4;   /* calculate this way due to compiler complaining in 32-bits mode */
+            if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
+            if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r = 0;
+            _BitScanReverse( &r, (unsigned long)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_clz((U32)val) >> 3);
+#       else
+            unsigned r;
+            if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+    }   }
+}
+
+
+MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* const pInLimit)
+{
+    const BYTE* const pStart = pIn;
+    const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
+
+    while (pIn < pInLoopLimit) {
+        size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+        if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
+        pIn += ZSTD_NbCommonBytes(diff);
+        return (size_t)(pIn - pStart);
+    }
+    if (MEM_64bits()) if ((pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
+    if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
+    if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
+    return (size_t)(pIn - pStart);
+}
+
+/** ZSTD_count_2segments() :
+*   can count match length with `ip` & `match` in 2 different segments.
+*   convention : on reaching mEnd, match count continue starting from iStart
+*/
+MEM_STATIC size_t ZSTD_count_2segments(const BYTE* ip, const BYTE* match, const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
+{
+    const BYTE* const vEnd = MIN( ip + (mEnd - match), iEnd);
+    size_t const matchLength = ZSTD_count(ip, match, vEnd);
+    if (match + matchLength != mEnd) return matchLength;
+    return matchLength + ZSTD_count(ip+matchLength, iStart, iEnd);
+}
+
+
+/*-*************************************
+*  Hashes
+***************************************/
+static const U32 prime3bytes = 506832829U;
+static U32    ZSTD_hash3(U32 u, U32 h) { return ((u << (32-24)) * prime3bytes)  >> (32-h) ; }
+MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); } /* only in zstd_opt.h */
+
+static const U32 prime4bytes = 2654435761U;
+static U32    ZSTD_hash4(U32 u, U32 h) { return (u * prime4bytes) >> (32-h) ; }
+static size_t ZSTD_hash4Ptr(const void* ptr, U32 h) { return ZSTD_hash4(MEM_read32(ptr), h); }
+
+static const U64 prime5bytes = 889523592379ULL;
+static size_t ZSTD_hash5(U64 u, U32 h) { return (size_t)(((u  << (64-40)) * prime5bytes) >> (64-h)) ; }
+static size_t ZSTD_hash5Ptr(const void* p, U32 h) { return ZSTD_hash5(MEM_readLE64(p), h); }
+
+static const U64 prime6bytes = 227718039650203ULL;
+static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u  << (64-48)) * prime6bytes) >> (64-h)) ; }
+static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
+
+static const U64 prime7bytes = 58295818150454627ULL;
+static size_t ZSTD_hash7(U64 u, U32 h) { return (size_t)(((u  << (64-56)) * prime7bytes) >> (64-h)) ; }
+static size_t ZSTD_hash7Ptr(const void* p, U32 h) { return ZSTD_hash7(MEM_readLE64(p), h); }
+
+static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
+static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
+static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
+
+MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
+{
+    switch(mls)
+    {
+    default:
+    case 4: return ZSTD_hash4Ptr(p, hBits);
+    case 5: return ZSTD_hash5Ptr(p, hBits);
+    case 6: return ZSTD_hash6Ptr(p, hBits);
+    case 7: return ZSTD_hash7Ptr(p, hBits);
+    case 8: return ZSTD_hash8Ptr(p, hBits);
+    }
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_COMPRESS_H */
diff --git a/contrib/zstd/zstd_decompress.c b/contrib/zstd/zstd_decompress.c
index 379842b57..91518990e 100644
--- a/contrib/zstd/zstd_decompress.c
+++ b/contrib/zstd/zstd_decompress.c
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 
@@ -54,17 +55,8 @@
 #endif
 
 
-#if defined(_MSC_VER)
-#  include <mmintrin.h>   /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */
-#  define ZSTD_PREFETCH(ptr)   _mm_prefetch((const char*)ptr, _MM_HINT_T0)
-#elif defined(__GNUC__)
-#  define ZSTD_PREFETCH(ptr)   __builtin_prefetch(ptr, 0, 0)
-#else
-#  define ZSTD_PREFETCH(ptr)   /* disabled */
-#endif
-
 /*-*************************************
-*  Macros
+*  Errors
 ***************************************/
 #define ZSTD_isError ERR_isError   /* for inlining */
 #define FSE_isError  ERR_isError
@@ -85,13 +77,17 @@ typedef enum { ZSTDds_getFrameHeaderSize, ZSTDds_decodeFrameHeader,
                ZSTDds_decompressLastBlock, ZSTDds_checkChecksum,
                ZSTDds_decodeSkippableHeader, ZSTDds_skipFrame } ZSTD_dStage;
 
+typedef enum { zdss_init=0, zdss_loadHeader,
+               zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage;
+
 typedef struct {
     FSE_DTable LLTable[FSE_DTABLE_SIZE_U32(LLFSELog)];
     FSE_DTable OFTable[FSE_DTABLE_SIZE_U32(OffFSELog)];
     FSE_DTable MLTable[FSE_DTABLE_SIZE_U32(MLFSELog)];
     HUF_DTable hufTable[HUF_DTABLE_SIZE(HufLog)];  /* can accommodate HUF_decompress4X */
+    U32 workspace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
     U32 rep[ZSTD_REP_NUM];
-} ZSTD_entropyTables_t;
+} ZSTD_entropyDTables_t;
 
 struct ZSTD_DCtx_s
 {
@@ -99,14 +95,15 @@ struct ZSTD_DCtx_s
     const FSE_DTable* MLTptr;
     const FSE_DTable* OFTptr;
     const HUF_DTable* HUFptr;
-    ZSTD_entropyTables_t entropy;
+    ZSTD_entropyDTables_t entropy;
     const void* previousDstEnd;   /* detect continuity */
     const void* base;             /* start of current segment */
     const void* vBase;            /* virtual start of previous segment if it was just before current one */
     const void* dictEnd;          /* end of previous segment */
     size_t expected;
     ZSTD_frameHeader fParams;
-    blockType_e bType;   /* used in ZSTD_decompressContinue(), to transfer blockType between header decoding and block decoding stages */
+    U64 decodedSize;
+    blockType_e bType;            /* used in ZSTD_decompressContinue(), store blockType between block header decoding and block decompression stages */
     ZSTD_dStage stage;
     U32 litEntropy;
     U32 fseEntropy;
@@ -117,11 +114,38 @@ struct ZSTD_DCtx_s
     ZSTD_customMem customMem;
     size_t litSize;
     size_t rleSize;
-    BYTE litBuffer[ZSTD_BLOCKSIZE_ABSOLUTEMAX + WILDCOPY_OVERLENGTH];
+    size_t staticSize;
+
+    /* streaming */
+    ZSTD_DDict* ddictLocal;
+    const ZSTD_DDict* ddict;
+    ZSTD_dStreamStage streamStage;
+    char*  inBuff;
+    size_t inBuffSize;
+    size_t inPos;
+    size_t maxWindowSize;
+    char*  outBuff;
+    size_t outBuffSize;
+    size_t outStart;
+    size_t outEnd;
+    size_t lhSize;
+    void* legacyContext;
+    U32 previousLegacyVersion;
+    U32 legacyVersion;
+    U32 hostageByte;
+
+    /* workspace */
+    BYTE litBuffer[ZSTD_BLOCKSIZE_MAX + WILDCOPY_OVERLENGTH];
     BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];
 };  /* typedef'd to ZSTD_DCtx within "zstd.h" */
 
-size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx) { return (dctx==NULL) ? 0 : sizeof(ZSTD_DCtx); }
+size_t ZSTD_sizeof_DCtx (const ZSTD_DCtx* dctx)
+{
+    if (dctx==NULL) return 0;   /* support sizeof NULL */
+    return sizeof(*dctx)
+           + ZSTD_sizeof_DDict(dctx->ddictLocal)
+           + dctx->inBuffSize + dctx->outBuffSize;
+}
 
 size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
 
@@ -129,6 +153,7 @@ size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
 {
     dctx->expected = ZSTD_frameHeaderSize_prefix;
     dctx->stage = ZSTDds_getFrameHeaderSize;
+    dctx->decodedSize = 0;
     dctx->previousDstEnd = NULL;
     dctx->base = NULL;
     dctx->vBase = NULL;
@@ -145,40 +170,76 @@ size_t ZSTD_decompressBegin(ZSTD_DCtx* dctx)
     return 0;
 }
 
+static void ZSTD_initDCtx_internal(ZSTD_DCtx* dctx)
+{
+    ZSTD_decompressBegin(dctx);   /* cannot fail */
+    dctx->staticSize  = 0;
+    dctx->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
+    dctx->ddict       = NULL;
+    dctx->ddictLocal  = NULL;
+    dctx->inBuff      = NULL;
+    dctx->inBuffSize  = 0;
+    dctx->outBuffSize = 0;
+    dctx->streamStage = zdss_init;
+}
+
 ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem)
 {
-    ZSTD_DCtx* dctx;
+    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
+
+    {   ZSTD_DCtx* const dctx = (ZSTD_DCtx*)ZSTD_malloc(sizeof(*dctx), customMem);
+        if (!dctx) return NULL;
+        dctx->customMem = customMem;
+        dctx->legacyContext = NULL;
+        dctx->previousLegacyVersion = 0;
+        ZSTD_initDCtx_internal(dctx);
+        return dctx;
+    }
+}
+
+ZSTD_DCtx* ZSTD_initStaticDCtx(void *workspace, size_t workspaceSize)
+{
+    ZSTD_DCtx* dctx = (ZSTD_DCtx*) workspace;
 
-    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
-    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+    if ((size_t)workspace & 7) return NULL;  /* 8-aligned */
+    if (workspaceSize < sizeof(ZSTD_DCtx)) return NULL;  /* minimum size */
 
-    dctx = (ZSTD_DCtx*)ZSTD_malloc(sizeof(ZSTD_DCtx), customMem);
-    if (!dctx) return NULL;
-    memcpy(&dctx->customMem, &customMem, sizeof(customMem));
-    ZSTD_decompressBegin(dctx);
+    ZSTD_initDCtx_internal(dctx);
+    dctx->staticSize = workspaceSize;
+    dctx->inBuff = (char*)(dctx+1);
     return dctx;
 }
 
 ZSTD_DCtx* ZSTD_createDCtx(void)
 {
-    return ZSTD_createDCtx_advanced(defaultCustomMem);
+    return ZSTD_createDCtx_advanced(ZSTD_defaultCMem);
 }
 
 size_t ZSTD_freeDCtx(ZSTD_DCtx* dctx)
 {
     if (dctx==NULL) return 0;   /* support free on NULL */
-    ZSTD_free(dctx, dctx->customMem);
-    return 0;   /* reserved as a potential error code in the future */
+    if (dctx->staticSize) return ERROR(memory_allocation);   /* not compatible with static DCtx */
+    {   ZSTD_customMem const cMem = dctx->customMem;
+        ZSTD_freeDDict(dctx->ddictLocal);
+        dctx->ddictLocal = NULL;
+        ZSTD_free(dctx->inBuff, cMem);
+        dctx->inBuff = NULL;
+#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
+        if (dctx->legacyContext)
+            ZSTD_freeLegacyStreamContext(dctx->legacyContext, dctx->previousLegacyVersion);
+#endif
+        ZSTD_free(dctx, cMem);
+        return 0;
+    }
 }
 
+/* no longer useful */
 void ZSTD_copyDCtx(ZSTD_DCtx* dstDCtx, const ZSTD_DCtx* srcDCtx)
 {
-    size_t const workSpaceSize = (ZSTD_BLOCKSIZE_ABSOLUTEMAX+WILDCOPY_OVERLENGTH) + ZSTD_frameHeaderSize_max;
-    memcpy(dstDCtx, srcDCtx, sizeof(ZSTD_DCtx) - workSpaceSize);  /* no need to copy workspace */
+    size_t const toCopy = (size_t)((char*)(&dstDCtx->inBuff) - (char*)dstDCtx);
+    memcpy(dstDCtx, srcDCtx, toCopy);  /* no need to copy workspace */
 }
 
-static void ZSTD_refDDict(ZSTD_DCtx* dstDCtx, const ZSTD_DDict* ddict);
-
 
 /*-*************************************************************
 *   Decompression section
@@ -206,7 +267,7 @@ unsigned ZSTD_isFrame(const void* buffer, size_t size)
 /** ZSTD_frameHeaderSize() :
 *   srcSize must be >= ZSTD_frameHeaderSize_prefix.
 *   @return : size of the Frame Header */
-static size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize)
+size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize)
 {
     if (srcSize < ZSTD_frameHeaderSize_prefix) return ERROR(srcSize_wrong);
     {   BYTE const fhd = ((const BYTE*)src)[4];
@@ -232,18 +293,21 @@ size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t src
     if (MEM_readLE32(src) != ZSTD_MAGICNUMBER) {
         if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
             /* skippable frame */
-            if (srcSize < ZSTD_skippableHeaderSize) return ZSTD_skippableHeaderSize; /* magic number + skippable frame length */
+            if (srcSize < ZSTD_skippableHeaderSize)
+                return ZSTD_skippableHeaderSize; /* magic number + frame length */
             memset(zfhPtr, 0, sizeof(*zfhPtr));
             zfhPtr->frameContentSize = MEM_readLE32((const char *)src + 4);
-            zfhPtr->windowSize = 0; /* windowSize==0 means a frame is skippable */
+            zfhPtr->frameType = ZSTD_skippableFrame;
             return 0;
         }
         return ERROR(prefix_unknown);
     }
 
     /* ensure there is enough `srcSize` to fully read/decode frame header */
-    { size_t const fhsize = ZSTD_frameHeaderSize(src, srcSize);
-      if (srcSize < fhsize) return fhsize; }
+    {   size_t const fhsize = ZSTD_frameHeaderSize(src, srcSize);
+        if (srcSize < fhsize) return fhsize;
+        zfhPtr->headerSize = (U32)fhsize;
+    }
 
     {   BYTE const fhdByte = ip[4];
         size_t pos = 5;
@@ -251,22 +315,23 @@ size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t src
         U32 const checksumFlag = (fhdByte>>2)&1;
         U32 const singleSegment = (fhdByte>>5)&1;
         U32 const fcsID = fhdByte>>6;
-        U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX;
-        U32 windowSize = 0;
+        U64 windowSize = 0;
         U32 dictID = 0;
-        U64 frameContentSize = 0;
-        if ((fhdByte & 0x08) != 0) return ERROR(frameParameter_unsupported);   /* reserved bits, which must be zero */
+        U64 frameContentSize = ZSTD_CONTENTSIZE_UNKNOWN;
+        if ((fhdByte & 0x08) != 0)
+            return ERROR(frameParameter_unsupported); /* reserved bits, must be zero */
+
         if (!singleSegment) {
             BYTE const wlByte = ip[pos++];
             U32 const windowLog = (wlByte >> 3) + ZSTD_WINDOWLOG_ABSOLUTEMIN;
-            if (windowLog > ZSTD_WINDOWLOG_MAX) return ERROR(frameParameter_windowTooLarge);  /* avoids issue with 1 << windowLog */
-            windowSize = (1U << windowLog);
+            if (windowLog > ZSTD_WINDOWLOG_MAX)
+                return ERROR(frameParameter_windowTooLarge);
+            windowSize = (1ULL << windowLog);
             windowSize += (windowSize >> 3) * (wlByte&7);
         }
-
         switch(dictIDSizeCode)
         {
-            default:   /* impossible */
+            default: assert(0);  /* impossible */
             case 0 : break;
             case 1 : dictID = ip[pos]; pos++; break;
             case 2 : dictID = MEM_readLE16(ip+pos); pos+=2; break;
@@ -274,16 +339,18 @@ size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t src
         }
         switch(fcsID)
         {
-            default:   /* impossible */
+            default: assert(0);  /* impossible */
             case 0 : if (singleSegment) frameContentSize = ip[pos]; break;
             case 1 : frameContentSize = MEM_readLE16(ip+pos)+256; break;
             case 2 : frameContentSize = MEM_readLE32(ip+pos); break;
             case 3 : frameContentSize = MEM_readLE64(ip+pos); break;
         }
-        if (!windowSize) windowSize = (U32)frameContentSize;
-        if (windowSize > windowSizeMax) return ERROR(frameParameter_windowTooLarge);
+        if (singleSegment) windowSize = frameContentSize;
+
+        zfhPtr->frameType = ZSTD_frame;
         zfhPtr->frameContentSize = frameContentSize;
         zfhPtr->windowSize = windowSize;
+        zfhPtr->blockSizeMax = (unsigned) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
         zfhPtr->dictID = dictID;
         zfhPtr->checksumFlag = checksumFlag;
     }
@@ -291,10 +358,10 @@ size_t ZSTD_getFrameHeader(ZSTD_frameHeader* zfhPtr, const void* src, size_t src
 }
 
 /** ZSTD_getFrameContentSize() :
-*   compatible with legacy mode
-*   @return : decompressed size of the single frame pointed to be `src` if known, otherwise
-*             - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
-*             - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */
+ *  compatible with legacy mode
+ * @return : decompressed size of the single frame pointed to be `src` if known, otherwise
+ *         - ZSTD_CONTENTSIZE_UNKNOWN if the size cannot be determined
+ *         - ZSTD_CONTENTSIZE_ERROR if an error occurred (e.g. invalid magic number, srcSize too small) */
 unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
 {
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
@@ -303,17 +370,14 @@ unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
         return ret == 0 ? ZSTD_CONTENTSIZE_UNKNOWN : ret;
     }
 #endif
-    {   ZSTD_frameHeader fParams;
-        if (ZSTD_getFrameHeader(&fParams, src, srcSize) != 0) return ZSTD_CONTENTSIZE_ERROR;
-        if (fParams.windowSize == 0) {
-            /* Either skippable or empty frame, size == 0 either way */
+    {   ZSTD_frameHeader zfh;
+        if (ZSTD_getFrameHeader(&zfh, src, srcSize) != 0)
+            return ZSTD_CONTENTSIZE_ERROR;
+        if (zfh.frameType == ZSTD_skippableFrame) {
             return 0;
-        } else if (fParams.frameContentSize != 0) {
-            return fParams.frameContentSize;
         } else {
-            return ZSTD_CONTENTSIZE_UNKNOWN;
-        }
-    }
+            return zfh.frameContentSize;
+    }   }
 }
 
 /** ZSTD_findDecompressedSize() :
@@ -323,58 +387,56 @@ unsigned long long ZSTD_getFrameContentSize(const void *src, size_t srcSize)
  *  @return : decompressed size of the frames contained */
 unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
 {
-    {
-        unsigned long long totalDstSize = 0;
-        while (srcSize >= ZSTD_frameHeaderSize_prefix) {
-            const U32 magicNumber = MEM_readLE32(src);
+    unsigned long long totalDstSize = 0;
 
-            if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
-                size_t skippableSize;
-                if (srcSize < ZSTD_skippableHeaderSize)
-                    return ERROR(srcSize_wrong);
-                skippableSize = MEM_readLE32((const BYTE *)src + 4) +
-                                ZSTD_skippableHeaderSize;
-                if (srcSize < skippableSize) {
-                    return ZSTD_CONTENTSIZE_ERROR;
-                }
-
-                src = (const BYTE *)src + skippableSize;
-                srcSize -= skippableSize;
-                continue;
+    while (srcSize >= ZSTD_frameHeaderSize_prefix) {
+        const U32 magicNumber = MEM_readLE32(src);
+
+        if ((magicNumber & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
+            size_t skippableSize;
+            if (srcSize < ZSTD_skippableHeaderSize)
+                return ERROR(srcSize_wrong);
+            skippableSize = MEM_readLE32((const BYTE *)src + 4) +
+                            ZSTD_skippableHeaderSize;
+            if (srcSize < skippableSize) {
+                return ZSTD_CONTENTSIZE_ERROR;
             }
 
-            {
-                unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
-                if (ret >= ZSTD_CONTENTSIZE_ERROR) return ret;
+            src = (const BYTE *)src + skippableSize;
+            srcSize -= skippableSize;
+            continue;
+        }
 
-                /* check for overflow */
-                if (totalDstSize + ret < totalDstSize) return ZSTD_CONTENTSIZE_ERROR;
-                totalDstSize += ret;
-            }
-            {
-                size_t const frameSrcSize = ZSTD_findFrameCompressedSize(src, srcSize);
-                if (ZSTD_isError(frameSrcSize)) {
-                    return ZSTD_CONTENTSIZE_ERROR;
-                }
+        {   unsigned long long const ret = ZSTD_getFrameContentSize(src, srcSize);
+            if (ret >= ZSTD_CONTENTSIZE_ERROR) return ret;
 
-                src = (const BYTE *)src + frameSrcSize;
-                srcSize -= frameSrcSize;
-            }
+            /* check for overflow */
+            if (totalDstSize + ret < totalDstSize) return ZSTD_CONTENTSIZE_ERROR;
+            totalDstSize += ret;
         }
+        {   size_t const frameSrcSize = ZSTD_findFrameCompressedSize(src, srcSize);
+            if (ZSTD_isError(frameSrcSize)) {
+                return ZSTD_CONTENTSIZE_ERROR;
+            }
 
-        if (srcSize) {
-            return ZSTD_CONTENTSIZE_ERROR;
+            src = (const BYTE *)src + frameSrcSize;
+            srcSize -= frameSrcSize;
         }
+    }
 
-        return totalDstSize;
+    if (srcSize) {
+        return ZSTD_CONTENTSIZE_ERROR;
     }
+
+    return totalDstSize;
 }
 
 /** ZSTD_getDecompressedSize() :
 *   compatible with legacy mode
 *   @return : decompressed size if known, 0 otherwise
               note : 0 can mean any of the following :
-                   - decompressed size is not present within frame header
+                   - frame content is empty
+                   - decompressed size field is not present in frame header
                    - frame header unknown / not supported
                    - frame header not complete (`srcSize` too small) */
 unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize)
@@ -392,19 +454,13 @@ static size_t ZSTD_decodeFrameHeader(ZSTD_DCtx* dctx, const void* src, size_t he
     size_t const result = ZSTD_getFrameHeader(&(dctx->fParams), src, headerSize);
     if (ZSTD_isError(result)) return result;  /* invalid header */
     if (result>0) return ERROR(srcSize_wrong);   /* headerSize too small */
-    if (dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID)) return ERROR(dictionary_wrong);
+    if (dctx->fParams.dictID && (dctx->dictID != dctx->fParams.dictID))
+        return ERROR(dictionary_wrong);
     if (dctx->fParams.checksumFlag) XXH64_reset(&dctx->xxhState, 0);
     return 0;
 }
 
 
-typedef struct
-{
-    blockType_e blockType;
-    U32 lastBlock;
-    U32 origSize;
-} blockProperties_t;
-
 /*! ZSTD_getcBlockSize() :
 *   Provides the size of compressed block from block header `src` */
 size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
@@ -485,7 +541,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                     litCSize = (lhc >> 22) + (istart[4] << 10);
                     break;
                 }
-                if (litSize > ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(corruption_detected);
+                if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
                 if (litCSize + lhSize > srcSize) return ERROR(corruption_detected);
 
                 if (HUF_isError((litEncType==set_repeat) ?
@@ -493,8 +549,10 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                                         HUF_decompress1X_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr) :
                                         HUF_decompress4X_usingDTable(dctx->litBuffer, litSize, istart+lhSize, litCSize, dctx->HUFptr) ) :
                                     ( singleStream ?
-                                        HUF_decompress1X2_DCtx(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize) :
-                                        HUF_decompress4X_hufOnly (dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize)) ))
+                                        HUF_decompress1X2_DCtx_wksp(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize,
+                                                                    dctx->entropy.workspace, sizeof(dctx->entropy.workspace)) :
+                                        HUF_decompress4X_hufOnly_wksp(dctx->entropy.hufTable, dctx->litBuffer, litSize, istart+lhSize, litCSize,
+                                                                      dctx->entropy.workspace, sizeof(dctx->entropy.workspace)))))
                     return ERROR(corruption_detected);
 
                 dctx->litPtr = dctx->litBuffer;
@@ -557,7 +615,7 @@ size_t ZSTD_decodeLiteralsBlock(ZSTD_DCtx* dctx,
                     if (srcSize<4) return ERROR(corruption_detected);   /* srcSize >= MIN_CBLOCK_SIZE == 3; here we need lhSize+1 = 4 */
                     break;
                 }
-                if (litSize > ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(corruption_detected);
+                if (litSize > ZSTD_BLOCKSIZE_MAX) return ERROR(corruption_detected);
                 memset(dctx->litBuffer, istart[lhSize], litSize + WILDCOPY_OVERLENGTH);
                 dctx->litPtr = dctx->litBuffer;
                 dctx->litSize = litSize;
@@ -684,6 +742,7 @@ size_t ZSTD_decodeSeqHeaders(ZSTD_DCtx* dctx, int* nbSeqPtr,
     const BYTE* const istart = (const BYTE* const)src;
     const BYTE* const iend = istart + srcSize;
     const BYTE* ip = istart;
+    DEBUGLOG(5, "ZSTD_decodeSeqHeaders");
 
     /* check */
     if (srcSize < MIN_SEQUENCES_SIZE) return ERROR(srcSize_wrong);
@@ -801,7 +860,10 @@ size_t ZSTD_execSequenceLast7(BYTE* op,
 }
 
 
-static seq_t ZSTD_decodeSequence(seqState_t* seqState)
+typedef enum { ZSTD_lo_isRegularOffset, ZSTD_lo_isLongOffset=1 } ZSTD_longOffset_e;
+
+
+static seq_t ZSTD_decodeSequence(seqState_t* seqState, const ZSTD_longOffset_e longOffsets)
 {
     seq_t seq;
 
@@ -841,8 +903,16 @@ static seq_t ZSTD_decodeSequence(seqState_t* seqState)
         if (!ofCode)
             offset = 0;
         else {
-            offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
-            if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+            ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
+            if (longOffsets) {
+                int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN);
+                offset = OF_base[ofCode] + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
+                if (MEM_32bits() || extraBits) BIT_reloadDStream(&seqState->DStream);
+                if (extraBits) offset += BIT_readBitsFast(&seqState->DStream, extraBits);
+            } else {
+                offset = OF_base[ofCode] + BIT_readBitsFast(&seqState->DStream, ofBits);   /* <=  (ZSTD_WINDOWLOG_MAX-1) bits */
+                if (MEM_32bits()) BIT_reloadDStream(&seqState->DStream);
+            }
         }
 
         if (ofCode <= 1) {
@@ -864,12 +934,18 @@ static seq_t ZSTD_decodeSequence(seqState_t* seqState)
         seq.offset = offset;
     }
 
-    seq.matchLength = ML_base[mlCode] + ((mlCode>31) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0);  /* <=  16 bits */
+    seq.matchLength = ML_base[mlCode]
+                    + ((mlCode>31) ? BIT_readBitsFast(&seqState->DStream, mlBits) : 0);  /* <=  16 bits */
     if (MEM_32bits() && (mlBits+llBits>24)) BIT_reloadDStream(&seqState->DStream);
 
-    seq.litLength = LL_base[llCode] + ((llCode>15) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0);    /* <=  16 bits */
-    if (MEM_32bits() ||
-       (totalBits > 64 - 7 - (LLFSELog+MLFSELog+OffFSELog)) ) BIT_reloadDStream(&seqState->DStream);
+    seq.litLength = LL_base[llCode]
+                  + ((llCode>15) ? BIT_readBitsFast(&seqState->DStream, llBits) : 0);    /* <=  16 bits */
+    if (  MEM_32bits()
+      || (totalBits > 64 - 7 - (LLFSELog+MLFSELog+OffFSELog)) )
+       BIT_reloadDStream(&seqState->DStream);
+
+    DEBUGLOG(6, "seq: litL=%u, matchL=%u, offset=%u",
+                (U32)seq.litLength, (U32)seq.matchLength, (U32)seq.offset);
 
     /* ANS state update */
     FSE_updateState(&seqState->stateLL, &seqState->DStream);    /* <=  9 bits */
@@ -881,7 +957,7 @@ static seq_t ZSTD_decodeSequence(seqState_t* seqState)
 }
 
 
-FORCE_INLINE
+HINT_INLINE
 size_t ZSTD_execSequence(BYTE* op,
                          BYTE* const oend, seq_t sequence,
                          const BYTE** litPtr, const BYTE* const litLimit,
@@ -909,7 +985,8 @@ size_t ZSTD_execSequence(BYTE* op,
     /* copy Match */
     if (sequence.offset > (size_t)(oLitEnd - base)) {
         /* offset beyond prefix -> go into extDict */
-        if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
+        if (sequence.offset > (size_t)(oLitEnd - vBase))
+            return ERROR(corruption_detected);
         match = dictEnd + (match - base);
         if (match + sequence.matchLength <= dictEnd) {
             memmove(oLitEnd, match, sequence.matchLength);
@@ -964,7 +1041,8 @@ size_t ZSTD_execSequence(BYTE* op,
 static size_t ZSTD_decompressSequences(
                                ZSTD_DCtx* dctx,
                                void* dst, size_t maxDstSize,
-                         const void* seqStart, size_t seqSize)
+                         const void* seqStart, size_t seqSize,
+                         const ZSTD_longOffset_e isLongOffset)
 {
     const BYTE* ip = (const BYTE*)seqStart;
     const BYTE* const iend = ip + seqSize;
@@ -977,9 +1055,12 @@ static size_t ZSTD_decompressSequences(
     const BYTE* const vBase = (const BYTE*) (dctx->vBase);
     const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
     int nbSeq;
+    DEBUGLOG(5, "ZSTD_decompressSequences");
 
     /* Build Decoding Tables */
     {   size_t const seqHSize = ZSTD_decodeSeqHeaders(dctx, &nbSeq, ip, seqSize);
+        DEBUGLOG(5, "ZSTD_decodeSeqHeaders: size=%u, nbSeq=%i",
+                    (U32)seqHSize, nbSeq);
         if (ZSTD_isError(seqHSize)) return seqHSize;
         ip += seqHSize;
     }
@@ -996,13 +1077,15 @@ static size_t ZSTD_decompressSequences(
 
         for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && nbSeq ; ) {
             nbSeq--;
-            {   seq_t const sequence = ZSTD_decodeSequence(&seqState);
+            {   seq_t const sequence = ZSTD_decodeSequence(&seqState, isLongOffset);
                 size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequence, &litPtr, litEnd, base, vBase, dictEnd);
+                DEBUGLOG(6, "regenerated sequence size : %u", (U32)oneSeqSize);
                 if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
                 op += oneSeqSize;
         }   }
 
         /* check if reached exact end */
+        DEBUGLOG(5, "after decode loop, remaining nbSeq : %i", nbSeq);
         if (nbSeq) return ERROR(corruption_detected);
         /* save reps for next block */
         { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) dctx->entropy.rep[i] = (U32)(seqState.prevOffset[i]); }
@@ -1019,7 +1102,9 @@ static size_t ZSTD_decompressSequences(
 }
 
 
-FORCE_INLINE seq_t ZSTD_decodeSequenceLong_generic(seqState_t* seqState, int const longOffsets)
+
+HINT_INLINE
+seq_t ZSTD_decodeSequenceLong(seqState_t* seqState, ZSTD_longOffset_e const longOffsets)
 {
     seq_t seq;
 
@@ -1059,6 +1144,7 @@ FORCE_INLINE seq_t ZSTD_decodeSequenceLong_generic(seqState_t* seqState, int con
         if (!ofCode)
             offset = 0;
         else {
+            ZSTD_STATIC_ASSERT(ZSTD_lo_isLongOffset == 1);
             if (longOffsets) {
                 int const extraBits = ofBits - MIN(ofBits, STREAM_ACCUMULATOR_MIN);
                 offset = OF_base[ofCode] + (BIT_readBitsFast(&seqState->DStream, ofBits - extraBits) << extraBits);
@@ -1111,19 +1197,12 @@ FORCE_INLINE seq_t ZSTD_decodeSequenceLong_generic(seqState_t* seqState, int con
     return seq;
 }
 
-static seq_t ZSTD_decodeSequenceLong(seqState_t* seqState, unsigned const windowSize) {
-    if (ZSTD_highbit32(windowSize) > STREAM_ACCUMULATOR_MIN) {
-        return ZSTD_decodeSequenceLong_generic(seqState, 1);
-    } else {
-        return ZSTD_decodeSequenceLong_generic(seqState, 0);
-    }
-}
 
-FORCE_INLINE
+HINT_INLINE
 size_t ZSTD_execSequenceLong(BYTE* op,
-                                BYTE* const oend, seq_t sequence,
-                                const BYTE** litPtr, const BYTE* const litLimit,
-                                const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
+                             BYTE* const oend, seq_t sequence,
+                             const BYTE** litPtr, const BYTE* const litLimit,
+                             const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
 {
     BYTE* const oLitEnd = op + sequence.litLength;
     size_t const sequenceLength = sequence.litLength + sequence.matchLength;
@@ -1133,11 +1212,9 @@ size_t ZSTD_execSequenceLong(BYTE* op,
     const BYTE* match = sequence.match;
 
     /* check */
-#if 1
     if (oMatchEnd>oend) return ERROR(dstSize_tooSmall); /* last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend */
     if (iLitEnd > litLimit) return ERROR(corruption_detected);   /* over-read beyond lit buffer */
     if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, base, vBase, dictEnd);
-#endif
 
     /* copy Literals */
     ZSTD_copy8(op, *litPtr);
@@ -1147,7 +1224,6 @@ size_t ZSTD_execSequenceLong(BYTE* op,
     *litPtr = iLitEnd;   /* update for next sequence */
 
     /* copy Match */
-#if 1
     if (sequence.offset > (size_t)(oLitEnd - base)) {
         /* offset beyond prefix */
         if (sequence.offset > (size_t)(oLitEnd - vBase)) return ERROR(corruption_detected);
@@ -1167,8 +1243,8 @@ size_t ZSTD_execSequenceLong(BYTE* op,
               return sequenceLength;
             }
     }   }
-    /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */
-#endif
+    assert(op <= oend_w);
+    assert(sequence.matchLength >= MINMATCH);
 
     /* match within prefix */
     if (sequence.offset < 8) {
@@ -1204,7 +1280,8 @@ size_t ZSTD_execSequenceLong(BYTE* op,
 static size_t ZSTD_decompressSequencesLong(
                                ZSTD_DCtx* dctx,
                                void* dst, size_t maxDstSize,
-                         const void* seqStart, size_t seqSize)
+                         const void* seqStart, size_t seqSize,
+                         const ZSTD_longOffset_e isLongOffset)
 {
     const BYTE* ip = (const BYTE*)seqStart;
     const BYTE* const iend = ip + seqSize;
@@ -1216,7 +1293,6 @@ static size_t ZSTD_decompressSequencesLong(
     const BYTE* const base = (const BYTE*) (dctx->base);
     const BYTE* const vBase = (const BYTE*) (dctx->vBase);
     const BYTE* const dictEnd = (const BYTE*) (dctx->dictEnd);
-    unsigned const windowSize = dctx->fParams.windowSize;
     int nbSeq;
 
     /* Build Decoding Tables */
@@ -1246,16 +1322,16 @@ static size_t ZSTD_decompressSequencesLong(
 
         /* prepare in advance */
         for (seqNb=0; (BIT_reloadDStream(&seqState.DStream) <= BIT_DStream_completed) && seqNb<seqAdvance; seqNb++) {
-            sequences[seqNb] = ZSTD_decodeSequenceLong(&seqState, windowSize);
+            sequences[seqNb] = ZSTD_decodeSequenceLong(&seqState, isLongOffset);
         }
         if (seqNb<seqAdvance) return ERROR(corruption_detected);
 
         /* decode and decompress */
         for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && seqNb<nbSeq ; seqNb++) {
-            seq_t const sequence = ZSTD_decodeSequenceLong(&seqState, windowSize);
+            seq_t const sequence = ZSTD_decodeSequenceLong(&seqState, isLongOffset);
             size_t const oneSeqSize = ZSTD_execSequenceLong(op, oend, sequences[(seqNb-ADVANCED_SEQS) & STOSEQ_MASK], &litPtr, litEnd, base, vBase, dictEnd);
             if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
-            ZSTD_PREFETCH(sequence.match);
+            PREFETCH(sequence.match);
             sequences[seqNb&STOSEQ_MASK] = sequence;
             op += oneSeqSize;
         }
@@ -1286,25 +1362,33 @@ static size_t ZSTD_decompressSequencesLong(
 
 static size_t ZSTD_decompressBlock_internal(ZSTD_DCtx* dctx,
                             void* dst, size_t dstCapacity,
-                      const void* src, size_t srcSize)
+                      const void* src, size_t srcSize, const int frame)
 {   /* blockType == blockCompressed */
     const BYTE* ip = (const BYTE*)src;
-
-    if (srcSize >= ZSTD_BLOCKSIZE_ABSOLUTEMAX) return ERROR(srcSize_wrong);
+    /* isLongOffset must be true if there are long offsets.
+     * Offsets are long if they are larger than 2^STREAM_ACCUMULATOR_MIN.
+     * We don't expect that to be the case in 64-bit mode.
+     * If we are in block mode we don't know the window size, so we have to be
+     * conservative.
+     */
+    ZSTD_longOffset_e const isLongOffset = (ZSTD_longOffset_e)(MEM_32bits() && (!frame || dctx->fParams.windowSize > (1ULL << STREAM_ACCUMULATOR_MIN)));
+    /* windowSize could be any value at this point, since it is only validated
+     * in the streaming API.
+     */
+    DEBUGLOG(5, "ZSTD_decompressBlock_internal");
+
+    if (srcSize >= ZSTD_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);
 
     /* Decode literals section */
     {   size_t const litCSize = ZSTD_decodeLiteralsBlock(dctx, src, srcSize);
+        DEBUGLOG(5, "ZSTD_decodeLiteralsBlock : %u", (U32)litCSize);
         if (ZSTD_isError(litCSize)) return litCSize;
         ip += litCSize;
         srcSize -= litCSize;
     }
-    if (sizeof(size_t) > 4)  /* do not enable prefetching on 32-bits x86, as it's performance detrimental */
-                             /* likely because of register pressure */
-                             /* if that's the correct cause, then 32-bits ARM should be affected differently */
-                             /* it would be good to test this on ARM real hardware, to see if prefetch version improves speed */
-        if (dctx->fParams.windowSize > (1<<23))
-            return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize);
-    return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize);
+    if (frame && dctx->fParams.windowSize > (1<<23))
+        return ZSTD_decompressSequencesLong(dctx, dst, dstCapacity, ip, srcSize, isLongOffset);
+    return ZSTD_decompressSequences(dctx, dst, dstCapacity, ip, srcSize, isLongOffset);
 }
 
 
@@ -1324,7 +1408,7 @@ size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx,
 {
     size_t dSize;
     ZSTD_checkContinuity(dctx, dst);
-    dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize);
+    dSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 0);
     dctx->previousDstEnd = (char*)dst + dSize;
     return dSize;
 }
@@ -1355,28 +1439,26 @@ size_t ZSTD_generateNxBytes(void* dst, size_t dstCapacity, BYTE byte, size_t len
 size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
 {
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
-    if (ZSTD_isLegacy(src, srcSize)) return ZSTD_findFrameCompressedSizeLegacy(src, srcSize);
+    if (ZSTD_isLegacy(src, srcSize))
+        return ZSTD_findFrameCompressedSizeLegacy(src, srcSize);
 #endif
-    if (srcSize >= ZSTD_skippableHeaderSize &&
-            (MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {
+    if ( (srcSize >= ZSTD_skippableHeaderSize)
+      && (MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START ) {
         return ZSTD_skippableHeaderSize + MEM_readLE32((const BYTE*)src + 4);
     } else {
         const BYTE* ip = (const BYTE*)src;
         const BYTE* const ipstart = ip;
         size_t remainingSize = srcSize;
-        ZSTD_frameHeader fParams;
+        ZSTD_frameHeader zfh;
 
-        size_t const headerSize = ZSTD_frameHeaderSize(ip, remainingSize);
-        if (ZSTD_isError(headerSize)) return headerSize;
-
-        /* Frame Header */
-        {   size_t const ret = ZSTD_getFrameHeader(&fParams, ip, remainingSize);
+        /* Extract Frame Header */
+        {   size_t const ret = ZSTD_getFrameHeader(&zfh, src, srcSize);
             if (ZSTD_isError(ret)) return ret;
             if (ret > 0) return ERROR(srcSize_wrong);
         }
 
-        ip += headerSize;
-        remainingSize -= headerSize;
+        ip += zfh.headerSize;
+        remainingSize -= zfh.headerSize;
 
         /* Loop on each block */
         while (1) {
@@ -1384,7 +1466,8 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
             size_t const cBlockSize = ZSTD_getcBlockSize(ip, remainingSize, &blockProperties);
             if (ZSTD_isError(cBlockSize)) return cBlockSize;
 
-            if (ZSTD_blockHeaderSize + cBlockSize > remainingSize) return ERROR(srcSize_wrong);
+            if (ZSTD_blockHeaderSize + cBlockSize > remainingSize)
+                return ERROR(srcSize_wrong);
 
             ip += ZSTD_blockHeaderSize + cBlockSize;
             remainingSize -= ZSTD_blockHeaderSize + cBlockSize;
@@ -1392,7 +1475,7 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
             if (blockProperties.lastBlock) break;
         }
 
-        if (fParams.checksumFlag) {   /* Frame content checksum */
+        if (zfh.checksumFlag) {   /* Final frame content checksum */
             if (remainingSize < 4) return ERROR(srcSize_wrong);
             ip += 4;
             remainingSize -= 4;
@@ -1405,8 +1488,8 @@ size_t ZSTD_findFrameCompressedSize(const void *src, size_t srcSize)
 /*! ZSTD_decompressFrame() :
 *   @dctx must be properly initialized */
 static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
-                                 void* dst, size_t dstCapacity,
-                                 const void** srcPtr, size_t *srcSizePtr)
+                                   void* dst, size_t dstCapacity,
+                             const void** srcPtr, size_t *srcSizePtr)
 {
     const BYTE* ip = (const BYTE*)(*srcPtr);
     BYTE* const ostart = (BYTE* const)dst;
@@ -1415,13 +1498,15 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
     size_t remainingSize = *srcSizePtr;
 
     /* check */
-    if (remainingSize < ZSTD_frameHeaderSize_min+ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
+    if (remainingSize < ZSTD_frameHeaderSize_min+ZSTD_blockHeaderSize)
+        return ERROR(srcSize_wrong);
 
     /* Frame Header */
     {   size_t const frameHeaderSize = ZSTD_frameHeaderSize(ip, ZSTD_frameHeaderSize_prefix);
         if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
-        if (remainingSize < frameHeaderSize+ZSTD_blockHeaderSize) return ERROR(srcSize_wrong);
-        CHECK_F(ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize));
+        if (remainingSize < frameHeaderSize+ZSTD_blockHeaderSize)
+            return ERROR(srcSize_wrong);
+        CHECK_F( ZSTD_decodeFrameHeader(dctx, ip, frameHeaderSize) );
         ip += frameHeaderSize; remainingSize -= frameHeaderSize;
     }
 
@@ -1439,7 +1524,7 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
         switch(blockProperties.blockType)
         {
         case bt_compressed:
-            decodedSize = ZSTD_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize);
+            decodedSize = ZSTD_decompressBlock_internal(dctx, op, oend-op, ip, cBlockSize, /* frame */ 1);
             break;
         case bt_raw :
             decodedSize = ZSTD_copyRawBlock(op, oend-op, ip, cBlockSize);
@@ -1453,14 +1538,15 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
         }
 
         if (ZSTD_isError(decodedSize)) return decodedSize;
-        if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, op, decodedSize);
+        if (dctx->fParams.checksumFlag)
+            XXH64_update(&dctx->xxhState, op, decodedSize);
         op += decodedSize;
         ip += cBlockSize;
         remainingSize -= cBlockSize;
         if (blockProperties.lastBlock) break;
     }
 
-    if (dctx->fParams.checksumFlag) {   /* Frame content checksum verification */
+    if (dctx->fParams.checksumFlag) { /* Frame content checksum verification */
         U32 const checkCalc = (U32)XXH64_digest(&dctx->xxhState);
         U32 checkRead;
         if (remainingSize<4) return ERROR(checksum_wrong);
@@ -1482,17 +1568,13 @@ static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict);
 static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
                                         void* dst, size_t dstCapacity,
                                   const void* src, size_t srcSize,
-                                  const void *dict, size_t dictSize,
+                                  const void* dict, size_t dictSize,
                                   const ZSTD_DDict* ddict)
 {
     void* const dststart = dst;
+    assert(dict==NULL || ddict==NULL);  /* either dict or ddict set, not both */
 
     if (ddict) {
-        if (dict) {
-            /* programmer error, these two cases should be mutually exclusive */
-            return ERROR(GENERIC);
-        }
-
         dict = ZSTD_DDictDictContent(ddict);
         dictSize = ZSTD_DDictDictSize(ddict);
     }
@@ -1505,6 +1587,8 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
             size_t decodedSize;
             size_t const frameSize = ZSTD_findFrameCompressedSizeLegacy(src, srcSize);
             if (ZSTD_isError(frameSize)) return frameSize;
+            /* legacy support is not compatible with static dctx */
+            if (dctx->staticSize) return ERROR(memory_allocation);
 
             decodedSize = ZSTD_decompressLegacy(dst, dstCapacity, src, frameSize, dict, dictSize);
 
@@ -1526,21 +1610,18 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
                     return ERROR(srcSize_wrong);
                 skippableSize = MEM_readLE32((const BYTE *)src + 4) +
                                 ZSTD_skippableHeaderSize;
-                if (srcSize < skippableSize) {
-                    return ERROR(srcSize_wrong);
-                }
+                if (srcSize < skippableSize) return ERROR(srcSize_wrong);
 
                 src = (const BYTE *)src + skippableSize;
                 srcSize -= skippableSize;
                 continue;
-            } else {
-                return ERROR(prefix_unknown);
             }
+            return ERROR(prefix_unknown);
         }
 
         if (ddict) {
             /* we were called from ZSTD_decompress_usingDDict */
-            ZSTD_refDDict(dctx, ddict);
+            CHECK_F(ZSTD_decompressBegin_usingDDict(dctx, ddict));
         } else {
             /* this will initialize correctly with no dict if dict == NULL, so
              * use this in all cases but ddict */
@@ -1551,12 +1632,11 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
         {   const size_t res = ZSTD_decompressFrame(dctx, dst, dstCapacity,
                                                     &src, &srcSize);
             if (ZSTD_isError(res)) return res;
-            /* don't need to bounds check this, ZSTD_decompressFrame will have
-             * already */
+            /* no need to bound check, ZSTD_decompressFrame already has */
             dst = (BYTE*)dst + res;
             dstCapacity -= res;
         }
-    }
+    }  /* while (srcSize >= ZSTD_frameHeaderSize_prefix) */
 
     if (srcSize) return ERROR(srcSize_wrong); /* input not entirely consumed */
 
@@ -1580,7 +1660,7 @@ size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const
 
 size_t ZSTD_decompress(void* dst, size_t dstCapacity, const void* src, size_t srcSize)
 {
-#if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE==1)
+#if defined(ZSTD_HEAPMODE) && (ZSTD_HEAPMODE>=1)
     size_t regenSize;
     ZSTD_DCtx* const dctx = ZSTD_createDCtx();
     if (dctx==NULL) return ERROR(memory_allocation);
@@ -1604,6 +1684,7 @@ ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
     switch(dctx->stage)
     {
     default:   /* should not happen */
+        assert(0);
     case ZSTDds_getFrameHeaderSize:
     case ZSTDds_decodeFrameHeader:
         return ZSTDnit_frameHeader;
@@ -1621,21 +1702,24 @@ ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx) {
     }
 }
 
-int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; }   /* for zbuff */
+static int ZSTD_isSkipFrame(ZSTD_DCtx* dctx) { return dctx->stage == ZSTDds_skipFrame; }
 
 /** ZSTD_decompressContinue() :
-*   @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity)
-*             or an error code, which can be tested using ZSTD_isError() */
+ *  srcSize : must be the exact nb of bytes expected (see ZSTD_nextSrcSizeToDecompress())
+ *  @return : nb of bytes generated into `dst` (necessarily <= `dstCapacity)
+ *            or an error code, which can be tested using ZSTD_isError() */
 size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize)
 {
+    DEBUGLOG(5, "ZSTD_decompressContinue");
     /* Sanity check */
-    if (srcSize != dctx->expected) return ERROR(srcSize_wrong);
+    if (srcSize != dctx->expected) return ERROR(srcSize_wrong);   /* unauthorized */
     if (dstCapacity) ZSTD_checkContinuity(dctx, dst);
 
     switch (dctx->stage)
     {
     case ZSTDds_getFrameHeaderSize :
-        if (srcSize != ZSTD_frameHeaderSize_prefix) return ERROR(srcSize_wrong);      /* impossible */
+        if (srcSize != ZSTD_frameHeaderSize_prefix) return ERROR(srcSize_wrong);      /* unauthorized */
+        assert(src != NULL);
         if ((MEM_readLE32(src) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {        /* skippable frame */
             memcpy(dctx->headerBuffer, src, ZSTD_frameHeaderSize_prefix);
             dctx->expected = ZSTD_skippableHeaderSize - ZSTD_frameHeaderSize_prefix;  /* magic number + skippable frame length */
@@ -1651,8 +1735,9 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
             return 0;
         }
         dctx->expected = 0;   /* not necessary to copy more */
-
+        /* fall-through */
     case ZSTDds_decodeFrameHeader:
+        assert(src != NULL);
         memcpy(dctx->headerBuffer + ZSTD_frameHeaderSize_prefix, src, dctx->expected);
         CHECK_F(ZSTD_decodeFrameHeader(dctx, dctx->headerBuffer, dctx->headerSize));
         dctx->expected = ZSTD_blockHeaderSize;
@@ -1680,18 +1765,20 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
                     dctx->stage = ZSTDds_getFrameHeaderSize;
                 }
             } else {
-                dctx->expected = 3;  /* go directly to next header */
+                dctx->expected = ZSTD_blockHeaderSize;  /* jump to next header */
                 dctx->stage = ZSTDds_decodeBlockHeader;
             }
             return 0;
         }
     case ZSTDds_decompressLastBlock:
     case ZSTDds_decompressBlock:
+        DEBUGLOG(5, "case ZSTDds_decompressBlock");
         {   size_t rSize;
             switch(dctx->bType)
             {
             case bt_compressed:
-                rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize);
+                DEBUGLOG(5, "case bt_compressed");
+                rSize = ZSTD_decompressBlock_internal(dctx, dst, dstCapacity, src, srcSize, /* frame */ 1);
                 break;
             case bt_raw :
                 rSize = ZSTD_copyRawBlock(dst, dstCapacity, src, srcSize);
@@ -1704,9 +1791,16 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
                 return ERROR(corruption_detected);
             }
             if (ZSTD_isError(rSize)) return rSize;
+            DEBUGLOG(5, "decoded size from block : %u", (U32)rSize);
+            dctx->decodedSize += rSize;
             if (dctx->fParams.checksumFlag) XXH64_update(&dctx->xxhState, dst, rSize);
 
             if (dctx->stage == ZSTDds_decompressLastBlock) {   /* end of frame */
+                DEBUGLOG(4, "decoded size from frame : %u", (U32)dctx->decodedSize);
+                if (dctx->fParams.frameContentSize != ZSTD_CONTENTSIZE_UNKNOWN) {
+                    if (dctx->decodedSize != dctx->fParams.frameContentSize) {
+                        return ERROR(corruption_detected);
+                }   }
                 if (dctx->fParams.checksumFlag) {  /* another round for frame checksum */
                     dctx->expected = 4;
                     dctx->stage = ZSTDds_checkChecksum;
@@ -1722,15 +1816,19 @@ size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, c
             return rSize;
         }
     case ZSTDds_checkChecksum:
+        DEBUGLOG(4, "case ZSTDds_checkChecksum");
+        assert(srcSize == 4);  /* guaranteed by dctx->expected */
         {   U32 const h32 = (U32)XXH64_digest(&dctx->xxhState);
-            U32 const check32 = MEM_readLE32(src);   /* srcSize == 4, guaranteed by dctx->expected */
+            U32 const check32 = MEM_readLE32(src);
+            DEBUGLOG(4, "calculated %08X :: %08X read", h32, check32);
             if (check32 != h32) return ERROR(checksum_wrong);
             dctx->expected = 0;
             dctx->stage = ZSTDds_getFrameHeaderSize;
             return 0;
         }
     case ZSTDds_decodeSkippableHeader:
-        {   memcpy(dctx->headerBuffer + ZSTD_frameHeaderSize_prefix, src, dctx->expected);
+        {   assert(src != NULL);
+            memcpy(dctx->headerBuffer + ZSTD_frameHeaderSize_prefix, src, dctx->expected);
             dctx->expected = MEM_readLE32(dctx->headerBuffer + 4);
             dctx->stage = ZSTDds_skipFrame;
             return 0;
@@ -1758,7 +1856,7 @@ static size_t ZSTD_refDictContent(ZSTD_DCtx* dctx, const void* dict, size_t dict
 /* ZSTD_loadEntropy() :
  * dict : must point at beginning of a valid zstd dictionary
  * @return : size of entropy tables read */
-static size_t ZSTD_loadEntropy(ZSTD_entropyTables_t* entropy, const void* const dict, size_t const dictSize)
+static size_t ZSTD_loadEntropy(ZSTD_entropyDTables_t* entropy, const void* const dict, size_t const dictSize)
 {
     const BYTE* dictPtr = (const BYTE*)dict;
     const BYTE* const dictEnd = dictPtr + dictSize;
@@ -1767,7 +1865,9 @@ static size_t ZSTD_loadEntropy(ZSTD_entropyTables_t* entropy, const void* const
     dictPtr += 8;   /* skip header = magic + dictID */
 
 
-    {   size_t const hSize = HUF_readDTableX4(entropy->hufTable, dictPtr, dictEnd-dictPtr);
+    {   size_t const hSize = HUF_readDTableX4_wksp(
+            entropy->hufTable, dictPtr, dictEnd - dictPtr,
+            entropy->workspace, sizeof(entropy->workspace));
         if (HUF_isError(hSize)) return ERROR(dictionary_corrupted);
         dictPtr += hSize;
     }
@@ -1815,7 +1915,7 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
 {
     if (dictSize < 8) return ZSTD_refDictContent(dctx, dict, dictSize);
     {   U32 const magic = MEM_readLE32(dict);
-        if (magic != ZSTD_DICT_MAGIC) {
+        if (magic != ZSTD_MAGIC_DICTIONARY) {
             return ZSTD_refDictContent(dctx, dict, dictSize);   /* pure content mode */
     }   }
     dctx->dictID = MEM_readLE32((const char*)dict + 4);
@@ -1834,8 +1934,9 @@ static size_t ZSTD_decompress_insertDictionary(ZSTD_DCtx* dctx, const void* dict
 
 size_t ZSTD_decompressBegin_usingDict(ZSTD_DCtx* dctx, const void* dict, size_t dictSize)
 {
-    CHECK_F(ZSTD_decompressBegin(dctx));
-    if (dict && dictSize) CHECK_E(ZSTD_decompress_insertDictionary(dctx, dict, dictSize), dictionary_corrupted);
+    CHECK_F( ZSTD_decompressBegin(dctx) );
+    if (dict && dictSize)
+        CHECK_E(ZSTD_decompress_insertDictionary(dctx, dict, dictSize), dictionary_corrupted);
     return 0;
 }
 
@@ -1846,7 +1947,7 @@ struct ZSTD_DDict_s {
     void* dictBuffer;
     const void* dictContent;
     size_t dictSize;
-    ZSTD_entropyTables_t entropy;
+    ZSTD_entropyDTables_t entropy;
     U32 dictID;
     U32 entropyPresent;
     ZSTD_customMem cMem;
@@ -1862,10 +1963,10 @@ static size_t ZSTD_DDictDictSize(const ZSTD_DDict* ddict)
     return ddict->dictSize;
 }
 
-static void ZSTD_refDDict(ZSTD_DCtx* dstDCtx, const ZSTD_DDict* ddict)
+size_t ZSTD_decompressBegin_usingDDict(ZSTD_DCtx* dstDCtx, const ZSTD_DDict* ddict)
 {
-    ZSTD_decompressBegin(dstDCtx);  /* init */
-    if (ddict) {   /* support refDDict on NULL */
+    CHECK_F( ZSTD_decompressBegin(dstDCtx) );
+    if (ddict) {   /* support begin on NULL */
         dstDCtx->dictID = ddict->dictID;
         dstDCtx->base = ddict->dictContent;
         dstDCtx->vBase = ddict->dictContent;
@@ -1886,6 +1987,7 @@ static void ZSTD_refDDict(ZSTD_DCtx* dstDCtx, const ZSTD_DDict* ddict)
             dstDCtx->fseEntropy = 0;
         }
     }
+    return 0;
 }
 
 static size_t ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict)
@@ -1894,7 +1996,7 @@ static size_t ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict)
     ddict->entropyPresent = 0;
     if (ddict->dictSize < 8) return 0;
     {   U32 const magic = MEM_readLE32(ddict->dictContent);
-        if (magic != ZSTD_DICT_MAGIC) return 0;   /* pure content mode */
+        if (magic != ZSTD_MAGIC_DICTIONARY) return 0;   /* pure content mode */
     }
     ddict->dictID = MEM_readLE32((const char*)ddict->dictContent + 4);
 
@@ -1905,33 +2007,39 @@ static size_t ZSTD_loadEntropy_inDDict(ZSTD_DDict* ddict)
 }
 
 
-ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize, unsigned byReference, ZSTD_customMem customMem)
+static size_t ZSTD_initDDict_internal(ZSTD_DDict* ddict, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod)
+{
+    if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dict) || (!dictSize)) {
+        ddict->dictBuffer = NULL;
+        ddict->dictContent = dict;
+    } else {
+        void* const internalBuffer = ZSTD_malloc(dictSize, ddict->cMem);
+        ddict->dictBuffer = internalBuffer;
+        ddict->dictContent = internalBuffer;
+        if (!internalBuffer) return ERROR(memory_allocation);
+        memcpy(internalBuffer, dict, dictSize);
+    }
+    ddict->dictSize = dictSize;
+    ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001);  /* cover both little and big endian */
+
+    /* parse dictionary content */
+    CHECK_F( ZSTD_loadEntropy_inDDict(ddict) );
+
+    return 0;
+}
+
+ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_customMem customMem)
 {
-    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
-    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+    if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
 
     {   ZSTD_DDict* const ddict = (ZSTD_DDict*) ZSTD_malloc(sizeof(ZSTD_DDict), customMem);
         if (!ddict) return NULL;
         ddict->cMem = customMem;
 
-        if ((byReference) || (!dict) || (!dictSize)) {
-            ddict->dictBuffer = NULL;
-            ddict->dictContent = dict;
-        } else {
-            void* const internalBuffer = ZSTD_malloc(dictSize, customMem);
-            if (!internalBuffer) { ZSTD_freeDDict(ddict); return NULL; }
-            memcpy(internalBuffer, dict, dictSize);
-            ddict->dictBuffer = internalBuffer;
-            ddict->dictContent = internalBuffer;
+        if (ZSTD_isError( ZSTD_initDDict_internal(ddict, dict, dictSize, dictLoadMethod) )) {
+            ZSTD_freeDDict(ddict);
+            return NULL;
         }
-        ddict->dictSize = dictSize;
-        ddict->entropy.hufTable[0] = (HUF_DTable)((HufLog)*0x1000001);  /* cover both little and big endian */
-        /* parse dictionary content */
-        {   size_t const errorCode = ZSTD_loadEntropy_inDDict(ddict);
-            if (ZSTD_isError(errorCode)) {
-                ZSTD_freeDDict(ddict);
-                return NULL;
-        }   }
 
         return ddict;
     }
@@ -1944,10 +2052,9 @@ ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize, unsigne
 ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize)
 {
     ZSTD_customMem const allocator = { NULL, NULL, NULL };
-    return ZSTD_createDDict_advanced(dict, dictSize, 0, allocator);
+    return ZSTD_createDDict_advanced(dict, dictSize, ZSTD_dlm_byCopy, allocator);
 }
 
-
 /*! ZSTD_createDDict_byReference() :
  *  Create a digested dictionary, to start decompression without startup delay.
  *  Dictionary content is simply referenced, it will be accessed during decompression.
@@ -1955,7 +2062,28 @@ ZSTD_DDict* ZSTD_createDDict(const void* dict, size_t dictSize)
 ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize)
 {
     ZSTD_customMem const allocator = { NULL, NULL, NULL };
-    return ZSTD_createDDict_advanced(dictBuffer, dictSize, 1, allocator);
+    return ZSTD_createDDict_advanced(dictBuffer, dictSize, ZSTD_dlm_byRef, allocator);
+}
+
+
+ZSTD_DDict* ZSTD_initStaticDDict(void* workspace, size_t workspaceSize,
+                                 const void* dict, size_t dictSize,
+                                 ZSTD_dictLoadMethod_e dictLoadMethod)
+{
+    size_t const neededSpace =
+            sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+    ZSTD_DDict* const ddict = (ZSTD_DDict*)workspace;
+    assert(workspace != NULL);
+    assert(dict != NULL);
+    if ((size_t)workspace & 7) return NULL;  /* 8-aligned */
+    if (workspaceSize < neededSpace) return NULL;
+    if (dictLoadMethod == ZSTD_dlm_byCopy) {
+        memcpy(ddict+1, dict, dictSize);  /* local copy */
+        dict = ddict+1;
+    }
+    if (ZSTD_isError( ZSTD_initDDict_internal(ddict, dict, dictSize, ZSTD_dlm_byRef) ))
+        return NULL;
+    return ddict;
 }
 
 
@@ -1971,10 +2099,10 @@ size_t ZSTD_freeDDict(ZSTD_DDict* ddict)
 
 /*! ZSTD_estimateDDictSize() :
  *  Estimate amount of memory that will be needed to create a dictionary for decompression.
- *  Note : if dictionary is created "byReference", reduce this amount by dictSize */
-size_t ZSTD_estimateDDictSize(size_t dictSize)
+ *  Note : dictionary created by reference using ZSTD_dlm_byRef are smaller */
+size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod)
 {
-    return dictSize + sizeof(ZSTD_DDict);
+    return sizeof(ZSTD_DDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
 }
 
 size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict)
@@ -1990,7 +2118,7 @@ size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict)
 unsigned ZSTD_getDictID_fromDict(const void* dict, size_t dictSize)
 {
     if (dictSize < 8) return 0;
-    if (MEM_readLE32(dict) != ZSTD_DICT_MAGIC) return 0;
+    if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) return 0;
     return MEM_readLE32((const char*)dict + 4);
 }
 
@@ -2019,7 +2147,7 @@ unsigned ZSTD_getDictID_fromDDict(const ZSTD_DDict* ddict)
  *  ZSTD_getFrameHeader(), which will provide a more precise error code. */
 unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize)
 {
-    ZSTD_frameHeader zfp = { 0 , 0 , 0 , 0 };
+    ZSTD_frameHeader zfp = { 0, 0, 0, ZSTD_frame, 0, 0, 0 };
     size_t const hError = ZSTD_getFrameHeader(&zfp, src, srcSize);
     if (ZSTD_isError(hError)) return 0;
     return zfp.dictID;
@@ -2045,88 +2173,35 @@ size_t ZSTD_decompress_usingDDict(ZSTD_DCtx* dctx,
 *   Streaming decompression
 *====================================*/
 
-typedef enum { zdss_init, zdss_loadHeader,
-               zdss_read, zdss_load, zdss_flush } ZSTD_dStreamStage;
-
-/* *** Resource management *** */
-struct ZSTD_DStream_s {
-    ZSTD_DCtx* dctx;
-    ZSTD_DDict* ddictLocal;
-    const ZSTD_DDict* ddict;
-    ZSTD_frameHeader fParams;
-    ZSTD_dStreamStage stage;
-    char*  inBuff;
-    size_t inBuffSize;
-    size_t inPos;
-    size_t maxWindowSize;
-    char*  outBuff;
-    size_t outBuffSize;
-    size_t outStart;
-    size_t outEnd;
-    size_t blockSize;
-    BYTE headerBuffer[ZSTD_FRAMEHEADERSIZE_MAX];   /* tmp buffer to store frame header */
-    size_t lhSize;
-    ZSTD_customMem customMem;
-    void* legacyContext;
-    U32 previousLegacyVersion;
-    U32 legacyVersion;
-    U32 hostageByte;
-};   /* typedef'd to ZSTD_DStream within "zstd.h" */
-
-
 ZSTD_DStream* ZSTD_createDStream(void)
 {
-    return ZSTD_createDStream_advanced(defaultCustomMem);
+    return ZSTD_createDStream_advanced(ZSTD_defaultCMem);
 }
 
-ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
+ZSTD_DStream* ZSTD_initStaticDStream(void *workspace, size_t workspaceSize)
 {
-    ZSTD_DStream* zds;
-
-    if (!customMem.customAlloc && !customMem.customFree) customMem = defaultCustomMem;
-    if (!customMem.customAlloc || !customMem.customFree) return NULL;
+    return ZSTD_initStaticDCtx(workspace, workspaceSize);
+}
 
-    zds = (ZSTD_DStream*) ZSTD_malloc(sizeof(ZSTD_DStream), customMem);
-    if (zds==NULL) return NULL;
-    memset(zds, 0, sizeof(ZSTD_DStream));
-    memcpy(&zds->customMem, &customMem, sizeof(ZSTD_customMem));
-    zds->dctx = ZSTD_createDCtx_advanced(customMem);
-    if (zds->dctx == NULL) { ZSTD_freeDStream(zds); return NULL; }
-    zds->stage = zdss_init;
-    zds->maxWindowSize = ZSTD_MAXWINDOWSIZE_DEFAULT;
-    return zds;
+ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem)
+{
+    return ZSTD_createDCtx_advanced(customMem);
 }
 
 size_t ZSTD_freeDStream(ZSTD_DStream* zds)
 {
-    if (zds==NULL) return 0;   /* support free on null */
-    {   ZSTD_customMem const cMem = zds->customMem;
-        ZSTD_freeDCtx(zds->dctx);
-        zds->dctx = NULL;
-        ZSTD_freeDDict(zds->ddictLocal);
-        zds->ddictLocal = NULL;
-        ZSTD_free(zds->inBuff, cMem);
-        zds->inBuff = NULL;
-        ZSTD_free(zds->outBuff, cMem);
-        zds->outBuff = NULL;
-#if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
-        if (zds->legacyContext)
-            ZSTD_freeLegacyStreamContext(zds->legacyContext, zds->previousLegacyVersion);
-#endif
-        ZSTD_free(zds, cMem);
-        return 0;
-    }
+    return ZSTD_freeDCtx(zds);
 }
 
 
 /* *** Initialization *** */
 
-size_t ZSTD_DStreamInSize(void)  { return ZSTD_BLOCKSIZE_ABSOLUTEMAX + ZSTD_blockHeaderSize; }
-size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_ABSOLUTEMAX; }
+size_t ZSTD_DStreamInSize(void)  { return ZSTD_BLOCKSIZE_MAX + ZSTD_blockHeaderSize; }
+size_t ZSTD_DStreamOutSize(void) { return ZSTD_BLOCKSIZE_MAX; }
 
 size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
 {
-    zds->stage = zdss_loadHeader;
+    zds->streamStage = zdss_loadHeader;
     zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
     ZSTD_freeDDict(zds->ddictLocal);
     if (dict && dictSize >= 8) {
@@ -2155,7 +2230,7 @@ size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict)
 
 size_t ZSTD_resetDStream(ZSTD_DStream* zds)
 {
-    zds->stage = zdss_loadHeader;
+    zds->streamStage = zdss_loadHeader;
     zds->lhSize = zds->inPos = zds->outStart = zds->outEnd = 0;
     zds->legacyVersion = 0;
     zds->hostageByte = 0;
@@ -2167,7 +2242,7 @@ size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds,
 {
     switch(paramType)
     {
-        default : return ERROR(parameter_unknown);
+        default : return ERROR(parameter_unsupported);
         case DStream_p_maxWindowSize : zds->maxWindowSize = paramValue ? paramValue : (U32)(-1); break;
     }
     return 0;
@@ -2176,20 +2251,37 @@ size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds,
 
 size_t ZSTD_sizeof_DStream(const ZSTD_DStream* zds)
 {
-    if (zds==NULL) return 0;   /* support sizeof NULL */
-    return sizeof(*zds)
-           + ZSTD_sizeof_DCtx(zds->dctx)
-           + ZSTD_sizeof_DDict(zds->ddictLocal)
-           + zds->inBuffSize + zds->outBuffSize;
+    return ZSTD_sizeof_DCtx(zds);
+}
+
+size_t ZSTD_decodingBufferSize_min(unsigned long long windowSize, unsigned long long frameContentSize)
+{
+    size_t const blockSize = (size_t) MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
+    unsigned long long const neededRBSize = windowSize + blockSize + (WILDCOPY_OVERLENGTH * 2);
+    unsigned long long const neededSize = MIN(frameContentSize, neededRBSize);
+    size_t const minRBSize = (size_t) neededSize;
+    if ((unsigned long long)minRBSize != neededSize) return ERROR(frameParameter_windowTooLarge);
+    return minRBSize;
 }
 
-size_t ZSTD_estimateDStreamSize(ZSTD_frameHeader fHeader)
+size_t ZSTD_estimateDStreamSize(size_t windowSize)
 {
-    size_t const windowSize = fHeader.windowSize;
-    size_t const blockSize = MIN(windowSize, ZSTD_BLOCKSIZE_ABSOLUTEMAX);
+    size_t const blockSize = MIN(windowSize, ZSTD_BLOCKSIZE_MAX);
     size_t const inBuffSize = blockSize;  /* no block can be larger */
-    size_t const outBuffSize = windowSize + blockSize + (WILDCOPY_OVERLENGTH * 2);
-    return sizeof(ZSTD_DStream) + ZSTD_estimateDCtxSize() + inBuffSize + outBuffSize;
+    size_t const outBuffSize = ZSTD_decodingBufferSize_min(windowSize, ZSTD_CONTENTSIZE_UNKNOWN);
+    return ZSTD_estimateDCtxSize() + inBuffSize + outBuffSize;
+}
+
+ZSTDLIB_API size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize)
+{
+    U32 const windowSizeMax = 1U << ZSTD_WINDOWLOG_MAX;   /* note : should be user-selectable */
+    ZSTD_frameHeader zfh;
+    size_t const err = ZSTD_getFrameHeader(&zfh, src, srcSize);
+    if (ZSTD_isError(err)) return err;
+    if (err>0) return ERROR(srcSize_wrong);
+    if (zfh.windowSize > windowSizeMax)
+        return ERROR(frameParameter_windowTooLarge);
+    return ZSTD_estimateDStreamSize((size_t)zfh.windowSize);
 }
 
 
@@ -2213,13 +2305,18 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
     char* op = ostart;
     U32 someMoreWork = 1;
 
+    DEBUGLOG(5, "ZSTD_decompressStream");
+    DEBUGLOG(5, "input size : %u", (U32)(input->size - input->pos));
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
-    if (zds->legacyVersion)
+    if (zds->legacyVersion) {
+        /* legacy support is incompatible with static dctx */
+        if (zds->staticSize) return ERROR(memory_allocation);
         return ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input);
+    }
 #endif
 
     while (someMoreWork) {
-        switch(zds->stage)
+        switch(zds->streamStage)
         {
         case zdss_init :
             ZSTD_resetDStream(zds);   /* transparent reset on starting decoding a new frame */
@@ -2227,30 +2324,34 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
 
         case zdss_loadHeader :
             {   size_t const hSize = ZSTD_getFrameHeader(&zds->fParams, zds->headerBuffer, zds->lhSize);
-                if (ZSTD_isError(hSize))
+                if (ZSTD_isError(hSize)) {
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT>=1)
-                {   U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart);
+                    U32 const legacyVersion = ZSTD_isLegacy(istart, iend-istart);
                     if (legacyVersion) {
                         const void* const dict = zds->ddict ? zds->ddict->dictContent : NULL;
                         size_t const dictSize = zds->ddict ? zds->ddict->dictSize : 0;
-                        CHECK_F(ZSTD_initLegacyStream(&zds->legacyContext, zds->previousLegacyVersion, legacyVersion,
-                                                       dict, dictSize));
+                        /* legacy support is incompatible with static dctx */
+                        if (zds->staticSize) return ERROR(memory_allocation);
+                        CHECK_F(ZSTD_initLegacyStream(&zds->legacyContext,
+                                    zds->previousLegacyVersion, legacyVersion,
+                                    dict, dictSize));
                         zds->legacyVersion = zds->previousLegacyVersion = legacyVersion;
-                        return ZSTD_decompressLegacyStream(zds->legacyContext, zds->legacyVersion, output, input);
-                    } else {
-                        return hSize; /* error */
-                }   }
-#else
-                return hSize;
+                        return ZSTD_decompressLegacyStream(zds->legacyContext, legacyVersion, output, input);
+                    }
 #endif
+                    return hSize; /* error */
+                }
                 if (hSize != 0) {   /* need more input */
                     size_t const toLoad = hSize - zds->lhSize;   /* if hSize!=0, hSize > zds->lhSize */
                     if (toLoad > (size_t)(iend-ip)) {   /* not enough input to load full header */
-                        memcpy(zds->headerBuffer + zds->lhSize, ip, iend-ip);
-                        zds->lhSize += iend-ip;
+                        if (iend-ip > 0) {
+                            memcpy(zds->headerBuffer + zds->lhSize, ip, iend-ip);
+                            zds->lhSize += iend-ip;
+                        }
                         input->pos = input->size;
                         return (MAX(ZSTD_frameHeaderSize_min, hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
                     }
+                    assert(ip != NULL);
                     memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
                     break;
             }   }
@@ -2260,74 +2361,88 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                 && (U64)(size_t)(oend-op) >= zds->fParams.frameContentSize) {
                 size_t const cSize = ZSTD_findFrameCompressedSize(istart, iend-istart);
                 if (cSize <= (size_t)(iend-istart)) {
-                    size_t const decompressedSize = ZSTD_decompress_usingDDict(zds->dctx, op, oend-op, istart, cSize, zds->ddict);
+                    size_t const decompressedSize = ZSTD_decompress_usingDDict(zds, op, oend-op, istart, cSize, zds->ddict);
                     if (ZSTD_isError(decompressedSize)) return decompressedSize;
                     ip = istart + cSize;
                     op += decompressedSize;
-                    zds->dctx->expected = 0;
-                    zds->stage = zdss_init;
+                    zds->expected = 0;
+                    zds->streamStage = zdss_init;
                     someMoreWork = 0;
                     break;
             }   }
 
-            /* Consume header */
-            ZSTD_refDDict(zds->dctx, zds->ddict);
-            {   size_t const h1Size = ZSTD_nextSrcSizeToDecompress(zds->dctx);  /* == ZSTD_frameHeaderSize_prefix */
-                CHECK_F(ZSTD_decompressContinue(zds->dctx, NULL, 0, zds->headerBuffer, h1Size));
-                {   size_t const h2Size = ZSTD_nextSrcSizeToDecompress(zds->dctx);
-                    CHECK_F(ZSTD_decompressContinue(zds->dctx, NULL, 0, zds->headerBuffer+h1Size, h2Size));
-            }   }
+            /* Consume header (see ZSTDds_decodeFrameHeader) */
+            DEBUGLOG(4, "Consume header");
+            CHECK_F(ZSTD_decompressBegin_usingDDict(zds, zds->ddict));
+
+            if ((MEM_readLE32(zds->headerBuffer) & 0xFFFFFFF0U) == ZSTD_MAGIC_SKIPPABLE_START) {  /* skippable frame */
+                zds->expected = MEM_readLE32(zds->headerBuffer + 4);
+                zds->stage = ZSTDds_skipFrame;
+            } else {
+                CHECK_F(ZSTD_decodeFrameHeader(zds, zds->headerBuffer, zds->lhSize));
+                zds->expected = ZSTD_blockHeaderSize;
+                zds->stage = ZSTDds_decodeBlockHeader;
+            }
 
+            /* control buffer memory usage */
+            DEBUGLOG(4, "Control max buffer memory usage");
             zds->fParams.windowSize = MAX(zds->fParams.windowSize, 1U << ZSTD_WINDOWLOG_ABSOLUTEMIN);
             if (zds->fParams.windowSize > zds->maxWindowSize) return ERROR(frameParameter_windowTooLarge);
 
             /* Adapt buffer sizes to frame header instructions */
-            {   size_t const blockSize = MIN(zds->fParams.windowSize, ZSTD_BLOCKSIZE_ABSOLUTEMAX);
-                size_t const neededOutSize = zds->fParams.windowSize + blockSize + WILDCOPY_OVERLENGTH * 2;
-                zds->blockSize = blockSize;
-                if (zds->inBuffSize < blockSize) {
-                    ZSTD_free(zds->inBuff, zds->customMem);
-                    zds->inBuffSize = 0;
-                    zds->inBuff = (char*)ZSTD_malloc(blockSize, zds->customMem);
-                    if (zds->inBuff == NULL) return ERROR(memory_allocation);
-                    zds->inBuffSize = blockSize;
-                }
-                if (zds->outBuffSize < neededOutSize) {
-                    ZSTD_free(zds->outBuff, zds->customMem);
-                    zds->outBuffSize = 0;
-                    zds->outBuff = (char*)ZSTD_malloc(neededOutSize, zds->customMem);
-                    if (zds->outBuff == NULL) return ERROR(memory_allocation);
-                    zds->outBuffSize = neededOutSize;
+            {   size_t const neededInBuffSize = MAX(zds->fParams.blockSizeMax, 4 /* frame checksum */);
+                size_t const neededOutBuffSize = ZSTD_decodingBufferSize_min(zds->fParams.windowSize, zds->fParams.frameContentSize);
+                if ((zds->inBuffSize < neededInBuffSize) || (zds->outBuffSize < neededOutBuffSize)) {
+                    size_t const bufferSize = neededInBuffSize + neededOutBuffSize;
+                    DEBUGLOG(4, "inBuff  : from %u to %u",
+                                (U32)zds->inBuffSize, (U32)neededInBuffSize);
+                    DEBUGLOG(4, "outBuff : from %u to %u",
+                                (U32)zds->outBuffSize, (U32)neededOutBuffSize);
+                    if (zds->staticSize) {  /* static DCtx */
+                        DEBUGLOG(4, "staticSize : %u", (U32)zds->staticSize);
+                        assert(zds->staticSize >= sizeof(ZSTD_DCtx));  /* controlled at init */
+                        if (bufferSize > zds->staticSize - sizeof(ZSTD_DCtx))
+                            return ERROR(memory_allocation);
+                    } else {
+                        ZSTD_free(zds->inBuff, zds->customMem);
+                        zds->inBuffSize = 0;
+                        zds->outBuffSize = 0;
+                        zds->inBuff = (char*)ZSTD_malloc(bufferSize, zds->customMem);
+                        if (zds->inBuff == NULL) return ERROR(memory_allocation);
+                    }
+                    zds->inBuffSize = neededInBuffSize;
+                    zds->outBuff = zds->inBuff + zds->inBuffSize;
+                    zds->outBuffSize = neededOutBuffSize;
             }   }
-            zds->stage = zdss_read;
-            /* pass-through */
+            zds->streamStage = zdss_read;
+            /* fall-through */
 
         case zdss_read:
-            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+            DEBUGLOG(5, "stage zdss_read");
+            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds);
+                DEBUGLOG(5, "neededInSize = %u", (U32)neededInSize);
                 if (neededInSize==0) {  /* end of frame */
-                    zds->stage = zdss_init;
+                    zds->streamStage = zdss_init;
                     someMoreWork = 0;
                     break;
                 }
                 if ((size_t)(iend-ip) >= neededInSize) {  /* decode directly from src */
-                    const int isSkipFrame = ZSTD_isSkipFrame(zds->dctx);
-                    size_t const decodedSize = ZSTD_decompressContinue(zds->dctx,
+                    int const isSkipFrame = ZSTD_isSkipFrame(zds);
+                    size_t const decodedSize = ZSTD_decompressContinue(zds,
                         zds->outBuff + zds->outStart, (isSkipFrame ? 0 : zds->outBuffSize - zds->outStart),
                         ip, neededInSize);
                     if (ZSTD_isError(decodedSize)) return decodedSize;
                     ip += neededInSize;
                     if (!decodedSize && !isSkipFrame) break;   /* this was just a header */
                     zds->outEnd = zds->outStart + decodedSize;
-                    zds->stage = zdss_flush;
+                    zds->streamStage = zdss_flush;
                     break;
-                }
-                if (ip==iend) { someMoreWork = 0; break; }   /* no more input */
-                zds->stage = zdss_load;
-                /* pass-through */
-            }
-
+            }   }
+            if (ip==iend) { someMoreWork = 0; break; }   /* no more input */
+            zds->streamStage = zdss_load;
+            /* fall-through */
         case zdss_load:
-            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+            {   size_t const neededInSize = ZSTD_nextSrcSizeToDecompress(zds);
                 size_t const toLoad = neededInSize - zds->inPos;   /* should always be <= remaining space within inBuff */
                 size_t loadedSize;
                 if (toLoad > zds->inBuffSize - zds->inPos) return ERROR(corruption_detected);   /* should never happen */
@@ -2337,55 +2452,63 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                 if (loadedSize < toLoad) { someMoreWork = 0; break; }   /* not enough input, wait for more */
 
                 /* decode loaded input */
-                {  const int isSkipFrame = ZSTD_isSkipFrame(zds->dctx);
-                   size_t const decodedSize = ZSTD_decompressContinue(zds->dctx,
+                {  const int isSkipFrame = ZSTD_isSkipFrame(zds);
+                   size_t const decodedSize = ZSTD_decompressContinue(zds,
                         zds->outBuff + zds->outStart, zds->outBuffSize - zds->outStart,
                         zds->inBuff, neededInSize);
                     if (ZSTD_isError(decodedSize)) return decodedSize;
                     zds->inPos = 0;   /* input is consumed */
-                    if (!decodedSize && !isSkipFrame) { zds->stage = zdss_read; break; }   /* this was just a header */
+                    if (!decodedSize && !isSkipFrame) { zds->streamStage = zdss_read; break; }   /* this was just a header */
                     zds->outEnd = zds->outStart +  decodedSize;
-                    zds->stage = zdss_flush;
-                    /* pass-through */
             }   }
-
+            zds->streamStage = zdss_flush;
+            /* fall-through */
         case zdss_flush:
             {   size_t const toFlushSize = zds->outEnd - zds->outStart;
                 size_t const flushedSize = ZSTD_limitCopy(op, oend-op, zds->outBuff + zds->outStart, toFlushSize);
                 op += flushedSize;
                 zds->outStart += flushedSize;
                 if (flushedSize == toFlushSize) {  /* flush completed */
-                    zds->stage = zdss_read;
-                    if (zds->outStart + zds->blockSize > zds->outBuffSize)
+                    zds->streamStage = zdss_read;
+                    if ( (zds->outBuffSize < zds->fParams.frameContentSize)
+                      && (zds->outStart + zds->fParams.blockSizeMax > zds->outBuffSize) ) {
+                        DEBUGLOG(5, "restart filling outBuff from beginning (left:%i, needed:%u)",
+                                (int)(zds->outBuffSize - zds->outStart),
+                                (U32)zds->fParams.blockSizeMax);
                         zds->outStart = zds->outEnd = 0;
+                    }
                     break;
-                }
-                /* cannot complete flush */
-                someMoreWork = 0;
-                break;
-            }
+            }   }
+            /* cannot complete flush */
+            someMoreWork = 0;
+            break;
+
         default: return ERROR(GENERIC);   /* impossible */
     }   }
 
     /* result */
     input->pos += (size_t)(ip-istart);
     output->pos += (size_t)(op-ostart);
-    {   size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds->dctx);
+    {   size_t nextSrcSizeHint = ZSTD_nextSrcSizeToDecompress(zds);
         if (!nextSrcSizeHint) {   /* frame fully decoded */
             if (zds->outEnd == zds->outStart) {  /* output fully flushed */
                 if (zds->hostageByte) {
-                    if (input->pos >= input->size) { zds->stage = zdss_read; return 1; }  /* can't release hostage (not present) */
+                    if (input->pos >= input->size) {
+                        /* can't release hostage (not present) */
+                        zds->streamStage = zdss_read;
+                        return 1;
+                    }
                     input->pos++;  /* release hostage */
-                }
+                }   /* zds->hostageByte */
                 return 0;
-            }
+            }  /* zds->outEnd == zds->outStart */
             if (!zds->hostageByte) { /* output not fully flushed; keep last byte as hostage; will be released when all output is flushed */
                 input->pos--;   /* note : pos > 0, otherwise, impossible to finish reading last block */
                 zds->hostageByte=1;
             }
             return 1;
-        }
-        nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds->dctx) == ZSTDnit_block);   /* preload header of next block */
+        }  /* nextSrcSizeHint==0 */
+        nextSrcSizeHint += ZSTD_blockHeaderSize * (ZSTD_nextInputType(zds) == ZSTDnit_block);   /* preload header of next block */
         if (zds->inPos > nextSrcSizeHint) return ERROR(GENERIC);   /* should never happen */
         nextSrcSizeHint -= zds->inPos;   /* already loaded*/
         return nextSrcSizeHint;
diff --git a/contrib/zstd/zstd_double_fast.c b/contrib/zstd/zstd_double_fast.c
new file mode 100644
index 000000000..876a36042
--- /dev/null
+++ b/contrib/zstd/zstd_double_fast.c
@@ -0,0 +1,308 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_double_fast.h"
+
+
+void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls)
+{
+    U32* const hashLarge = cctx->hashTable;
+    U32  const hBitsL = cctx->appliedParams.cParams.hashLog;
+    U32* const hashSmall = cctx->chainTable;
+    U32  const hBitsS = cctx->appliedParams.cParams.chainLog;
+    const BYTE* const base = cctx->base;
+    const BYTE* ip = base + cctx->nextToUpdate;
+    const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
+    const size_t fastHashFillStep = 3;
+
+    while(ip <= iend) {
+        hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip - base);
+        hashLarge[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip - base);
+        ip += fastHashFillStep;
+    }
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* const hashLong = cctx->hashTable;
+    const U32 hBitsL = cctx->appliedParams.cParams.hashLog;
+    U32* const hashSmall = cctx->chainTable;
+    const U32 hBitsS = cctx->appliedParams.cParams.chainLog;
+    seqStore_t* seqStorePtr = &(cctx->seqStore);
+    const BYTE* const base = cctx->base;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32 lowestIndex = cctx->dictLimit;
+    const BYTE* const lowest = base + lowestIndex;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - HASH_READ_SIZE;
+    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+    U32 offsetSaved = 0;
+
+    /* init */
+    ip += (ip==lowest);
+    {   U32 const maxRep = (U32)(ip-lowest);
+        if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
+        if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
+    }
+
+    /* Main Search Loop */
+    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
+        size_t mLength;
+        size_t const h2 = ZSTD_hashPtr(ip, hBitsL, 8);
+        size_t const h = ZSTD_hashPtr(ip, hBitsS, mls);
+        U32 const current = (U32)(ip-base);
+        U32 const matchIndexL = hashLong[h2];
+        U32 const matchIndexS = hashSmall[h];
+        const BYTE* matchLong = base + matchIndexL;
+        const BYTE* match = base + matchIndexS;
+        hashLong[h2] = hashSmall[h] = current;   /* update hash tables */
+
+        assert(offset_1 <= current);   /* supposed guaranteed by construction */
+        if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
+            /* favor repcode */
+            mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            U32 offset;
+            if ( (matchIndexL > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip)) ) {
+                mLength = ZSTD_count(ip+8, matchLong+8, iend) + 8;
+                offset = (U32)(ip-matchLong);
+                while (((ip>anchor) & (matchLong>lowest)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; } /* catch up */
+            } else if ( (matchIndexS > lowestIndex) && (MEM_read32(match) == MEM_read32(ip)) ) {
+                size_t const hl3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
+                U32 const matchIndexL3 = hashLong[hl3];
+                const BYTE* matchL3 = base + matchIndexL3;
+                hashLong[hl3] = current + 1;
+                if ( (matchIndexL3 > lowestIndex) && (MEM_read64(matchL3) == MEM_read64(ip+1)) ) {
+                    mLength = ZSTD_count(ip+9, matchL3+8, iend) + 8;
+                    ip++;
+                    offset = (U32)(ip-matchL3);
+                    while (((ip>anchor) & (matchL3>lowest)) && (ip[-1] == matchL3[-1])) { ip--; matchL3--; mLength++; } /* catch up */
+                } else {
+                    mLength = ZSTD_count(ip+4, match+4, iend) + 4;
+                    offset = (U32)(ip-match);
+                    while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+                }
+            } else {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+            }
+
+            offset_2 = offset_1;
+            offset_1 = offset;
+
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }
+
+        /* match found */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashLong[ZSTD_hashPtr(base+current+2, hBitsL, 8)] =
+                hashSmall[ZSTD_hashPtr(base+current+2, hBitsS, mls)] = current+2;  /* here because current+2 could be > iend-8 */
+            hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] =
+                hashSmall[ZSTD_hashPtr(ip-2, hBitsS, mls)] = (U32)(ip-2-base);
+
+            /* check immediate repcode */
+            while ( (ip <= ilimit)
+                 && ( (offset_2>0)
+                 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+                /* store sequence */
+                size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
+                hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
+                hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+                ip += rLength;
+                anchor = ip;
+                continue;   /* faster when present ... (?) */
+    }   }   }
+
+    /* save reps for next block */
+    seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
+    seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    switch(mls)
+    {
+    default: /* includes case 3 */
+    case 4 :
+        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 4);
+    case 5 :
+        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 5);
+    case 6 :
+        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 6);
+    case 7 :
+        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 7);
+    }
+}
+
+
+static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* const hashLong = ctx->hashTable;
+    U32  const hBitsL = ctx->appliedParams.cParams.hashLog;
+    U32* const hashSmall = ctx->chainTable;
+    U32  const hBitsS = ctx->appliedParams.cParams.chainLog;
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const base = ctx->base;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = ctx->lowLimit;
+    const BYTE* const dictStart = dictBase + lowestIndex;
+    const U32   dictLimit = ctx->dictLimit;
+    const BYTE* const lowPrefixPtr = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+
+    /* Search Loop */
+    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
+        const size_t hSmall = ZSTD_hashPtr(ip, hBitsS, mls);
+        const U32 matchIndex = hashSmall[hSmall];
+        const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
+        const BYTE* match = matchBase + matchIndex;
+
+        const size_t hLong = ZSTD_hashPtr(ip, hBitsL, 8);
+        const U32 matchLongIndex = hashLong[hLong];
+        const BYTE* matchLongBase = matchLongIndex < dictLimit ? dictBase : base;
+        const BYTE* matchLong = matchLongBase + matchLongIndex;
+
+        const U32 current = (U32)(ip-base);
+        const U32 repIndex = current + 1 - offset_1;   /* offset_1 expected <= current +1 */
+        const BYTE* repBase = repIndex < dictLimit ? dictBase : base;
+        const BYTE* repMatch = repBase + repIndex;
+        size_t mLength;
+        hashSmall[hSmall] = hashLong[hLong] = current;   /* update hash table */
+
+        if ( (((U32)((dictLimit-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > lowestIndex))
+           && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+            const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
+            mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            if ((matchLongIndex > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
+                const BYTE* matchEnd = matchLongIndex < dictLimit ? dictEnd : iend;
+                const BYTE* lowMatchPtr = matchLongIndex < dictLimit ? dictStart : lowPrefixPtr;
+                U32 offset;
+                mLength = ZSTD_count_2segments(ip+8, matchLong+8, iend, matchEnd, lowPrefixPtr) + 8;
+                offset = current - matchLongIndex;
+                while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; }   /* catch up */
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+            } else if ((matchIndex > lowestIndex) && (MEM_read32(match) == MEM_read32(ip))) {
+                size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
+                U32 const matchIndex3 = hashLong[h3];
+                const BYTE* const match3Base = matchIndex3 < dictLimit ? dictBase : base;
+                const BYTE* match3 = match3Base + matchIndex3;
+                U32 offset;
+                hashLong[h3] = current + 1;
+                if ( (matchIndex3 > lowestIndex) && (MEM_read64(match3) == MEM_read64(ip+1)) ) {
+                    const BYTE* matchEnd = matchIndex3 < dictLimit ? dictEnd : iend;
+                    const BYTE* lowMatchPtr = matchIndex3 < dictLimit ? dictStart : lowPrefixPtr;
+                    mLength = ZSTD_count_2segments(ip+9, match3+8, iend, matchEnd, lowPrefixPtr) + 8;
+                    ip++;
+                    offset = current+1 - matchIndex3;
+                    while (((ip>anchor) & (match3>lowMatchPtr)) && (ip[-1] == match3[-1])) { ip--; match3--; mLength++; } /* catch up */
+                } else {
+                    const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
+                    const BYTE* lowMatchPtr = matchIndex < dictLimit ? dictStart : lowPrefixPtr;
+                    mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, lowPrefixPtr) + 4;
+                    offset = current - matchIndex;
+                    while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
+                }
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+
+            } else {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+        }   }
+
+        /* found a match : store it */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashSmall[ZSTD_hashPtr(base+current+2, hBitsS, mls)] = current+2;
+            hashLong[ZSTD_hashPtr(base+current+2, hBitsL, 8)] = current+2;
+            hashSmall[ZSTD_hashPtr(ip-2, hBitsS, mls)] = (U32)(ip-2-base);
+            hashLong[ZSTD_hashPtr(ip-2, hBitsL, 8)] = (U32)(ip-2-base);
+            /* check immediate repcode */
+            while (ip <= ilimit) {
+                U32 const current2 = (U32)(ip-base);
+                U32 const repIndex2 = current2 - offset_2;
+                const BYTE* repMatch2 = repIndex2 < dictLimit ? dictBase + repIndex2 : base + repIndex2;
+                if ( (((U32)((dictLimit-1) - repIndex2) >= 3) & (repIndex2 > lowestIndex))  /* intentional overflow */
+                   && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+                    const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
+                    size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
+                    U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
+                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
+                    hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
+                    hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
+                    ip += repLength2;
+                    anchor = ip;
+                    continue;
+                }
+                break;
+    }   }   }
+
+    /* save reps for next block */
+    seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx,
+                         const void* src, size_t srcSize)
+{
+    U32 const mls = ctx->appliedParams.cParams.searchLength;
+    switch(mls)
+    {
+    default: /* includes case 3 */
+    case 4 :
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 4);
+    case 5 :
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 5);
+    case 6 :
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 6);
+    case 7 :
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 7);
+    }
+}
diff --git a/contrib/zstd/zstd_double_fast.h b/contrib/zstd/zstd_double_fast.h
new file mode 100644
index 000000000..3dba6c710
--- /dev/null
+++ b/contrib/zstd/zstd_double_fast.h
@@ -0,0 +1,28 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_DOUBLE_FAST_H
+#define ZSTD_DOUBLE_FAST_H
+
+#include "zstd_compress.h"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls);
+size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_DOUBLE_FAST_H */
diff --git a/contrib/zstd/zstd_errors.h b/contrib/zstd/zstd_errors.h
index de0fc8984..bde4304c9 100644
--- a/contrib/zstd/zstd_errors.h
+++ b/contrib/zstd/zstd_errors.h
@@ -1,10 +1,11 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 #ifndef ZSTD_ERRORS_H_398273423
@@ -19,10 +20,12 @@ extern "C" {
 
 
 /* =====   ZSTDERRORLIB_API : control library symbols visibility   ===== */
-#if defined(__GNUC__) && (__GNUC__ >= 4)
-#  define ZSTDERRORLIB_VISIBILITY __attribute__ ((visibility ("default")))
-#else
-#  define ZSTDERRORLIB_VISIBILITY
+#ifndef ZSTDERRORLIB_VISIBILITY
+#  if defined(__GNUC__) && (__GNUC__ >= 4)
+#    define ZSTDERRORLIB_VISIBILITY __attribute__ ((visibility ("default")))
+#  else
+#    define ZSTDERRORLIB_VISIBILITY
+#  endif
 #endif
 #if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
 #  define ZSTDERRORLIB_API __declspec(dllexport) ZSTDERRORLIB_VISIBILITY
@@ -33,41 +36,43 @@ extern "C" {
 #endif
 
 /*-****************************************
-*  error codes list
-******************************************/
+ *  error codes list
+ *  note : this API is still considered unstable
+ *         and shall not be used with a dynamic library.
+ *         only static linking is allowed
+ ******************************************/
 typedef enum {
-  ZSTD_error_no_error,
-  ZSTD_error_GENERIC,
-  ZSTD_error_prefix_unknown,
-  ZSTD_error_version_unsupported,
-  ZSTD_error_parameter_unknown,
-  ZSTD_error_frameParameter_unsupported,
-  ZSTD_error_frameParameter_unsupportedBy32bits,
-  ZSTD_error_frameParameter_windowTooLarge,
-  ZSTD_error_compressionParameter_unsupported,
-  ZSTD_error_init_missing,
-  ZSTD_error_memory_allocation,
-  ZSTD_error_stage_wrong,
-  ZSTD_error_dstSize_tooSmall,
-  ZSTD_error_srcSize_wrong,
-  ZSTD_error_corruption_detected,
-  ZSTD_error_checksum_wrong,
-  ZSTD_error_tableLog_tooLarge,
-  ZSTD_error_maxSymbolValue_tooLarge,
-  ZSTD_error_maxSymbolValue_tooSmall,
-  ZSTD_error_dictionary_corrupted,
-  ZSTD_error_dictionary_wrong,
-  ZSTD_error_dictionaryCreation_failed,
-  ZSTD_error_frameIndex_tooLarge,
-  ZSTD_error_seekableIO,
-  ZSTD_error_maxCode
+  ZSTD_error_no_error = 0,
+  ZSTD_error_GENERIC  = 1,
+  ZSTD_error_prefix_unknown                = 10,
+  ZSTD_error_version_unsupported           = 12,
+  ZSTD_error_frameParameter_unsupported    = 14,
+  ZSTD_error_frameParameter_windowTooLarge = 16,
+  ZSTD_error_corruption_detected = 20,
+  ZSTD_error_checksum_wrong      = 22,
+  ZSTD_error_dictionary_corrupted      = 30,
+  ZSTD_error_dictionary_wrong          = 32,
+  ZSTD_error_dictionaryCreation_failed = 34,
+  ZSTD_error_parameter_unsupported   = 40,
+  ZSTD_error_parameter_outOfBound    = 42,
+  ZSTD_error_tableLog_tooLarge       = 44,
+  ZSTD_error_maxSymbolValue_tooLarge = 46,
+  ZSTD_error_maxSymbolValue_tooSmall = 48,
+  ZSTD_error_stage_wrong       = 60,
+  ZSTD_error_init_missing      = 62,
+  ZSTD_error_memory_allocation = 64,
+  ZSTD_error_dstSize_tooSmall = 70,
+  ZSTD_error_srcSize_wrong    = 72,
+  ZSTD_error_frameIndex_tooLarge = 100,
+  ZSTD_error_seekableIO          = 102,
+  ZSTD_error_maxCode = 120  /* never EVER use this value directly, it may change in future versions! Use ZSTD_isError() instead */
 } ZSTD_ErrorCode;
 
 /*! ZSTD_getErrorCode() :
     convert a `size_t` function result into a `ZSTD_ErrorCode` enum type,
-    which can be used to compare directly with enum list published into "error_public.h" */
+    which can be used to compare with enum list published above */
 ZSTDERRORLIB_API ZSTD_ErrorCode ZSTD_getErrorCode(size_t functionResult);
-ZSTDERRORLIB_API const char* ZSTD_getErrorString(ZSTD_ErrorCode code);
+ZSTDERRORLIB_API const char* ZSTD_getErrorString(ZSTD_ErrorCode code);   /**< Same as ZSTD_getErrorName, but using a `ZSTD_ErrorCode` enum argument */
 
 
 #if defined (__cplusplus)
diff --git a/contrib/zstd/zstd_fast.c b/contrib/zstd/zstd_fast.c
new file mode 100644
index 000000000..2e057017b
--- /dev/null
+++ b/contrib/zstd/zstd_fast.c
@@ -0,0 +1,242 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_fast.h"
+
+
+void ZSTD_fillHashTable (ZSTD_CCtx* zc, const void* end, const U32 mls)
+{
+    U32* const hashTable = zc->hashTable;
+    U32  const hBits = zc->appliedParams.cParams.hashLog;
+    const BYTE* const base = zc->base;
+    const BYTE* ip = base + zc->nextToUpdate;
+    const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
+    const size_t fastHashFillStep = 3;
+
+    while(ip <= iend) {
+        hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip - base);
+        ip += fastHashFillStep;
+    }
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* const hashTable = cctx->hashTable;
+    U32  const hBits = cctx->appliedParams.cParams.hashLog;
+    seqStore_t* seqStorePtr = &(cctx->seqStore);
+    const BYTE* const base = cctx->base;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = cctx->dictLimit;
+    const BYTE* const lowest = base + lowestIndex;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - HASH_READ_SIZE;
+    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+    U32 offsetSaved = 0;
+
+    /* init */
+    ip += (ip==lowest);
+    {   U32 const maxRep = (U32)(ip-lowest);
+        if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
+        if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
+    }
+
+    /* Main Search Loop */
+    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
+        size_t mLength;
+        size_t const h = ZSTD_hashPtr(ip, hBits, mls);
+        U32 const current = (U32)(ip-base);
+        U32 const matchIndex = hashTable[h];
+        const BYTE* match = base + matchIndex;
+        hashTable[h] = current;   /* update hash table */
+
+        if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
+            mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            U32 offset;
+            if ( (matchIndex <= lowestIndex) || (MEM_read32(match) != MEM_read32(ip)) ) {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+            }
+            mLength = ZSTD_count(ip+4, match+4, iend) + 4;
+            offset = (U32)(ip-match);
+            while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+            offset_2 = offset_1;
+            offset_1 = offset;
+
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }
+
+        /* match found */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;  /* here because current+2 could be > iend-8 */
+            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
+            /* check immediate repcode */
+            while ( (ip <= ilimit)
+                 && ( (offset_2>0)
+                 & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+                /* store sequence */
+                size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; }  /* swap offset_2 <=> offset_1 */
+                hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip-base);
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+                ip += rLength;
+                anchor = ip;
+                continue;   /* faster when present ... (?) */
+    }   }   }
+
+    /* save reps for next block */
+    seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
+    seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
+                       const void* src, size_t srcSize)
+{
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    switch(mls)
+    {
+    default: /* includes case 3 */
+    case 4 :
+        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 4);
+    case 5 :
+        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 5);
+    case 6 :
+        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 6);
+    case 7 :
+        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 7);
+    }
+}
+
+
+static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
+                                 const void* src, size_t srcSize,
+                                 const U32 mls)
+{
+    U32* hashTable = ctx->hashTable;
+    const U32 hBits = ctx->appliedParams.cParams.hashLog;
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const base = ctx->base;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = ctx->lowLimit;
+    const BYTE* const dictStart = dictBase + lowestIndex;
+    const U32   dictLimit = ctx->dictLimit;
+    const BYTE* const lowPrefixPtr = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+
+    /* Search Loop */
+    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
+        const size_t h = ZSTD_hashPtr(ip, hBits, mls);
+        const U32 matchIndex = hashTable[h];
+        const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
+        const BYTE* match = matchBase + matchIndex;
+        const U32 current = (U32)(ip-base);
+        const U32 repIndex = current + 1 - offset_1;   /* offset_1 expected <= current +1 */
+        const BYTE* repBase = repIndex < dictLimit ? dictBase : base;
+        const BYTE* repMatch = repBase + repIndex;
+        size_t mLength;
+        hashTable[h] = current;   /* update hash table */
+
+        if ( (((U32)((dictLimit-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > lowestIndex))
+           && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+            const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
+            mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
+            ip++;
+            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+        } else {
+            if ( (matchIndex < lowestIndex) ||
+                 (MEM_read32(match) != MEM_read32(ip)) ) {
+                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                continue;
+            }
+            {   const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
+                const BYTE* lowMatchPtr = matchIndex < dictLimit ? dictStart : lowPrefixPtr;
+                U32 offset;
+                mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, lowPrefixPtr) + 4;
+                while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
+                offset = current - matchIndex;
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }   }
+
+        /* found a match : store it */
+        ip += mLength;
+        anchor = ip;
+
+        if (ip <= ilimit) {
+            /* Fill Table */
+            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;
+            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
+            /* check immediate repcode */
+            while (ip <= ilimit) {
+                U32 const current2 = (U32)(ip-base);
+                U32 const repIndex2 = current2 - offset_2;
+                const BYTE* repMatch2 = repIndex2 < dictLimit ? dictBase + repIndex2 : base + repIndex2;
+                if ( (((U32)((dictLimit-1) - repIndex2) >= 3) & (repIndex2 > lowestIndex))  /* intentional overflow */
+                   && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+                    const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
+                    size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
+                    U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
+                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
+                    hashTable[ZSTD_hashPtr(ip, hBits, mls)] = current2;
+                    ip += repLength2;
+                    anchor = ip;
+                    continue;
+                }
+                break;
+    }   }   }
+
+    /* save reps for next block */
+    seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
+                         const void* src, size_t srcSize)
+{
+    U32 const mls = ctx->appliedParams.cParams.searchLength;
+    switch(mls)
+    {
+    default: /* includes case 3 */
+    case 4 :
+        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 4);
+    case 5 :
+        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 5);
+    case 6 :
+        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 6);
+    case 7 :
+        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 7);
+    }
+}
diff --git a/contrib/zstd/zstd_fast.h b/contrib/zstd/zstd_fast.h
new file mode 100644
index 000000000..4205141a9
--- /dev/null
+++ b/contrib/zstd/zstd_fast.h
@@ -0,0 +1,30 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_FAST_H
+#define ZSTD_FAST_H
+
+#include "zstd_compress.h"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+void ZSTD_fillHashTable(ZSTD_CCtx* zc, const void* end, const U32 mls);
+size_t ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
+                         const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
+                         const void* src, size_t srcSize);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_FAST_H */
diff --git a/contrib/zstd/zstd_internal.h b/contrib/zstd/zstd_internal.h
index 2533333ba..e7d02b5a0 100644
--- a/contrib/zstd/zstd_internal.h
+++ b/contrib/zstd/zstd_internal.h
@@ -1,58 +1,73 @@
-/**
+/*
  * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
 #ifndef ZSTD_CCOMMON_H_MODULE
 #define ZSTD_CCOMMON_H_MODULE
 
-/*-*******************************************************
-*  Compiler specifics
-*********************************************************/
-#ifdef _MSC_VER    /* Visual Studio */
-#  define FORCE_INLINE static __forceinline
-#  include <intrin.h>                    /* For Visual 2005 */
-#  pragma warning(disable : 4100)        /* disable: C4100: unreferenced formal parameter */
-#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
-#  pragma warning(disable : 4324)        /* disable: C4324: padded structure */
-#else
-#  if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
-#    ifdef __GNUC__
-#      define FORCE_INLINE static inline __attribute__((always_inline))
-#    else
-#      define FORCE_INLINE static inline
-#    endif
-#  else
-#    define FORCE_INLINE static
-#  endif /* __STDC_VERSION__ */
-#endif
-
-#ifdef _MSC_VER
-#  define FORCE_NOINLINE static __declspec(noinline)
-#else
-#  ifdef __GNUC__
-#    define FORCE_NOINLINE static __attribute__((__noinline__))
-#  else
-#    define FORCE_NOINLINE static
-#  endif
-#endif
-
 
 /*-*************************************
 *  Dependencies
 ***************************************/
+#include "compiler.h"
 #include "mem.h"
 #include "error_private.h"
+#ifndef ZSTD_STATIC_LINKING_ONLY
 #define ZSTD_STATIC_LINKING_ONLY
+#endif
 #include "zstd.h"
+#define FSE_STATIC_LINKING_ONLY
+#include "fse.h"
+#define HUF_STATIC_LINKING_ONLY
+#include "huf.h"
 #ifndef XXH_STATIC_LINKING_ONLY
-#  define XXH_STATIC_LINKING_ONLY   /* XXH64_state_t */
+#  define XXH_STATIC_LINKING_ONLY  /* XXH64_state_t */
+#endif
+#include "xxhash.h"                /* XXH_reset, update, digest */
+
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+
+/*-*************************************
+*  Debug
+***************************************/
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
+#  include <assert.h>
+#else
+#  ifndef assert
+#    define assert(condition) ((void)0)
+#  endif
+#endif
+
+#define ZSTD_STATIC_ASSERT(c) { enum { ZSTD_static_assert = 1/(int)(!!(c)) }; }
+
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
+#  include <stdio.h>
+/* recommended values for ZSTD_DEBUG display levels :
+ * 1 : no display, enables assert() only
+ * 2 : reserved for currently active debugging path
+ * 3 : events once per object lifetime (CCtx, CDict)
+ * 4 : events once per frame
+ * 5 : events once per block
+ * 6 : events once per sequence (*very* verbose) */
+#  define DEBUGLOG(l, ...) {                         \
+                if (l<=ZSTD_DEBUG) {                 \
+                    fprintf(stderr, __FILE__ ": ");  \
+                    fprintf(stderr, __VA_ARGS__);    \
+                    fprintf(stderr, " \n");          \
+            }   }
+#else
+#  define DEBUGLOG(l, ...)      {}    /* disabled */
 #endif
-#include "xxhash.h"               /* XXH_reset, update, digest */
 
 
 /*-*************************************
@@ -70,7 +85,6 @@
 *  Common constants
 ***************************************/
 #define ZSTD_OPT_NUM    (1<<12)
-#define ZSTD_DICT_MAGIC  0xEC30A437   /* v0.7+ */
 
 #define ZSTD_REP_NUM      3                 /* number of repcodes */
 #define ZSTD_REP_CHECK    (ZSTD_REP_NUM)    /* number of repcodes to check by the optimal parser */
@@ -178,20 +192,6 @@ MEM_STATIC void ZSTD_wildcopy_e(void* dst, const void* src, void* dstEnd)   /* s
 *********************************************/
 typedef struct ZSTD_stats_s ZSTD_stats_t;
 
-typedef struct {
-    U32 off;
-    U32 len;
-} ZSTD_match_t;
-
-typedef struct {
-    U32 price;
-    U32 off;
-    U32 mlen;
-    U32 litlen;
-    U32 rep[ZSTD_REP_NUM];
-} ZSTD_optimal_t;
-
-
 typedef struct seqDef_s {
     U32 offset;
     U16 litLength;
@@ -209,13 +209,31 @@ typedef struct {
     BYTE* ofCode;
     U32   longLengthID;   /* 0 == no longLength; 1 == Lit.longLength; 2 == Match.longLength; */
     U32   longLengthPos;
-    /* opt */
-    ZSTD_optimal_t* priceTable;
-    ZSTD_match_t* matchTable;
-    U32* matchLengthFreq;
-    U32* litLengthFreq;
+    U32   rep[ZSTD_REP_NUM];
+    U32   repToConfirm[ZSTD_REP_NUM];
+} seqStore_t;
+
+typedef struct {
+    U32 off;
+    U32 len;
+} ZSTD_match_t;
+
+typedef struct {
+    U32 price;
+    U32 off;
+    U32 mlen;
+    U32 litlen;
+    U32 rep[ZSTD_REP_NUM];
+} ZSTD_optimal_t;
+
+typedef struct {
     U32* litFreq;
+    U32* litLengthFreq;
+    U32* matchLengthFreq;
     U32* offCodeFreq;
+    ZSTD_match_t* matchTable;
+    ZSTD_optimal_t* priceTable;
+
     U32  matchLengthSum;
     U32  matchSum;
     U32  litLengthSum;
@@ -231,19 +249,67 @@ typedef struct {
     U32  cachedPrice;
     U32  cachedLitLength;
     const BYTE* cachedLiterals;
-} seqStore_t;
+} optState_t;
+
+typedef struct {
+    U32 offset;
+    U32 checksum;
+} ldmEntry_t;
+
+typedef struct {
+    ldmEntry_t* hashTable;
+    BYTE* bucketOffsets;    /* Next position in bucket to insert entry */
+    U64 hashPower;          /* Used to compute the rolling hash.
+                             * Depends on ldmParams.minMatchLength */
+} ldmState_t;
+
+typedef struct {
+    U32 enableLdm;          /* 1 if enable long distance matching */
+    U32 hashLog;            /* Log size of hashTable */
+    U32 bucketSizeLog;      /* Log bucket size for collision resolution, at most 8 */
+    U32 minMatchLength;     /* Minimum match length */
+    U32 hashEveryLog;       /* Log number of entries to skip */
+} ldmParams_t;
+
+typedef struct {
+    U32 hufCTable[HUF_CTABLE_SIZE_U32(255)];
+    FSE_CTable offcodeCTable[FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)];
+    FSE_CTable matchlengthCTable[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)];
+    FSE_CTable litlengthCTable[FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)];
+    U32 workspace[HUF_WORKSPACE_SIZE_U32];
+    HUF_repeat hufCTable_repeatMode;
+    FSE_repeat offcode_repeatMode;
+    FSE_repeat matchlength_repeatMode;
+    FSE_repeat litlength_repeatMode;
+} ZSTD_entropyCTables_t;
+
+struct ZSTD_CCtx_params_s {
+    ZSTD_compressionParameters cParams;
+    ZSTD_frameParameters fParams;
+
+    int compressionLevel;
+    U32 forceWindow;           /* force back-references to respect limit of
+                                * 1<<wLog, even for dictionary */
+
+    /* Multithreading: used to pass parameters to mtctx */
+    U32 nbThreads;
+    unsigned jobSize;
+    unsigned overlapSizeLog;
+
+    /* Long distance matching parameters */
+    ldmParams_t ldmParams;
+
+    /* For use with createCCtxParams() and freeCCtxParams() only */
+    ZSTD_customMem customMem;
+
+};  /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
 
 const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx);
 void ZSTD_seqToCodes(const seqStore_t* seqStorePtr);
-int ZSTD_isSkipFrame(ZSTD_DCtx* dctx);
 
 /* custom memory allocation functions */
-void* ZSTD_defaultAllocFunction(void* opaque, size_t size);
-void ZSTD_defaultFreeFunction(void* opaque, void* address);
-#ifndef ZSTD_DLL_IMPORT
-static const ZSTD_customMem defaultCustomMem = { ZSTD_defaultAllocFunction, ZSTD_defaultFreeFunction, NULL };
-#endif
 void* ZSTD_malloc(size_t size, ZSTD_customMem customMem);
+void* ZSTD_calloc(size_t size, ZSTD_customMem customMem);
 void ZSTD_free(void* ptr, ZSTD_customMem customMem);
 
 
@@ -251,24 +317,27 @@ void ZSTD_free(void* ptr, ZSTD_customMem customMem);
 
 MEM_STATIC U32 ZSTD_highbit32(U32 val)
 {
+    assert(val != 0);
+    {
 #   if defined(_MSC_VER)   /* Visual */
-    unsigned long r=0;
-    _BitScanReverse(&r, val);
-    return (unsigned)r;
+        unsigned long r=0;
+        _BitScanReverse(&r, val);
+        return (unsigned)r;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* GCC Intrinsic */
-    return 31 - __builtin_clz(val);
+        return 31 - __builtin_clz(val);
 #   else   /* Software version */
-    static const int DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
-    U32 v = val;
-    int r;
-    v |= v >> 1;
-    v |= v >> 2;
-    v |= v >> 4;
-    v |= v >> 8;
-    v |= v >> 16;
-    r = DeBruijnClz[(U32)(v * 0x07C4ACDDU) >> 27];
-    return r;
+        static const int DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+        U32 v = val;
+        int r;
+        v |= v >> 1;
+        v |= v >> 2;
+        v |= v >> 4;
+        v |= v >> 8;
+        v |= v >> 16;
+        r = DeBruijnClz[(U32)(v * 0x07C4ACDDU) >> 27];
+        return r;
 #   endif
+    }
 }
 
 
@@ -281,4 +350,56 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)
 void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx);
 
 
+/*! ZSTD_initCStream_internal() :
+ *  Private use only. Init streaming operation.
+ *  expects params to be valid.
+ *  must receive dict, or cdict, or none, but not both.
+ *  @return : 0, or an error code */
+size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
+                     const void* dict, size_t dictSize,
+                     const ZSTD_CDict* cdict,
+                     ZSTD_CCtx_params  params, unsigned long long pledgedSrcSize);
+
+/*! ZSTD_compressStream_generic() :
+ *  Private use only. To be called from zstdmt_compress.c in single-thread mode. */
+size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
+                                   ZSTD_outBuffer* output,
+                                   ZSTD_inBuffer* input,
+                                   ZSTD_EndDirective const flushMode);
+
+/*! ZSTD_getCParamsFromCDict() :
+ *  as the name implies */
+ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict);
+
+/* ZSTD_compressBegin_advanced_internal() :
+ * Private use only. To be called from zstdmt_compress.c. */
+size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
+                                    const void* dict, size_t dictSize,
+                                    ZSTD_dictMode_e dictMode,
+                                    ZSTD_CCtx_params params,
+                                    unsigned long long pledgedSrcSize);
+
+/* ZSTD_compress_advanced_internal() :
+ * Private use only. To be called from zstdmt_compress.c. */
+size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
+                                       void* dst, size_t dstCapacity,
+                                 const void* src, size_t srcSize,
+                                 const void* dict,size_t dictSize,
+                                 ZSTD_CCtx_params params);
+
+typedef struct {
+    blockType_e blockType;
+    U32 lastBlock;
+    U32 origSize;
+} blockProperties_t;
+
+/*! ZSTD_getcBlockSize() :
+*   Provides the size of compressed block from block header `src` */
+size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
+                          blockProperties_t* bpPtr);
+
+#if defined (__cplusplus)
+}
+#endif
+
 #endif   /* ZSTD_CCOMMON_H_MODULE */
diff --git a/contrib/zstd/zstd_lazy.c b/contrib/zstd/zstd_lazy.c
new file mode 100644
index 000000000..2a7f6a0fe
--- /dev/null
+++ b/contrib/zstd/zstd_lazy.c
@@ -0,0 +1,749 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_lazy.h"
+
+
+/*-*************************************
+*  Binary Tree search
+***************************************/
+/** ZSTD_insertBt1() : add one or multiple positions to tree.
+*   ip : assumed <= iend-8 .
+*   @return : nb of positions added */
+static U32 ZSTD_insertBt1(ZSTD_CCtx* zc, const BYTE* const ip, const U32 mls, const BYTE* const iend, U32 nbCompares,
+                          U32 extDict)
+{
+    U32*   const hashTable = zc->hashTable;
+    U32    const hashLog = zc->appliedParams.cParams.hashLog;
+    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32*   const bt = zc->chainTable;
+    U32    const btLog  = zc->appliedParams.cParams.chainLog - 1;
+    U32    const btMask = (1 << btLog) - 1;
+    U32 matchIndex = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = zc->base;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* match;
+    const U32 current = (U32)(ip-base);
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = smallerPtr + 1;
+    U32 dummy32;   /* to be nullified at the end */
+    U32 const windowLow = zc->lowLimit;
+    U32 matchEndIdx = current+8;
+    size_t bestLength = 8;
+#ifdef ZSTD_C_PREDICT
+    U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
+    U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
+    predictedSmall += (predictedSmall>0);
+    predictedLarge += (predictedLarge>0);
+#endif /* ZSTD_C_PREDICT */
+
+    assert(ip <= iend-8);   /* required for h calculation */
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* const nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+
+#ifdef ZSTD_C_PREDICT   /* note : can create issues when hlog small <= 11 */
+        const U32* predictPtr = bt + 2*((matchIndex-1) & btMask);   /* written this way, as bt is a roll buffer */
+        if (matchIndex == predictedSmall) {
+            /* no need to check length, result known */
+            *smallerPtr = matchIndex;
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            predictedSmall = predictPtr[1] + (predictPtr[1]>0);
+            continue;
+        }
+        if (matchIndex == predictedLarge) {
+            *largerPtr = matchIndex;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+            predictedLarge = predictPtr[0] + (predictPtr[0]>0);
+            continue;
+        }
+#endif
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            match = base + matchIndex;
+            if (match[matchLength] == ip[matchLength])
+                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            bestLength = matchLength;
+            if (matchLength > matchEndIdx - matchIndex)
+                matchEndIdx = matchIndex + (U32)matchLength;
+        }
+
+        if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+            break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
+
+        if (match[matchLength] < ip[matchLength]) {  /* necessarily within buffer */
+            /* match+1 is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+    if (bestLength > 384) return MIN(192, (U32)(bestLength - 384));   /* speed optimization */
+    if (matchEndIdx > current + 8) return matchEndIdx - (current + 8);
+    return 1;
+}
+
+
+static size_t ZSTD_insertBtAndFindBestMatch (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iend,
+                        size_t* offsetPtr,
+                        U32 nbCompares, const U32 mls,
+                        U32 extDict)
+{
+    U32*   const hashTable = zc->hashTable;
+    U32    const hashLog = zc->appliedParams.cParams.hashLog;
+    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32*   const bt = zc->chainTable;
+    U32    const btLog  = zc->appliedParams.cParams.chainLog - 1;
+    U32    const btMask = (1 << btLog) - 1;
+    U32 matchIndex  = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = zc->base;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const U32 current = (U32)(ip-base);
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    const U32 windowLow = zc->lowLimit;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = bt + 2*(current&btMask) + 1;
+    U32 matchEndIdx = current+8;
+    U32 dummy32;   /* to be nullified at the end */
+    size_t bestLength = 0;
+
+    assert(ip <= iend-8);   /* required for h calculation */
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* const nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+        const BYTE* match;
+
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            match = base + matchIndex;
+            if (match[matchLength] == ip[matchLength])
+                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            if (matchLength > matchEndIdx - matchIndex)
+                matchEndIdx = matchIndex + (U32)matchLength;
+            if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
+                bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
+            if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+        }
+
+        if (match[matchLength] < ip[matchLength]) {
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+
+    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
+    return bestLength;
+}
+
+
+void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
+{
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+
+    while(idx < target)
+        idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 0);
+}
+
+/** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
+static size_t ZSTD_BtFindBestMatch (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 0);
+}
+
+
+static size_t ZSTD_BtFindBestMatch_selectMLS (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default : /* includes case 3 */
+    case 4 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
+    case 5 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 7 :
+    case 6 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    }
+}
+
+
+void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
+{
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+
+    while (idx < target) idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 1);
+}
+
+
+/** Tree updater, providing best match */
+static size_t ZSTD_BtFindBestMatch_extDict (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 1);
+}
+
+
+static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default : /* includes case 3 */
+    case 4 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
+    case 5 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 7 :
+    case 6 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    }
+}
+
+
+
+/* *********************************
+*  Hash Chain
+***********************************/
+#define NEXT_IN_CHAIN(d, mask)   chainTable[(d) & mask]
+
+/* Update chains up to ip (excluded)
+   Assumption : always within prefix (i.e. not within extDict) */
+U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
+{
+    U32* const hashTable  = zc->hashTable;
+    const U32 hashLog = zc->appliedParams.cParams.hashLog;
+    U32* const chainTable = zc->chainTable;
+    const U32 chainMask = (1 << zc->appliedParams.cParams.chainLog) - 1;
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+
+    while(idx < target) { /* catch up */
+        size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
+        NEXT_IN_CHAIN(idx, chainMask) = hashTable[h];
+        hashTable[h] = idx;
+        idx++;
+    }
+
+    zc->nextToUpdate = target;
+    return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
+}
+
+
+/* inlining is important to hardwire a hot branch (template emulation) */
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_HcFindBestMatch_generic (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls, const U32 extDict)
+{
+    U32* const chainTable = zc->chainTable;
+    const U32 chainSize = (1 << zc->appliedParams.cParams.chainLog);
+    const U32 chainMask = chainSize-1;
+    const BYTE* const base = zc->base;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const U32 lowLimit = zc->lowLimit;
+    const U32 current = (U32)(ip-base);
+    const U32 minChain = current > chainSize ? current - chainSize : 0;
+    int nbAttempts=maxNbAttempts;
+    size_t ml=4-1;
+
+    /* HC4 match finder */
+    U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls);
+
+    for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
+        const BYTE* match;
+        size_t currentMl=0;
+        if ((!extDict) || matchIndex >= dictLimit) {
+            match = base + matchIndex;
+            if (match[ml] == ip[ml])   /* potentially better */
+                currentMl = ZSTD_count(ip, match, iLimit);
+        } else {
+            match = dictBase + matchIndex;
+            if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+                currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
+        }
+
+        /* save best solution */
+        if (currentMl > ml) {
+            ml = currentMl;
+            *offsetPtr = current - matchIndex + ZSTD_REP_MOVE;
+            if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+        }
+
+        if (matchIndex <= minChain) break;
+        matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
+    }
+
+    return ml;
+}
+
+
+FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
+                        ZSTD_CCtx* zc,
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default : /* includes case 3 */
+    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 0);
+    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 0);
+    case 7 :
+    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 0);
+    }
+}
+
+
+FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
+                        ZSTD_CCtx* zc,
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default : /* includes case 3 */
+    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 1);
+    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 1);
+    case 7 :
+    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 1);
+    }
+}
+
+
+/* *******************************
+*  Common parser - lazy strategy
+*********************************/
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
+                                       const void* src, size_t srcSize,
+                                       const U32 searchMethod, const U32 depth)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base + ctx->dictLimit;
+
+    U32 const maxSearches = 1 << ctx->appliedParams.cParams.searchLog;
+    U32 const mls = ctx->appliedParams.cParams.searchLength;
+
+    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
+                        size_t* offsetPtr,
+                        U32 maxNbAttempts, U32 matchLengthSearch);
+    searchMax_f const searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS;
+    U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1], savedOffset=0;
+
+    /* init */
+    ip += (ip==base);
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    {   U32 const maxRep = (U32)(ip-base);
+        if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
+        if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
+    }
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        size_t matchLength=0;
+        size_t offset=0;
+        const BYTE* start=ip+1;
+
+        /* check repCode */
+        if ((offset_1>0) & (MEM_read32(ip+1) == MEM_read32(ip+1 - offset_1))) {
+            /* repcode : we take it */
+            matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
+            if (depth==0) goto _storeSequence;
+        }
+
+        /* first search (depth 0) */
+        {   size_t offsetFound = 99999999;
+            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
+            if (ml2 > matchLength)
+                matchLength = ml2, start = ip, offset=offsetFound;
+        }
+
+        if (matchLength < 4) {
+            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
+            continue;
+        }
+
+        /* let's try to find a better solution */
+        if (depth>=1)
+        while (ip<ilimit) {
+            ip ++;
+            if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+                size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
+                int const gain2 = (int)(mlRep * 3);
+                int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
+                if ((mlRep >= 4) && (gain2 > gain1))
+                    matchLength = mlRep, offset = 0, start = ip;
+            }
+            {   size_t offset2=99999999;
+                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
+                if ((ml2 >= 4) && (gain2 > gain1)) {
+                    matchLength = ml2, offset = offset2, start = ip;
+                    continue;   /* search a better one */
+            }   }
+
+            /* let's find an even better one */
+            if ((depth==2) && (ip<ilimit)) {
+                ip ++;
+                if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+                    size_t const ml2 = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
+                    int const gain2 = (int)(ml2 * 4);
+                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
+                    if ((ml2 >= 4) && (gain2 > gain1))
+                        matchLength = ml2, offset = 0, start = ip;
+                }
+                {   size_t offset2=99999999;
+                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                    int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
+                    if ((ml2 >= 4) && (gain2 > gain1)) {
+                        matchLength = ml2, offset = offset2, start = ip;
+                        continue;
+            }   }   }
+            break;  /* nothing found : store previous solution */
+        }
+
+        /* NOTE:
+         * start[-offset+ZSTD_REP_MOVE-1] is undefined behavior.
+         * (-offset+ZSTD_REP_MOVE-1) is unsigned, and is added to start, which
+         * overflows the pointer, which is undefined behavior.
+         */
+        /* catch up */
+        if (offset) {
+            while ( (start > anchor)
+                 && (start > base+offset-ZSTD_REP_MOVE)
+                 && (start[-1] == (start-offset+ZSTD_REP_MOVE)[-1]) )  /* only search for offset within prefix */
+                { start--; matchLength++; }
+            offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
+        }
+        /* store sequence */
+_storeSequence:
+        {   size_t const litLength = start - anchor;
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            anchor = ip = start + matchLength;
+        }
+
+        /* check immediate repcode */
+        while ( (ip <= ilimit)
+             && ((offset_2>0)
+             & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+            /* store sequence */
+            matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
+            offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
+            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
+            ip += matchLength;
+            anchor = ip;
+            continue;   /* faster when present ... (?) */
+    }   }
+
+    /* Save reps for next block */
+    seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : savedOffset;
+    seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : savedOffset;
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 1, 2);
+}
+
+size_t ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 2);
+}
+
+size_t ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 1);
+}
+
+size_t ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 0);
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
+                                     const void* src, size_t srcSize,
+                                     const U32 searchMethod, const U32 depth)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base;
+    const U32 dictLimit = ctx->dictLimit;
+    const U32 lowestIndex = ctx->lowLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const dictEnd  = dictBase + dictLimit;
+    const BYTE* const dictStart  = dictBase + ctx->lowLimit;
+
+    const U32 maxSearches = 1 << ctx->appliedParams.cParams.searchLog;
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+
+    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
+                        size_t* offsetPtr,
+                        U32 maxNbAttempts, U32 matchLengthSearch);
+    searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS;
+
+    U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1];
+
+    /* init */
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ip += (ip == prefixStart);
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        size_t matchLength=0;
+        size_t offset=0;
+        const BYTE* start=ip+1;
+        U32 current = (U32)(ip-base);
+
+        /* check repCode */
+        {   const U32 repIndex = (U32)(current+1 - offset_1);
+            const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+            const BYTE* const repMatch = repBase + repIndex;
+            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))   /* intentional overflow */
+            if (MEM_read32(ip+1) == MEM_read32(repMatch)) {
+                /* repcode detected we should take it */
+                const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+                if (depth==0) goto _storeSequence;
+        }   }
+
+        /* first search (depth 0) */
+        {   size_t offsetFound = 99999999;
+            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
+            if (ml2 > matchLength)
+                matchLength = ml2, start = ip, offset=offsetFound;
+        }
+
+         if (matchLength < 4) {
+            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
+            continue;
+        }
+
+        /* let's try to find a better solution */
+        if (depth>=1)
+        while (ip<ilimit) {
+            ip ++;
+            current++;
+            /* check repCode */
+            if (offset) {
+                const U32 repIndex = (U32)(current - offset_1);
+                const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+                const BYTE* const repMatch = repBase + repIndex;
+                if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+                if (MEM_read32(ip) == MEM_read32(repMatch)) {
+                    /* repcode detected */
+                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                    size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+                    int const gain2 = (int)(repLength * 3);
+                    int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
+                    if ((repLength >= 4) && (gain2 > gain1))
+                        matchLength = repLength, offset = 0, start = ip;
+            }   }
+
+            /* search match, depth 1 */
+            {   size_t offset2=99999999;
+                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
+                if ((ml2 >= 4) && (gain2 > gain1)) {
+                    matchLength = ml2, offset = offset2, start = ip;
+                    continue;   /* search a better one */
+            }   }
+
+            /* let's find an even better one */
+            if ((depth==2) && (ip<ilimit)) {
+                ip ++;
+                current++;
+                /* check repCode */
+                if (offset) {
+                    const U32 repIndex = (U32)(current - offset_1);
+                    const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+                    const BYTE* const repMatch = repBase + repIndex;
+                    if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+                    if (MEM_read32(ip) == MEM_read32(repMatch)) {
+                        /* repcode detected */
+                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                        size_t const repLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+                        int const gain2 = (int)(repLength * 4);
+                        int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
+                        if ((repLength >= 4) && (gain2 > gain1))
+                            matchLength = repLength, offset = 0, start = ip;
+                }   }
+
+                /* search match, depth 2 */
+                {   size_t offset2=99999999;
+                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                    int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
+                    int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
+                    if ((ml2 >= 4) && (gain2 > gain1)) {
+                        matchLength = ml2, offset = offset2, start = ip;
+                        continue;
+            }   }   }
+            break;  /* nothing found : store previous solution */
+        }
+
+        /* catch up */
+        if (offset) {
+            U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
+            const BYTE* match = (matchIndex < dictLimit) ? dictBase + matchIndex : base + matchIndex;
+            const BYTE* const mStart = (matchIndex < dictLimit) ? dictStart : prefixStart;
+            while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; }  /* catch up */
+            offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
+        }
+
+        /* store sequence */
+_storeSequence:
+        {   size_t const litLength = start - anchor;
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            anchor = ip = start + matchLength;
+        }
+
+        /* check immediate repcode */
+        while (ip <= ilimit) {
+            const U32 repIndex = (U32)((ip-base) - offset_2);
+            const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+            const BYTE* const repMatch = repBase + repIndex;
+            if (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex > lowestIndex))  /* intentional overflow */
+            if (MEM_read32(ip) == MEM_read32(repMatch)) {
+                /* repcode detected we should take it */
+                const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
+                offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset history */
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
+                ip += matchLength;
+                anchor = ip;
+                continue;   /* faster when present ... (?) */
+            }
+            break;
+    }   }
+
+    /* Save reps for next block */
+    seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 0);
+}
+
+size_t ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 1);
+}
+
+size_t ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 2);
+}
+
+size_t ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 1, 2);
+}
diff --git a/contrib/zstd/zstd_lazy.h b/contrib/zstd/zstd_lazy.h
new file mode 100644
index 000000000..a9c4daed2
--- /dev/null
+++ b/contrib/zstd/zstd_lazy.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_LAZY_H
+#define ZSTD_LAZY_H
+
+#include "zstd_compress.h"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls);
+void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls);
+void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls);
+
+size_t ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+
+size_t ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_LAZY_H */
diff --git a/contrib/zstd/zstd_ldm.c b/contrib/zstd/zstd_ldm.c
new file mode 100644
index 000000000..e40007c19
--- /dev/null
+++ b/contrib/zstd/zstd_ldm.c
@@ -0,0 +1,703 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ */
+
+#include "zstd_ldm.h"
+
+#include "zstd_fast.h"          /* ZSTD_fillHashTable() */
+#include "zstd_double_fast.h"   /* ZSTD_fillDoubleHashTable() */
+
+#define LDM_BUCKET_SIZE_LOG 3
+#define LDM_MIN_MATCH_LENGTH 64
+#define LDM_HASH_LOG 20
+#define LDM_HASH_CHAR_OFFSET 10
+
+size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm)
+{
+    ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
+    params->enableLdm = enableLdm>0;
+    params->hashLog = LDM_HASH_LOG;
+    params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
+    params->minMatchLength = LDM_MIN_MATCH_LENGTH;
+    params->hashEveryLog = ZSTD_LDM_HASHEVERYLOG_NOTSET;
+    return 0;
+}
+
+void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog)
+{
+    if (params->hashEveryLog == ZSTD_LDM_HASHEVERYLOG_NOTSET) {
+        params->hashEveryLog =
+                windowLog < params->hashLog ? 0 : windowLog - params->hashLog;
+    }
+    params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
+}
+
+size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog) {
+    size_t const ldmHSize = ((size_t)1) << hashLog;
+    size_t const ldmBucketSizeLog = MIN(bucketSizeLog, hashLog);
+    size_t const ldmBucketSize =
+        ((size_t)1) << (hashLog - ldmBucketSizeLog);
+    return ldmBucketSize + (ldmHSize * (sizeof(ldmEntry_t)));
+}
+
+/** ZSTD_ldm_getSmallHash() :
+ *  numBits should be <= 32
+ *  If numBits==0, returns 0.
+ *  @return : the most significant numBits of value. */
+static U32 ZSTD_ldm_getSmallHash(U64 value, U32 numBits)
+{
+    assert(numBits <= 32);
+    return numBits == 0 ? 0 : (U32)(value >> (64 - numBits));
+}
+
+/** ZSTD_ldm_getChecksum() :
+ *  numBitsToDiscard should be <= 32
+ *  @return : the next most significant 32 bits after numBitsToDiscard */
+static U32 ZSTD_ldm_getChecksum(U64 hash, U32 numBitsToDiscard)
+{
+    assert(numBitsToDiscard <= 32);
+    return (hash >> (64 - 32 - numBitsToDiscard)) & 0xFFFFFFFF;
+}
+
+/** ZSTD_ldm_getTag() ;
+ *  Given the hash, returns the most significant numTagBits bits
+ *  after (32 + hbits) bits.
+ *
+ *  If there are not enough bits remaining, return the last
+ *  numTagBits bits. */
+static U32 ZSTD_ldm_getTag(U64 hash, U32 hbits, U32 numTagBits)
+{
+    assert(numTagBits < 32 && hbits <= 32);
+    if (32 - hbits < numTagBits) {
+        return hash & (((U32)1 << numTagBits) - 1);
+    } else {
+        return (hash >> (32 - hbits - numTagBits)) & (((U32)1 << numTagBits) - 1);
+    }
+}
+
+/** ZSTD_ldm_getBucket() :
+ *  Returns a pointer to the start of the bucket associated with hash. */
+static ldmEntry_t* ZSTD_ldm_getBucket(
+        ldmState_t* ldmState, size_t hash, ldmParams_t const ldmParams)
+{
+    return ldmState->hashTable + (hash << ldmParams.bucketSizeLog);
+}
+
+/** ZSTD_ldm_insertEntry() :
+ *  Insert the entry with corresponding hash into the hash table */
+static void ZSTD_ldm_insertEntry(ldmState_t* ldmState,
+                                 size_t const hash, const ldmEntry_t entry,
+                                 ldmParams_t const ldmParams)
+{
+    BYTE* const bucketOffsets = ldmState->bucketOffsets;
+    *(ZSTD_ldm_getBucket(ldmState, hash, ldmParams) + bucketOffsets[hash]) = entry;
+    bucketOffsets[hash]++;
+    bucketOffsets[hash] &= ((U32)1 << ldmParams.bucketSizeLog) - 1;
+}
+
+/** ZSTD_ldm_makeEntryAndInsertByTag() :
+ *
+ *  Gets the small hash, checksum, and tag from the rollingHash.
+ *
+ *  If the tag matches (1 << ldmParams.hashEveryLog)-1, then
+ *  creates an ldmEntry from the offset, and inserts it into the hash table.
+ *
+ *  hBits is the length of the small hash, which is the most significant hBits
+ *  of rollingHash. The checksum is the next 32 most significant bits, followed
+ *  by ldmParams.hashEveryLog bits that make up the tag. */
+static void ZSTD_ldm_makeEntryAndInsertByTag(ldmState_t* ldmState,
+                                             U64 const rollingHash,
+                                             U32 const hBits,
+                                             U32 const offset,
+                                             ldmParams_t const ldmParams)
+{
+    U32 const tag = ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog);
+    U32 const tagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
+    if (tag == tagMask) {
+        U32 const hash = ZSTD_ldm_getSmallHash(rollingHash, hBits);
+        U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
+        ldmEntry_t entry;
+        entry.offset = offset;
+        entry.checksum = checksum;
+        ZSTD_ldm_insertEntry(ldmState, hash, entry, ldmParams);
+    }
+}
+
+/** ZSTD_ldm_getRollingHash() :
+ *  Get a 64-bit hash using the first len bytes from buf.
+ *
+ *  Giving bytes s = s_1, s_2, ... s_k, the hash is defined to be
+ *  H(s) = s_1*(a^(k-1)) + s_2*(a^(k-2)) + ... + s_k*(a^0)
+ *
+ *  where the constant a is defined to be prime8bytes.
+ *
+ *  The implementation adds an offset to each byte, so
+ *  H(s) = (s_1 + HASH_CHAR_OFFSET)*(a^(k-1)) + ... */
+static U64 ZSTD_ldm_getRollingHash(const BYTE* buf, U32 len)
+{
+    U64 ret = 0;
+    U32 i;
+    for (i = 0; i < len; i++) {
+        ret *= prime8bytes;
+        ret += buf[i] + LDM_HASH_CHAR_OFFSET;
+    }
+    return ret;
+}
+
+/** ZSTD_ldm_ipow() :
+ *  Return base^exp. */
+static U64 ZSTD_ldm_ipow(U64 base, U64 exp)
+{
+    U64 ret = 1;
+    while (exp) {
+        if (exp & 1) { ret *= base; }
+        exp >>= 1;
+        base *= base;
+    }
+    return ret;
+}
+
+U64 ZSTD_ldm_getHashPower(U32 minMatchLength) {
+    assert(minMatchLength >= ZSTD_LDM_MINMATCH_MIN);
+    return ZSTD_ldm_ipow(prime8bytes, minMatchLength - 1);
+}
+
+/** ZSTD_ldm_updateHash() :
+ *  Updates hash by removing toRemove and adding toAdd. */
+static U64 ZSTD_ldm_updateHash(U64 hash, BYTE toRemove, BYTE toAdd, U64 hashPower)
+{
+    hash -= ((toRemove + LDM_HASH_CHAR_OFFSET) * hashPower);
+    hash *= prime8bytes;
+    hash += toAdd + LDM_HASH_CHAR_OFFSET;
+    return hash;
+}
+
+/** ZSTD_ldm_countBackwardsMatch() :
+ *  Returns the number of bytes that match backwards before pIn and pMatch.
+ *
+ *  We count only bytes where pMatch >= pBase and pIn >= pAnchor. */
+static size_t ZSTD_ldm_countBackwardsMatch(
+            const BYTE* pIn, const BYTE* pAnchor,
+            const BYTE* pMatch, const BYTE* pBase)
+{
+    size_t matchLength = 0;
+    while (pIn > pAnchor && pMatch > pBase && pIn[-1] == pMatch[-1]) {
+        pIn--;
+        pMatch--;
+        matchLength++;
+    }
+    return matchLength;
+}
+
+/** ZSTD_ldm_fillFastTables() :
+ *
+ *  Fills the relevant tables for the ZSTD_fast and ZSTD_dfast strategies.
+ *  This is similar to ZSTD_loadDictionaryContent.
+ *
+ *  The tables for the other strategies are filled within their
+ *  block compressors. */
+static size_t ZSTD_ldm_fillFastTables(ZSTD_CCtx* zc, const void* end)
+{
+    const BYTE* const iend = (const BYTE*)end;
+    const U32 mls = zc->appliedParams.cParams.searchLength;
+
+    switch(zc->appliedParams.cParams.strategy)
+    {
+    case ZSTD_fast:
+        ZSTD_fillHashTable(zc, iend, mls);
+        zc->nextToUpdate = (U32)(iend - zc->base);
+        break;
+
+    case ZSTD_dfast:
+        ZSTD_fillDoubleHashTable(zc, iend, mls);
+        zc->nextToUpdate = (U32)(iend - zc->base);
+        break;
+
+    case ZSTD_greedy:
+    case ZSTD_lazy:
+    case ZSTD_lazy2:
+    case ZSTD_btlazy2:
+    case ZSTD_btopt:
+    case ZSTD_btultra:
+        break;
+    default:
+        assert(0);  /* not possible : not a valid strategy id */
+    }
+
+    return 0;
+}
+
+/** ZSTD_ldm_fillLdmHashTable() :
+ *
+ *  Fills hashTable from (lastHashed + 1) to iend (non-inclusive).
+ *  lastHash is the rolling hash that corresponds to lastHashed.
+ *
+ *  Returns the rolling hash corresponding to position iend-1. */
+static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
+                                     U64 lastHash, const BYTE* lastHashed,
+                                     const BYTE* iend, const BYTE* base,
+                                     U32 hBits, ldmParams_t const ldmParams)
+{
+    U64 rollingHash = lastHash;
+    const BYTE* cur = lastHashed + 1;
+
+    while (cur < iend) {
+        rollingHash = ZSTD_ldm_updateHash(rollingHash, cur[-1],
+                                          cur[ldmParams.minMatchLength-1],
+                                          state->hashPower);
+        ZSTD_ldm_makeEntryAndInsertByTag(state,
+                                         rollingHash, hBits,
+                                         (U32)(cur - base), ldmParams);
+        ++cur;
+    }
+    return rollingHash;
+}
+
+
+/** ZSTD_ldm_limitTableUpdate() :
+ *
+ *  Sets cctx->nextToUpdate to a position corresponding closer to anchor
+ *  if it is far way
+ *  (after a long match, only update tables a limited amount). */
+static void ZSTD_ldm_limitTableUpdate(ZSTD_CCtx* cctx, const BYTE* anchor)
+{
+    U32 const current = (U32)(anchor - cctx->base);
+    if (current > cctx->nextToUpdate + 1024) {
+        cctx->nextToUpdate =
+            current - MIN(512, current - cctx->nextToUpdate - 1024);
+    }
+}
+
+typedef size_t (*ZSTD_blockCompressor) (ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+/* defined in zstd_compress.c */
+ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict);
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
+                                      const void* src, size_t srcSize)
+{
+    ldmState_t* const ldmState = &(cctx->ldmState);
+    const ldmParams_t ldmParams = cctx->appliedParams.ldmParams;
+    const U64 hashPower = ldmState->hashPower;
+    const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
+    const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
+    const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
+    seqStore_t* const seqStorePtr = &(cctx->seqStore);
+    const BYTE* const base = cctx->base;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = cctx->dictLimit;
+    const BYTE* const lowest = base + lowestIndex;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
+
+    const ZSTD_blockCompressor blockCompressor =
+        ZSTD_selectBlockCompressor(cctx->appliedParams.cParams.strategy, 0);
+    U32* const repToConfirm = seqStorePtr->repToConfirm;
+    U32 savedRep[ZSTD_REP_NUM];
+    U64 rollingHash = 0;
+    const BYTE* lastHashed = NULL;
+    size_t i, lastLiterals;
+
+    /* Save seqStorePtr->rep and copy repToConfirm */
+    for (i = 0; i < ZSTD_REP_NUM; i++)
+        savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
+
+    /* Main Search Loop */
+    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
+        size_t mLength;
+        U32 const current = (U32)(ip - base);
+        size_t forwardMatchLength = 0, backwardMatchLength = 0;
+        ldmEntry_t* bestEntry = NULL;
+        if (ip != istart) {
+            rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
+                                              lastHashed[ldmParams.minMatchLength],
+                                              hashPower);
+        } else {
+            rollingHash = ZSTD_ldm_getRollingHash(ip, ldmParams.minMatchLength);
+        }
+        lastHashed = ip;
+
+        /* Do not insert and do not look for a match */
+        if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
+                ldmTagMask) {
+           ip++;
+           continue;
+        }
+
+        /* Get the best entry and compute the match lengths */
+        {
+            ldmEntry_t* const bucket =
+                ZSTD_ldm_getBucket(ldmState,
+                                   ZSTD_ldm_getSmallHash(rollingHash, hBits),
+                                   ldmParams);
+            ldmEntry_t* cur;
+            size_t bestMatchLength = 0;
+            U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
+
+            for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
+                const BYTE* const pMatch = cur->offset + base;
+                size_t curForwardMatchLength, curBackwardMatchLength,
+                       curTotalMatchLength;
+                if (cur->checksum != checksum || cur->offset <= lowestIndex) {
+                    continue;
+                }
+
+                curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
+                if (curForwardMatchLength < ldmParams.minMatchLength) {
+                    continue;
+                }
+                curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
+                                             ip, anchor, pMatch, lowest);
+                curTotalMatchLength = curForwardMatchLength +
+                                      curBackwardMatchLength;
+
+                if (curTotalMatchLength > bestMatchLength) {
+                    bestMatchLength = curTotalMatchLength;
+                    forwardMatchLength = curForwardMatchLength;
+                    backwardMatchLength = curBackwardMatchLength;
+                    bestEntry = cur;
+                }
+            }
+        }
+
+        /* No match found -- continue searching */
+        if (bestEntry == NULL) {
+            ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash,
+                                             hBits, current,
+                                             ldmParams);
+            ip++;
+            continue;
+        }
+
+        /* Match found */
+        mLength = forwardMatchLength + backwardMatchLength;
+        ip -= backwardMatchLength;
+
+        /* Call the block compressor on the remaining literals */
+        {
+            U32 const matchIndex = bestEntry->offset;
+            const BYTE* const match = base + matchIndex - backwardMatchLength;
+            U32 const offset = (U32)(ip - match);
+
+            /* Overwrite rep codes */
+            for (i = 0; i < ZSTD_REP_NUM; i++)
+                seqStorePtr->rep[i] = repToConfirm[i];
+
+            /* Fill tables for block compressor */
+            ZSTD_ldm_limitTableUpdate(cctx, anchor);
+            ZSTD_ldm_fillFastTables(cctx, anchor);
+
+            /* Call block compressor and get remaining literals */
+            lastLiterals = blockCompressor(cctx, anchor, ip - anchor);
+            cctx->nextToUpdate = (U32)(ip - base);
+
+            /* Update repToConfirm with the new offset */
+            for (i = ZSTD_REP_NUM - 1; i > 0; i--)
+                repToConfirm[i] = repToConfirm[i-1];
+            repToConfirm[0] = offset;
+
+            /* Store the sequence with the leftover literals */
+            ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
+                          offset + ZSTD_REP_MOVE, mLength - MINMATCH);
+        }
+
+        /* Insert the current entry into the hash table */
+        ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
+                                         (U32)(lastHashed - base),
+                                         ldmParams);
+
+        assert(ip + backwardMatchLength == lastHashed);
+
+        /* Fill the hash table from lastHashed+1 to ip+mLength*/
+        /* Heuristic: don't need to fill the entire table at end of block */
+        if (ip + mLength < ilimit) {
+            rollingHash = ZSTD_ldm_fillLdmHashTable(
+                              ldmState, rollingHash, lastHashed,
+                              ip + mLength, base, hBits, ldmParams);
+            lastHashed = ip + mLength - 1;
+        }
+        ip += mLength;
+        anchor = ip;
+        /* Check immediate repcode */
+        while ( (ip < ilimit)
+             && ( (repToConfirm[1] > 0) && (repToConfirm[1] <= (U32)(ip-lowest))
+             && (MEM_read32(ip) == MEM_read32(ip - repToConfirm[1])) )) {
+
+            size_t const rLength = ZSTD_count(ip+4, ip+4-repToConfirm[1],
+                                              iend) + 4;
+            /* Swap repToConfirm[1] <=> repToConfirm[0] */
+            {
+                U32 const tmpOff = repToConfirm[1];
+                repToConfirm[1] = repToConfirm[0];
+                repToConfirm[0] = tmpOff;
+            }
+
+            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+
+            /* Fill the  hash table from lastHashed+1 to ip+rLength*/
+            if (ip + rLength < ilimit) {
+                rollingHash = ZSTD_ldm_fillLdmHashTable(
+                                ldmState, rollingHash, lastHashed,
+                                ip + rLength, base, hBits, ldmParams);
+                lastHashed = ip + rLength - 1;
+            }
+            ip += rLength;
+            anchor = ip;
+        }
+    }
+
+    /* Overwrite rep */
+    for (i = 0; i < ZSTD_REP_NUM; i++)
+        seqStorePtr->rep[i] = repToConfirm[i];
+
+    ZSTD_ldm_limitTableUpdate(cctx, anchor);
+    ZSTD_ldm_fillFastTables(cctx, anchor);
+
+    lastLiterals = blockCompressor(cctx, anchor, iend - anchor);
+    cctx->nextToUpdate = (U32)(iend - base);
+
+    /* Restore seqStorePtr->rep */
+    for (i = 0; i < ZSTD_REP_NUM; i++)
+        seqStorePtr->rep[i] = savedRep[i];
+
+    /* Return the last literals size */
+    return lastLiterals;
+}
+
+size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* ctx,
+                              const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_ldm_generic(ctx, src, srcSize);
+}
+
+static size_t ZSTD_compressBlock_ldm_extDict_generic(
+                                 ZSTD_CCtx* ctx,
+                                 const void* src, size_t srcSize)
+{
+    ldmState_t* const ldmState = &(ctx->ldmState);
+    const ldmParams_t ldmParams = ctx->appliedParams.ldmParams;
+    const U64 hashPower = ldmState->hashPower;
+    const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
+    const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
+    const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
+    seqStore_t* const seqStorePtr = &(ctx->seqStore);
+    const BYTE* const base = ctx->base;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const U32   lowestIndex = ctx->lowLimit;
+    const BYTE* const dictStart = dictBase + lowestIndex;
+    const U32   dictLimit = ctx->dictLimit;
+    const BYTE* const lowPrefixPtr = base + dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
+
+    const ZSTD_blockCompressor blockCompressor =
+        ZSTD_selectBlockCompressor(ctx->appliedParams.cParams.strategy, 1);
+    U32* const repToConfirm = seqStorePtr->repToConfirm;
+    U32 savedRep[ZSTD_REP_NUM];
+    U64 rollingHash = 0;
+    const BYTE* lastHashed = NULL;
+    size_t i, lastLiterals;
+
+    /* Save seqStorePtr->rep and copy repToConfirm */
+    for (i = 0; i < ZSTD_REP_NUM; i++) {
+        savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
+    }
+
+    /* Search Loop */
+    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
+        size_t mLength;
+        const U32 current = (U32)(ip-base);
+        size_t forwardMatchLength = 0, backwardMatchLength = 0;
+        ldmEntry_t* bestEntry = NULL;
+        if (ip != istart) {
+          rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
+                                       lastHashed[ldmParams.minMatchLength],
+                                       hashPower);
+        } else {
+            rollingHash = ZSTD_ldm_getRollingHash(ip, ldmParams.minMatchLength);
+        }
+        lastHashed = ip;
+
+        if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
+                ldmTagMask) {
+            /* Don't insert and don't look for a match */
+           ip++;
+           continue;
+        }
+
+        /* Get the best entry and compute the match lengths */
+        {
+            ldmEntry_t* const bucket =
+                ZSTD_ldm_getBucket(ldmState,
+                                   ZSTD_ldm_getSmallHash(rollingHash, hBits),
+                                   ldmParams);
+            ldmEntry_t* cur;
+            size_t bestMatchLength = 0;
+            U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
+
+            for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
+                const BYTE* const curMatchBase =
+                    cur->offset < dictLimit ? dictBase : base;
+                const BYTE* const pMatch = curMatchBase + cur->offset;
+                const BYTE* const matchEnd =
+                    cur->offset < dictLimit ? dictEnd : iend;
+                const BYTE* const lowMatchPtr =
+                    cur->offset < dictLimit ? dictStart : lowPrefixPtr;
+                size_t curForwardMatchLength, curBackwardMatchLength,
+                       curTotalMatchLength;
+
+                if (cur->checksum != checksum || cur->offset <= lowestIndex) {
+                    continue;
+                }
+
+                curForwardMatchLength = ZSTD_count_2segments(
+                                            ip, pMatch, iend,
+                                            matchEnd, lowPrefixPtr);
+                if (curForwardMatchLength < ldmParams.minMatchLength) {
+                    continue;
+                }
+                curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
+                                             ip, anchor, pMatch, lowMatchPtr);
+                curTotalMatchLength = curForwardMatchLength +
+                                      curBackwardMatchLength;
+
+                if (curTotalMatchLength > bestMatchLength) {
+                    bestMatchLength = curTotalMatchLength;
+                    forwardMatchLength = curForwardMatchLength;
+                    backwardMatchLength = curBackwardMatchLength;
+                    bestEntry = cur;
+                }
+            }
+        }
+
+        /* No match found -- continue searching */
+        if (bestEntry == NULL) {
+            ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
+                                             (U32)(lastHashed - base),
+                                             ldmParams);
+            ip++;
+            continue;
+        }
+
+        /* Match found */
+        mLength = forwardMatchLength + backwardMatchLength;
+        ip -= backwardMatchLength;
+
+        /* Call the block compressor on the remaining literals */
+        {
+            /* ip = current - backwardMatchLength
+             * The match is at (bestEntry->offset - backwardMatchLength) */
+            U32 const matchIndex = bestEntry->offset;
+            U32 const offset = current - matchIndex;
+
+            /* Overwrite rep codes */
+            for (i = 0; i < ZSTD_REP_NUM; i++)
+                seqStorePtr->rep[i] = repToConfirm[i];
+
+            /* Fill the hash table for the block compressor */
+            ZSTD_ldm_limitTableUpdate(ctx, anchor);
+            ZSTD_ldm_fillFastTables(ctx, anchor);
+
+            /* Call block compressor and get remaining literals  */
+            lastLiterals = blockCompressor(ctx, anchor, ip - anchor);
+            ctx->nextToUpdate = (U32)(ip - base);
+
+            /* Update repToConfirm with the new offset */
+            for (i = ZSTD_REP_NUM - 1; i > 0; i--)
+                repToConfirm[i] = repToConfirm[i-1];
+            repToConfirm[0] = offset;
+
+            /* Store the sequence with the leftover literals */
+            ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
+                          offset + ZSTD_REP_MOVE, mLength - MINMATCH);
+        }
+
+        /* Insert the current entry into the hash table */
+        ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
+                                         (U32)(lastHashed - base),
+                                         ldmParams);
+
+        /* Fill the hash table from lastHashed+1 to ip+mLength */
+        assert(ip + backwardMatchLength == lastHashed);
+        if (ip + mLength < ilimit) {
+            rollingHash = ZSTD_ldm_fillLdmHashTable(
+                              ldmState, rollingHash, lastHashed,
+                              ip + mLength, base, hBits,
+                              ldmParams);
+            lastHashed = ip + mLength - 1;
+        }
+        ip += mLength;
+        anchor = ip;
+
+        /* check immediate repcode */
+        while (ip < ilimit) {
+            U32 const current2 = (U32)(ip-base);
+            U32 const repIndex2 = current2 - repToConfirm[1];
+            const BYTE* repMatch2 = repIndex2 < dictLimit ?
+                                    dictBase + repIndex2 : base + repIndex2;
+            if ( (((U32)((dictLimit-1) - repIndex2) >= 3) &
+                        (repIndex2 > lowestIndex))  /* intentional overflow */
+               && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
+                const BYTE* const repEnd2 = repIndex2 < dictLimit ?
+                                            dictEnd : iend;
+                size_t const repLength2 =
+                        ZSTD_count_2segments(ip+4, repMatch2+4, iend,
+                                             repEnd2, lowPrefixPtr) + 4;
+
+                U32 tmpOffset = repToConfirm[1];
+                repToConfirm[1] = repToConfirm[0];
+                repToConfirm[0] = tmpOffset;
+
+                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
+
+                /* Fill the  hash table from lastHashed+1 to ip+repLength2*/
+                if (ip + repLength2 < ilimit) {
+                    rollingHash = ZSTD_ldm_fillLdmHashTable(
+                                      ldmState, rollingHash, lastHashed,
+                                      ip + repLength2, base, hBits,
+                                      ldmParams);
+                    lastHashed = ip + repLength2 - 1;
+                }
+                ip += repLength2;
+                anchor = ip;
+                continue;
+            }
+            break;
+        }
+    }
+
+    /* Overwrite rep */
+    for (i = 0; i < ZSTD_REP_NUM; i++)
+        seqStorePtr->rep[i] = repToConfirm[i];
+
+    ZSTD_ldm_limitTableUpdate(ctx, anchor);
+    ZSTD_ldm_fillFastTables(ctx, anchor);
+
+    /* Call the block compressor one last time on the last literals */
+    lastLiterals = blockCompressor(ctx, anchor, iend - anchor);
+    ctx->nextToUpdate = (U32)(iend - base);
+
+    /* Restore seqStorePtr->rep */
+    for (i = 0; i < ZSTD_REP_NUM; i++)
+        seqStorePtr->rep[i] = savedRep[i];
+
+    /* Return the last literals size */
+    return lastLiterals;
+}
+
+size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
+                                      const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_ldm_extDict_generic(ctx, src, srcSize);
+}
diff --git a/contrib/zstd/zstd_ldm.h b/contrib/zstd/zstd_ldm.h
new file mode 100644
index 000000000..7a6248399
--- /dev/null
+++ b/contrib/zstd/zstd_ldm.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ */
+
+#ifndef ZSTD_LDM_H
+#define ZSTD_LDM_H
+
+#include "zstd_compress.h"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-*************************************
+*  Long distance matching
+***************************************/
+
+#define ZSTD_LDM_WINDOW_LOG 27
+#define ZSTD_LDM_HASHEVERYLOG_NOTSET 9999
+
+/** ZSTD_compressBlock_ldm_generic() :
+ *
+ *  This is a block compressor intended for long distance matching.
+ *
+ *  The function searches for matches of length at least
+ *  ldmParams.minMatchLength using a hash table in cctx->ldmState.
+ *  Matches can be at a distance of up to cParams.windowLog.
+ *
+ *  Upon finding a match, the unmatched literals are compressed using a
+ *  ZSTD_blockCompressor (depending on the strategy in the compression
+ *  parameters), which stores the matched sequences. The "long distance"
+ *  match is then stored with the remaining literals from the
+ *  ZSTD_blockCompressor. */
+size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* cctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
+                                      const void* src, size_t srcSize);
+
+/** ZSTD_ldm_initializeParameters() :
+ *  Initialize the long distance matching parameters to their default values. */
+size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm);
+
+/** ZSTD_ldm_getTableSize() :
+ *  Estimate the space needed for long distance matching tables. */
+size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog);
+
+/** ZSTD_ldm_getTableSize() :
+ *  Return prime8bytes^(minMatchLength-1) */
+U64 ZSTD_ldm_getHashPower(U32 minMatchLength);
+
+/** ZSTD_ldm_adjustParameters() :
+ *  If the params->hashEveryLog is not set, set it to its default value based on
+ *  windowLog and params->hashLog.
+ *
+ *  Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
+ *  params->hashLog if it is not). */
+void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog);
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_FAST_H */
diff --git a/contrib/zstd/zstd_opt.c b/contrib/zstd/zstd_opt.c
new file mode 100644
index 000000000..fd102da2d
--- /dev/null
+++ b/contrib/zstd/zstd_opt.c
@@ -0,0 +1,949 @@
+/*
+ * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#include "zstd_opt.h"
+#include "zstd_lazy.h"
+
+
+#define ZSTD_LITFREQ_ADD    2
+#define ZSTD_FREQ_DIV       4
+#define ZSTD_MAX_PRICE      (1<<30)
+
+/*-*************************************
+*  Price functions for optimal parser
+***************************************/
+static void ZSTD_setLog2Prices(optState_t* optPtr)
+{
+    optPtr->log2matchLengthSum = ZSTD_highbit32(optPtr->matchLengthSum+1);
+    optPtr->log2litLengthSum = ZSTD_highbit32(optPtr->litLengthSum+1);
+    optPtr->log2litSum = ZSTD_highbit32(optPtr->litSum+1);
+    optPtr->log2offCodeSum = ZSTD_highbit32(optPtr->offCodeSum+1);
+    optPtr->factor = 1 + ((optPtr->litSum>>5) / optPtr->litLengthSum) + ((optPtr->litSum<<1) / (optPtr->litSum + optPtr->matchSum));
+}
+
+
+static void ZSTD_rescaleFreqs(optState_t* optPtr, const BYTE* src, size_t srcSize)
+{
+    unsigned u;
+
+    optPtr->cachedLiterals = NULL;
+    optPtr->cachedPrice = optPtr->cachedLitLength = 0;
+    optPtr->staticPrices = 0;
+
+    if (optPtr->litLengthSum == 0) {
+        if (srcSize <= 1024) optPtr->staticPrices = 1;
+
+        assert(optPtr->litFreq!=NULL);
+        for (u=0; u<=MaxLit; u++)
+            optPtr->litFreq[u] = 0;
+        for (u=0; u<srcSize; u++)
+            optPtr->litFreq[src[u]]++;
+
+        optPtr->litSum = 0;
+        optPtr->litLengthSum = MaxLL+1;
+        optPtr->matchLengthSum = MaxML+1;
+        optPtr->offCodeSum = (MaxOff+1);
+        optPtr->matchSum = (ZSTD_LITFREQ_ADD<<Litbits);
+
+        for (u=0; u<=MaxLit; u++) {
+            optPtr->litFreq[u] = 1 + (optPtr->litFreq[u]>>ZSTD_FREQ_DIV);
+            optPtr->litSum += optPtr->litFreq[u];
+        }
+        for (u=0; u<=MaxLL; u++)
+            optPtr->litLengthFreq[u] = 1;
+        for (u=0; u<=MaxML; u++)
+            optPtr->matchLengthFreq[u] = 1;
+        for (u=0; u<=MaxOff; u++)
+            optPtr->offCodeFreq[u] = 1;
+    } else {
+        optPtr->matchLengthSum = 0;
+        optPtr->litLengthSum = 0;
+        optPtr->offCodeSum = 0;
+        optPtr->matchSum = 0;
+        optPtr->litSum = 0;
+
+        for (u=0; u<=MaxLit; u++) {
+            optPtr->litFreq[u] = 1 + (optPtr->litFreq[u]>>(ZSTD_FREQ_DIV+1));
+            optPtr->litSum += optPtr->litFreq[u];
+        }
+        for (u=0; u<=MaxLL; u++) {
+            optPtr->litLengthFreq[u] = 1 + (optPtr->litLengthFreq[u]>>(ZSTD_FREQ_DIV+1));
+            optPtr->litLengthSum += optPtr->litLengthFreq[u];
+        }
+        for (u=0; u<=MaxML; u++) {
+            optPtr->matchLengthFreq[u] = 1 + (optPtr->matchLengthFreq[u]>>ZSTD_FREQ_DIV);
+            optPtr->matchLengthSum += optPtr->matchLengthFreq[u];
+            optPtr->matchSum += optPtr->matchLengthFreq[u] * (u + 3);
+        }
+        optPtr->matchSum *= ZSTD_LITFREQ_ADD;
+        for (u=0; u<=MaxOff; u++) {
+            optPtr->offCodeFreq[u] = 1 + (optPtr->offCodeFreq[u]>>ZSTD_FREQ_DIV);
+            optPtr->offCodeSum += optPtr->offCodeFreq[u];
+        }
+    }
+
+    ZSTD_setLog2Prices(optPtr);
+}
+
+
+static U32 ZSTD_getLiteralPrice(optState_t* optPtr, U32 litLength, const BYTE* literals)
+{
+    U32 price, u;
+
+    if (optPtr->staticPrices)
+        return ZSTD_highbit32((U32)litLength+1) + (litLength*6);
+
+    if (litLength == 0)
+        return optPtr->log2litLengthSum - ZSTD_highbit32(optPtr->litLengthFreq[0]+1);
+
+    /* literals */
+    if (optPtr->cachedLiterals == literals) {
+        U32 const additional = litLength - optPtr->cachedLitLength;
+        const BYTE* literals2 = optPtr->cachedLiterals + optPtr->cachedLitLength;
+        price = optPtr->cachedPrice + additional * optPtr->log2litSum;
+        for (u=0; u < additional; u++)
+            price -= ZSTD_highbit32(optPtr->litFreq[literals2[u]]+1);
+        optPtr->cachedPrice = price;
+        optPtr->cachedLitLength = litLength;
+    } else {
+        price = litLength * optPtr->log2litSum;
+        for (u=0; u < litLength; u++)
+            price -= ZSTD_highbit32(optPtr->litFreq[literals[u]]+1);
+
+        if (litLength >= 12) {
+            optPtr->cachedLiterals = literals;
+            optPtr->cachedPrice = price;
+            optPtr->cachedLitLength = litLength;
+        }
+    }
+
+    /* literal Length */
+    {   const BYTE LL_deltaCode = 19;
+        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+        price += LL_bits[llCode] + optPtr->log2litLengthSum - ZSTD_highbit32(optPtr->litLengthFreq[llCode]+1);
+    }
+
+    return price;
+}
+
+
+FORCE_INLINE_TEMPLATE U32 ZSTD_getPrice(optState_t* optPtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength, const int ultra)
+{
+    /* offset */
+    U32 price;
+    BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
+
+    if (optPtr->staticPrices)
+        return ZSTD_getLiteralPrice(optPtr, litLength, literals) + ZSTD_highbit32((U32)matchLength+1) + 16 + offCode;
+
+    price = offCode + optPtr->log2offCodeSum - ZSTD_highbit32(optPtr->offCodeFreq[offCode]+1);
+    if (!ultra && offCode >= 20) price += (offCode-19)*2;
+
+    /* match Length */
+    {   const BYTE ML_deltaCode = 36;
+        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
+        price += ML_bits[mlCode] + optPtr->log2matchLengthSum - ZSTD_highbit32(optPtr->matchLengthFreq[mlCode]+1);
+    }
+
+    return price + ZSTD_getLiteralPrice(optPtr, litLength, literals) + optPtr->factor;
+}
+
+
+static void ZSTD_updatePrice(optState_t* optPtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength)
+{
+    U32 u;
+
+    /* literals */
+    optPtr->litSum += litLength*ZSTD_LITFREQ_ADD;
+    for (u=0; u < litLength; u++)
+        optPtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
+
+    /* literal Length */
+    {   const BYTE LL_deltaCode = 19;
+        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+        optPtr->litLengthFreq[llCode]++;
+        optPtr->litLengthSum++;
+    }
+
+    /* match offset */
+    {   BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
+        optPtr->offCodeSum++;
+        optPtr->offCodeFreq[offCode]++;
+    }
+
+    /* match Length */
+    {   const BYTE ML_deltaCode = 36;
+        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
+        optPtr->matchLengthFreq[mlCode]++;
+        optPtr->matchLengthSum++;
+    }
+
+    ZSTD_setLog2Prices(optPtr);
+}
+
+
+#define SET_PRICE(pos, mlen_, offset_, litlen_, price_)   \
+    {                                                 \
+        while (last_pos < pos)  { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; } \
+        opt[pos].mlen = mlen_;                         \
+        opt[pos].off = offset_;                        \
+        opt[pos].litlen = litlen_;                     \
+        opt[pos].price = price_;                       \
+    }
+
+
+/* function safe only for comparisons */
+static U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
+{
+    switch (length)
+    {
+    default :
+    case 4 : return MEM_read32(memPtr);
+    case 3 : if (MEM_isLittleEndian())
+                return MEM_read32(memPtr)<<8;
+             else
+                return MEM_read32(memPtr)>>8;
+    }
+}
+
+
+/* Update hashTable3 up to ip (excluded)
+   Assumption : always within prefix (i.e. not within extDict) */
+static
+U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* zc, const BYTE* ip)
+{
+    U32* const hashTable3  = zc->hashTable3;
+    U32 const hashLog3  = zc->hashLog3;
+    const BYTE* const base = zc->base;
+    U32 idx = zc->nextToUpdate3;
+    const U32 target = zc->nextToUpdate3 = (U32)(ip - base);
+    const size_t hash3 = ZSTD_hash3Ptr(ip, hashLog3);
+
+    while(idx < target) {
+        hashTable3[ZSTD_hash3Ptr(base+idx, hashLog3)] = idx;
+        idx++;
+    }
+
+    return hashTable3[hash3];
+}
+
+
+/*-*************************************
+*  Binary Tree search
+***************************************/
+static U32 ZSTD_insertBtAndGetAllMatches (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        U32 nbCompares, const U32 mls,
+                        U32 extDict, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    const BYTE* const base = zc->base;
+    const U32 current = (U32)(ip-base);
+    const U32 hashLog = zc->appliedParams.cParams.hashLog;
+    const size_t h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32* const hashTable = zc->hashTable;
+    U32 matchIndex  = hashTable[h];
+    U32* const bt   = zc->chainTable;
+    const U32 btLog = zc->appliedParams.cParams.chainLog - 1;
+    const U32 btMask= (1U << btLog) - 1;
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const dictBase = zc->dictBase;
+    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    const U32 windowLow = zc->lowLimit;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = bt + 2*(current&btMask) + 1;
+    U32 matchEndIdx = current+8;
+    U32 dummy32;   /* to be nullified at the end */
+    U32 mnum = 0;
+
+    const U32 minMatch = (mls == 3) ? 3 : 4;
+    size_t bestLength = minMatchLen-1;
+
+    if (minMatch == 3) { /* HC3 match finder */
+        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3 (zc, ip);
+        if (matchIndex3>windowLow && (current - matchIndex3 < (1<<18))) {
+            const BYTE* match;
+            size_t currentMl=0;
+            if ((!extDict) || matchIndex3 >= dictLimit) {
+                match = base + matchIndex3;
+                if (match[bestLength] == ip[bestLength]) currentMl = ZSTD_count(ip, match, iLimit);
+            } else {
+                match = dictBase + matchIndex3;
+                if (ZSTD_readMINMATCH(match, MINMATCH) == ZSTD_readMINMATCH(ip, MINMATCH))    /* assumption : matchIndex3 <= dictLimit-4 (by table construction) */
+                    currentMl = ZSTD_count_2segments(ip+MINMATCH, match+MINMATCH, iLimit, dictEnd, prefixStart) + MINMATCH;
+            }
+
+            /* save best solution */
+            if (currentMl > bestLength) {
+                bestLength = currentMl;
+                matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex3;
+                matches[mnum].len = (U32)currentMl;
+                mnum++;
+                if (currentMl > ZSTD_OPT_NUM) goto update;
+                if (ip+currentMl == iLimit) goto update; /* best possible, and avoid read overflow*/
+            }
+        }
+    }
+
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+        const BYTE* match;
+
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            match = base + matchIndex;
+            if (match[matchLength] == ip[matchLength]) {
+                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iLimit) +1;
+            }
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            if (matchLength > matchEndIdx - matchIndex) matchEndIdx = matchIndex + (U32)matchLength;
+            bestLength = matchLength;
+            matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex;
+            matches[mnum].len = (U32)matchLength;
+            mnum++;
+            if (matchLength > ZSTD_OPT_NUM) break;
+            if (ip+matchLength == iLimit)   /* equal : no way to know if inf or sup */
+                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+        }
+
+        if (match[matchLength] < ip[matchLength]) {
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+
+update:
+    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
+    return mnum;
+}
+
+
+/** Tree updater, providing best match */
+static U32 ZSTD_BtGetAllMatches (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 0, matches, minMatchLen);
+}
+
+
+static U32 ZSTD_BtGetAllMatches_selectMLS (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iHighLimit,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    switch(matchLengthSearch)
+    {
+    case 3 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
+    default :
+    case 4 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
+    case 5 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 7 :
+    case 6 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
+    }
+}
+
+/** Tree updater, providing best match */
+static U32 ZSTD_BtGetAllMatches_extDict (
+                        ZSTD_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 1, matches, minMatchLen);
+}
+
+
+static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
+                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iHighLimit,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
+{
+    switch(matchLengthSearch)
+    {
+    case 3 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
+    default :
+    case 4 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
+    case 5 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 7 :
+    case 6 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
+    }
+}
+
+
+/*-*******************************
+*  Optimal parser
+*********************************/
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
+                                      const void* src, size_t srcSize, const int ultra)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    optState_t* optStatePtr = &(ctx->optState);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base;
+    const BYTE* const prefixStart = base + ctx->dictLimit;
+
+    const U32 maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
+    const U32 sufficient_len = ctx->appliedParams.cParams.targetLength;
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    const U32 minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
+
+    ZSTD_optimal_t* opt = optStatePtr->priceTable;
+    ZSTD_match_t* matches = optStatePtr->matchTable;
+    const BYTE* inr;
+    U32 offset, rep[ZSTD_REP_NUM];
+
+    /* init */
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize);
+    ip += (ip==prefixStart);
+    { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=seqStorePtr->rep[i]; }
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        U32 cur, match_num, last_pos, litlen, price;
+        U32 u, mlen, best_mlen, best_off, litLength;
+        memset(opt, 0, sizeof(ZSTD_optimal_t));
+        last_pos = 0;
+        litlen = (U32)(ip - anchor);
+
+        /* check repCode */
+        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
+            for (i=(ip == anchor); i<last_i; i++) {
+                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
+                if ( (repCur > 0) && (repCur < (S32)(ip-prefixStart))
+                    && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repCur, minMatch))) {
+                    mlen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repCur, iend) + minMatch;
+                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
+                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
+                        goto _storeSequence;
+                    }
+                    best_off = i - (ip == anchor);
+                    do {
+                        price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                        if (mlen > last_pos || price < opt[mlen].price)
+                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
+                        mlen--;
+                    } while (mlen >= minMatch);
+        }   }   }
+
+        match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, ip, iend, maxSearches, mls, matches, minMatch);
+
+        if (!last_pos && !match_num) { ip++; continue; }
+
+        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+            best_mlen = matches[match_num-1].len;
+            best_off = matches[match_num-1].off;
+            cur = 0;
+            last_pos = 1;
+            goto _storeSequence;
+        }
+
+        /* set prices using matches at position = 0 */
+        best_mlen = (last_pos) ? last_pos : minMatch;
+        for (u = 0; u < match_num; u++) {
+            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+            best_mlen = matches[u].len;
+            while (mlen <= best_mlen) {
+                price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                if (mlen > last_pos || price < opt[mlen].price)
+                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);   /* note : macro modifies last_pos */
+                mlen++;
+        }   }
+
+        if (last_pos < minMatch) { ip++; continue; }
+
+        /* initialize opt[0] */
+        { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
+        opt[0].mlen = 1;
+        opt[0].litlen = litlen;
+
+         /* check further positions */
+        for (cur = 1; cur <= last_pos; cur++) {
+           inr = ip + cur;
+
+           if (opt[cur-1].mlen == 1) {
+                litlen = opt[cur-1].litlen + 1;
+                if (cur > litlen) {
+                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-litlen);
+                } else
+                    price = ZSTD_getLiteralPrice(optStatePtr, litlen, anchor);
+           } else {
+                litlen = 1;
+                price = opt[cur - 1].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-1);
+           }
+
+           if (cur > last_pos || price <= opt[cur].price)
+                SET_PRICE(cur, 1, 0, litlen, price);
+
+           if (cur == last_pos) break;
+
+           if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
+               continue;
+
+           mlen = opt[cur].mlen;
+           if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
+                opt[cur].rep[2] = opt[cur-mlen].rep[1];
+                opt[cur].rep[1] = opt[cur-mlen].rep[0];
+                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
+           } else {
+                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
+                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
+                opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
+           }
+
+            best_mlen = minMatch;
+            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
+                for (i=(opt[cur].mlen != 1); i<last_i; i++) {  /* check rep */
+                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
+                    if ( (repCur > 0) && (repCur < (S32)(inr-prefixStart))
+                       && (ZSTD_readMINMATCH(inr, minMatch) == ZSTD_readMINMATCH(inr - repCur, minMatch))) {
+                       mlen = (U32)ZSTD_count(inr+minMatch, inr+minMatch - repCur, iend) + minMatch;
+
+                       if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
+                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
+                            goto _storeSequence;
+                       }
+
+                       best_off = i - (opt[cur].mlen != 1);
+                       if (mlen > best_mlen) best_mlen = mlen;
+
+                       do {
+                           if (opt[cur].mlen == 1) {
+                                litlen = opt[cur].litlen;
+                                if (cur > litlen) {
+                                    price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
+                                } else
+                                    price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                            } else {
+                                litlen = 0;
+                                price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
+                            }
+
+                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
+                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
+                            mlen--;
+                        } while (mlen >= minMatch);
+            }   }   }
+
+            match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, inr, iend, maxSearches, mls, matches, best_mlen);
+
+            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+                best_mlen = matches[match_num-1].len;
+                best_off = matches[match_num-1].off;
+                last_pos = cur + 1;
+                goto _storeSequence;
+            }
+
+            /* set prices using matches at position = cur */
+            for (u = 0; u < match_num; u++) {
+                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+                best_mlen = matches[u].len;
+
+                while (mlen <= best_mlen) {
+                    if (opt[cur].mlen == 1) {
+                        litlen = opt[cur].litlen;
+                        if (cur > litlen)
+                            price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
+                        else
+                            price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                    } else {
+                        litlen = 0;
+                        price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
+                    }
+
+                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
+                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
+
+                    mlen++;
+        }   }   }
+
+        best_mlen = opt[last_pos].mlen;
+        best_off = opt[last_pos].off;
+        cur = last_pos - best_mlen;
+
+        /* store sequence */
+_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
+        opt[0].mlen = 1;
+
+        while (1) {
+            mlen = opt[cur].mlen;
+            offset = opt[cur].off;
+            opt[cur].mlen = best_mlen;
+            opt[cur].off = best_off;
+            best_mlen = mlen;
+            best_off = offset;
+            if (mlen > cur) break;
+            cur -= mlen;
+        }
+
+        for (u = 0; u <= last_pos;) {
+            u += opt[u].mlen;
+        }
+
+        for (cur=0; cur < last_pos; ) {
+            mlen = opt[cur].mlen;
+            if (mlen == 1) { ip++; cur++; continue; }
+            offset = opt[cur].off;
+            cur += mlen;
+            litLength = (U32)(ip - anchor);
+
+            if (offset > ZSTD_REP_MOVE_OPT) {
+                rep[2] = rep[1];
+                rep[1] = rep[0];
+                rep[0] = offset - ZSTD_REP_MOVE_OPT;
+                offset--;
+            } else {
+                if (offset != 0) {
+                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
+                    if (offset != 1) rep[2] = rep[1];
+                    rep[1] = rep[0];
+                    rep[0] = best_off;
+                }
+                if (litLength==0) offset--;
+            }
+
+            ZSTD_updatePrice(optStatePtr, litLength, anchor, offset, mlen-MINMATCH);
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
+            anchor = ip = ip + mlen;
+    }    }   /* for (cur=0; cur < last_pos; ) */
+
+    /* Save reps for next block */
+    { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->repToConfirm[i] = rep[i]; }
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0);
+}
+
+size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 1);
+}
+
+
+FORCE_INLINE_TEMPLATE
+size_t ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
+                                     const void* src, size_t srcSize, const int ultra)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    optState_t* optStatePtr = &(ctx->optState);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+    const BYTE* const base = ctx->base;
+    const U32 lowestIndex = ctx->lowLimit;
+    const U32 dictLimit = ctx->dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const dictEnd  = dictBase + dictLimit;
+
+    const U32 maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
+    const U32 sufficient_len = ctx->appliedParams.cParams.targetLength;
+    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    const U32 minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
+
+    ZSTD_optimal_t* opt = optStatePtr->priceTable;
+    ZSTD_match_t* matches = optStatePtr->matchTable;
+    const BYTE* inr;
+
+    /* init */
+    U32 offset, rep[ZSTD_REP_NUM];
+    { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=seqStorePtr->rep[i]; }
+
+    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize);
+    ip += (ip==prefixStart);
+
+    /* Match Loop */
+    while (ip < ilimit) {
+        U32 cur, match_num, last_pos, litlen, price;
+        U32 u, mlen, best_mlen, best_off, litLength;
+        U32 current = (U32)(ip-base);
+        memset(opt, 0, sizeof(ZSTD_optimal_t));
+        last_pos = 0;
+        opt[0].litlen = (U32)(ip - anchor);
+
+        /* check repCode */
+        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
+            for (i = (ip==anchor); i<last_i; i++) {
+                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
+                const U32 repIndex = (U32)(current - repCur);
+                const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+                const BYTE* const repMatch = repBase + repIndex;
+                if ( (repCur > 0 && repCur <= (S32)current)
+                   && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
+                   && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
+                    /* repcode detected we should take it */
+                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                    mlen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
+
+                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
+                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
+                        goto _storeSequence;
+                    }
+
+                    best_off = i - (ip==anchor);
+                    litlen = opt[0].litlen;
+                    do {
+                        price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                        if (mlen > last_pos || price < opt[mlen].price)
+                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
+                        mlen--;
+                    } while (mlen >= minMatch);
+        }   }   }
+
+        match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, ip, iend, maxSearches, mls, matches, minMatch);  /* first search (depth 0) */
+
+        if (!last_pos && !match_num) { ip++; continue; }
+
+        { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
+        opt[0].mlen = 1;
+
+        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+            best_mlen = matches[match_num-1].len;
+            best_off = matches[match_num-1].off;
+            cur = 0;
+            last_pos = 1;
+            goto _storeSequence;
+        }
+
+        best_mlen = (last_pos) ? last_pos : minMatch;
+
+        /* set prices using matches at position = 0 */
+        for (u = 0; u < match_num; u++) {
+            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+            best_mlen = matches[u].len;
+            litlen = opt[0].litlen;
+            while (mlen <= best_mlen) {
+                price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                if (mlen > last_pos || price < opt[mlen].price)
+                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);
+                mlen++;
+        }   }
+
+        if (last_pos < minMatch) {
+            ip++; continue;
+        }
+
+        /* check further positions */
+        for (cur = 1; cur <= last_pos; cur++) {
+            inr = ip + cur;
+
+            if (opt[cur-1].mlen == 1) {
+                litlen = opt[cur-1].litlen + 1;
+                if (cur > litlen) {
+                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-litlen);
+                } else
+                    price = ZSTD_getLiteralPrice(optStatePtr, litlen, anchor);
+            } else {
+                litlen = 1;
+                price = opt[cur - 1].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-1);
+            }
+
+            if (cur > last_pos || price <= opt[cur].price)
+                SET_PRICE(cur, 1, 0, litlen, price);
+
+            if (cur == last_pos) break;
+
+            if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
+                continue;
+
+            mlen = opt[cur].mlen;
+            if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
+                opt[cur].rep[2] = opt[cur-mlen].rep[1];
+                opt[cur].rep[1] = opt[cur-mlen].rep[0];
+                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
+            } else {
+                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
+                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
+                opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
+            }
+
+            best_mlen = minMatch;
+            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
+                for (i = (mlen != 1); i<last_i; i++) {
+                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
+                    const U32 repIndex = (U32)(current+cur - repCur);
+                    const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
+                    const BYTE* const repMatch = repBase + repIndex;
+                    if ( (repCur > 0 && repCur <= (S32)(current+cur))
+                      && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
+                      && (ZSTD_readMINMATCH(inr, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
+                        /* repcode detected */
+                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
+                        mlen = (U32)ZSTD_count_2segments(inr+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
+
+                        if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
+                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
+                            goto _storeSequence;
+                        }
+
+                        best_off = i - (opt[cur].mlen != 1);
+                        if (mlen > best_mlen) best_mlen = mlen;
+
+                        do {
+                            if (opt[cur].mlen == 1) {
+                                litlen = opt[cur].litlen;
+                                if (cur > litlen) {
+                                    price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
+                                } else
+                                    price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
+                            } else {
+                                litlen = 0;
+                                price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
+                            }
+
+                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
+                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
+                            mlen--;
+                        } while (mlen >= minMatch);
+            }   }   }
+
+            match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, inr, iend, maxSearches, mls, matches, minMatch);
+
+            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
+                best_mlen = matches[match_num-1].len;
+                best_off = matches[match_num-1].off;
+                last_pos = cur + 1;
+                goto _storeSequence;
+            }
+
+            /* set prices using matches at position = cur */
+            for (u = 0; u < match_num; u++) {
+                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
+                best_mlen = matches[u].len;
+
+                while (mlen <= best_mlen) {
+                    if (opt[cur].mlen == 1) {
+                        litlen = opt[cur].litlen;
+                        if (cur > litlen)
+                            price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
+                        else
+                            price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
+                    } else {
+                        litlen = 0;
+                        price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
+                    }
+
+                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
+                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
+
+                    mlen++;
+        }   }   }   /* for (cur = 1; cur <= last_pos; cur++) */
+
+        best_mlen = opt[last_pos].mlen;
+        best_off = opt[last_pos].off;
+        cur = last_pos - best_mlen;
+
+        /* store sequence */
+_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
+        opt[0].mlen = 1;
+
+        while (1) {
+            mlen = opt[cur].mlen;
+            offset = opt[cur].off;
+            opt[cur].mlen = best_mlen;
+            opt[cur].off = best_off;
+            best_mlen = mlen;
+            best_off = offset;
+            if (mlen > cur) break;
+            cur -= mlen;
+        }
+
+        for (u = 0; u <= last_pos; ) {
+            u += opt[u].mlen;
+        }
+
+        for (cur=0; cur < last_pos; ) {
+            mlen = opt[cur].mlen;
+            if (mlen == 1) { ip++; cur++; continue; }
+            offset = opt[cur].off;
+            cur += mlen;
+            litLength = (U32)(ip - anchor);
+
+            if (offset > ZSTD_REP_MOVE_OPT) {
+                rep[2] = rep[1];
+                rep[1] = rep[0];
+                rep[0] = offset - ZSTD_REP_MOVE_OPT;
+                offset--;
+            } else {
+                if (offset != 0) {
+                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
+                    if (offset != 1) rep[2] = rep[1];
+                    rep[1] = rep[0];
+                    rep[0] = best_off;
+                }
+
+                if (litLength==0) offset--;
+            }
+
+            ZSTD_updatePrice(optStatePtr, litLength, anchor, offset, mlen-MINMATCH);
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
+            anchor = ip = ip + mlen;
+    }    }   /* for (cur=0; cur < last_pos; ) */
+
+    /* Save reps for next block */
+    { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->repToConfirm[i] = rep[i]; }
+
+    /* Return the last literals size */
+    return iend - anchor;
+}
+
+
+size_t ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 0);
+}
+
+size_t ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 1);
+}
diff --git a/contrib/zstd/zstd_opt.h b/contrib/zstd/zstd_opt.h
index 543761191..816a1fabb 100644
--- a/contrib/zstd/zstd_opt.h
+++ b/contrib/zstd/zstd_opt.h
@@ -1,921 +1,30 @@
-/**
- * Copyright (c) 2016-present, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
- * This source code is licensed under the BSD-style license found in the
- * LICENSE file in the root directory of this source tree. An additional grant
- * of patent rights can be found in the PATENTS file in the same directory.
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
  */
 
+#ifndef ZSTD_OPT_H
+#define ZSTD_OPT_H
 
-/* Note : this file is intended to be included within zstd_compress.c */
+#include "zstd_compress.h"
 
+#if defined (__cplusplus)
+extern "C" {
+#endif
 
-#ifndef ZSTD_OPT_H_91842398743
-#define ZSTD_OPT_H_91842398743
+size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
 
+size_t ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
 
-#define ZSTD_LITFREQ_ADD    2
-#define ZSTD_FREQ_DIV       4
-#define ZSTD_MAX_PRICE      (1<<30)
-
-/*-*************************************
-*  Price functions for optimal parser
-***************************************/
-FORCE_INLINE void ZSTD_setLog2Prices(seqStore_t* ssPtr)
-{
-    ssPtr->log2matchLengthSum = ZSTD_highbit32(ssPtr->matchLengthSum+1);
-    ssPtr->log2litLengthSum = ZSTD_highbit32(ssPtr->litLengthSum+1);
-    ssPtr->log2litSum = ZSTD_highbit32(ssPtr->litSum+1);
-    ssPtr->log2offCodeSum = ZSTD_highbit32(ssPtr->offCodeSum+1);
-    ssPtr->factor = 1 + ((ssPtr->litSum>>5) / ssPtr->litLengthSum) + ((ssPtr->litSum<<1) / (ssPtr->litSum + ssPtr->matchSum));
-}
-
-
-MEM_STATIC void ZSTD_rescaleFreqs(seqStore_t* ssPtr, const BYTE* src, size_t srcSize)
-{
-    unsigned u;
-
-    ssPtr->cachedLiterals = NULL;
-    ssPtr->cachedPrice = ssPtr->cachedLitLength = 0;
-    ssPtr->staticPrices = 0;
-
-    if (ssPtr->litLengthSum == 0) {
-        if (srcSize <= 1024) ssPtr->staticPrices = 1;
-
-        for (u=0; u<=MaxLit; u++)
-            ssPtr->litFreq[u] = 0;
-        for (u=0; u<srcSize; u++)
-            ssPtr->litFreq[src[u]]++;
-
-        ssPtr->litSum = 0;
-        ssPtr->litLengthSum = MaxLL+1;
-        ssPtr->matchLengthSum = MaxML+1;
-        ssPtr->offCodeSum = (MaxOff+1);
-        ssPtr->matchSum = (ZSTD_LITFREQ_ADD<<Litbits);
-
-        for (u=0; u<=MaxLit; u++) {
-            ssPtr->litFreq[u] = 1 + (ssPtr->litFreq[u]>>ZSTD_FREQ_DIV);
-            ssPtr->litSum += ssPtr->litFreq[u];
-        }
-        for (u=0; u<=MaxLL; u++)
-            ssPtr->litLengthFreq[u] = 1;
-        for (u=0; u<=MaxML; u++)
-            ssPtr->matchLengthFreq[u] = 1;
-        for (u=0; u<=MaxOff; u++)
-            ssPtr->offCodeFreq[u] = 1;
-    } else {
-        ssPtr->matchLengthSum = 0;
-        ssPtr->litLengthSum = 0;
-        ssPtr->offCodeSum = 0;
-        ssPtr->matchSum = 0;
-        ssPtr->litSum = 0;
-
-        for (u=0; u<=MaxLit; u++) {
-            ssPtr->litFreq[u] = 1 + (ssPtr->litFreq[u]>>(ZSTD_FREQ_DIV+1));
-            ssPtr->litSum += ssPtr->litFreq[u];
-        }
-        for (u=0; u<=MaxLL; u++) {
-            ssPtr->litLengthFreq[u] = 1 + (ssPtr->litLengthFreq[u]>>(ZSTD_FREQ_DIV+1));
-            ssPtr->litLengthSum += ssPtr->litLengthFreq[u];
-        }
-        for (u=0; u<=MaxML; u++) {
-            ssPtr->matchLengthFreq[u] = 1 + (ssPtr->matchLengthFreq[u]>>ZSTD_FREQ_DIV);
-            ssPtr->matchLengthSum += ssPtr->matchLengthFreq[u];
-            ssPtr->matchSum += ssPtr->matchLengthFreq[u] * (u + 3);
-        }
-        ssPtr->matchSum *= ZSTD_LITFREQ_ADD;
-        for (u=0; u<=MaxOff; u++) {
-            ssPtr->offCodeFreq[u] = 1 + (ssPtr->offCodeFreq[u]>>ZSTD_FREQ_DIV);
-            ssPtr->offCodeSum += ssPtr->offCodeFreq[u];
-        }
-    }
-
-    ZSTD_setLog2Prices(ssPtr);
-}
-
-
-FORCE_INLINE U32 ZSTD_getLiteralPrice(seqStore_t* ssPtr, U32 litLength, const BYTE* literals)
-{
-    U32 price, u;
-
-    if (ssPtr->staticPrices)
-        return ZSTD_highbit32((U32)litLength+1) + (litLength*6);
-
-    if (litLength == 0)
-        return ssPtr->log2litLengthSum - ZSTD_highbit32(ssPtr->litLengthFreq[0]+1);
-
-    /* literals */
-    if (ssPtr->cachedLiterals == literals) {
-        U32 const additional = litLength - ssPtr->cachedLitLength;
-        const BYTE* literals2 = ssPtr->cachedLiterals + ssPtr->cachedLitLength;
-        price = ssPtr->cachedPrice + additional * ssPtr->log2litSum;
-        for (u=0; u < additional; u++)
-            price -= ZSTD_highbit32(ssPtr->litFreq[literals2[u]]+1);
-        ssPtr->cachedPrice = price;
-        ssPtr->cachedLitLength = litLength;
-    } else {
-        price = litLength * ssPtr->log2litSum;
-        for (u=0; u < litLength; u++)
-            price -= ZSTD_highbit32(ssPtr->litFreq[literals[u]]+1);
-
-        if (litLength >= 12) {
-            ssPtr->cachedLiterals = literals;
-            ssPtr->cachedPrice = price;
-            ssPtr->cachedLitLength = litLength;
-        }
-    }
-
-    /* literal Length */
-    {   const BYTE LL_deltaCode = 19;
-        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
-        price += LL_bits[llCode] + ssPtr->log2litLengthSum - ZSTD_highbit32(ssPtr->litLengthFreq[llCode]+1);
-    }
-
-    return price;
-}
-
-
-FORCE_INLINE U32 ZSTD_getPrice(seqStore_t* seqStorePtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength, const int ultra)
-{
-    /* offset */
-    U32 price;
-    BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
-
-    if (seqStorePtr->staticPrices)
-        return ZSTD_getLiteralPrice(seqStorePtr, litLength, literals) + ZSTD_highbit32((U32)matchLength+1) + 16 + offCode;
-
-    price = offCode + seqStorePtr->log2offCodeSum - ZSTD_highbit32(seqStorePtr->offCodeFreq[offCode]+1);
-    if (!ultra && offCode >= 20) price += (offCode-19)*2;
-
-    /* match Length */
-    {   const BYTE ML_deltaCode = 36;
-        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
-        price += ML_bits[mlCode] + seqStorePtr->log2matchLengthSum - ZSTD_highbit32(seqStorePtr->matchLengthFreq[mlCode]+1);
-    }
-
-    return price + ZSTD_getLiteralPrice(seqStorePtr, litLength, literals) + seqStorePtr->factor;
-}
-
-
-MEM_STATIC void ZSTD_updatePrice(seqStore_t* seqStorePtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength)
-{
-    U32 u;
-
-    /* literals */
-    seqStorePtr->litSum += litLength*ZSTD_LITFREQ_ADD;
-    for (u=0; u < litLength; u++)
-        seqStorePtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
-
-    /* literal Length */
-    {   const BYTE LL_deltaCode = 19;
-        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
-        seqStorePtr->litLengthFreq[llCode]++;
-        seqStorePtr->litLengthSum++;
-    }
-
-    /* match offset */
-    {   BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
-        seqStorePtr->offCodeSum++;
-        seqStorePtr->offCodeFreq[offCode]++;
-    }
-
-    /* match Length */
-    {   const BYTE ML_deltaCode = 36;
-        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
-        seqStorePtr->matchLengthFreq[mlCode]++;
-        seqStorePtr->matchLengthSum++;
-    }
-
-    ZSTD_setLog2Prices(seqStorePtr);
-}
-
-
-#define SET_PRICE(pos, mlen_, offset_, litlen_, price_)   \
-    {                                                 \
-        while (last_pos < pos)  { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; } \
-        opt[pos].mlen = mlen_;                         \
-        opt[pos].off = offset_;                        \
-        opt[pos].litlen = litlen_;                     \
-        opt[pos].price = price_;                       \
-    }
-
-
-
-/* Update hashTable3 up to ip (excluded)
-   Assumption : always within prefix (i.e. not within extDict) */
-FORCE_INLINE
-U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* zc, const BYTE* ip)
-{
-    U32* const hashTable3  = zc->hashTable3;
-    U32 const hashLog3  = zc->hashLog3;
-    const BYTE* const base = zc->base;
-    U32 idx = zc->nextToUpdate3;
-    const U32 target = zc->nextToUpdate3 = (U32)(ip - base);
-    const size_t hash3 = ZSTD_hash3Ptr(ip, hashLog3);
-
-    while(idx < target) {
-        hashTable3[ZSTD_hash3Ptr(base+idx, hashLog3)] = idx;
-        idx++;
-    }
-
-    return hashTable3[hash3];
-}
-
-
-/*-*************************************
-*  Binary Tree search
-***************************************/
-static U32 ZSTD_insertBtAndGetAllMatches (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        U32 nbCompares, const U32 mls,
-                        U32 extDict, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    const BYTE* const base = zc->base;
-    const U32 current = (U32)(ip-base);
-    const U32 hashLog = zc->params.cParams.hashLog;
-    const size_t h  = ZSTD_hashPtr(ip, hashLog, mls);
-    U32* const hashTable = zc->hashTable;
-    U32 matchIndex  = hashTable[h];
-    U32* const bt   = zc->chainTable;
-    const U32 btLog = zc->params.cParams.chainLog - 1;
-    const U32 btMask= (1U << btLog) - 1;
-    size_t commonLengthSmaller=0, commonLengthLarger=0;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const U32 btLow = btMask >= current ? 0 : current - btMask;
-    const U32 windowLow = zc->lowLimit;
-    U32* smallerPtr = bt + 2*(current&btMask);
-    U32* largerPtr  = bt + 2*(current&btMask) + 1;
-    U32 matchEndIdx = current+8;
-    U32 dummy32;   /* to be nullified at the end */
-    U32 mnum = 0;
-
-    const U32 minMatch = (mls == 3) ? 3 : 4;
-    size_t bestLength = minMatchLen-1;
-
-    if (minMatch == 3) { /* HC3 match finder */
-        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3 (zc, ip);
-        if (matchIndex3>windowLow && (current - matchIndex3 < (1<<18))) {
-            const BYTE* match;
-            size_t currentMl=0;
-            if ((!extDict) || matchIndex3 >= dictLimit) {
-                match = base + matchIndex3;
-                if (match[bestLength] == ip[bestLength]) currentMl = ZSTD_count(ip, match, iLimit);
-            } else {
-                match = dictBase + matchIndex3;
-                if (MEM_readMINMATCH(match, MINMATCH) == MEM_readMINMATCH(ip, MINMATCH))    /* assumption : matchIndex3 <= dictLimit-4 (by table construction) */
-                    currentMl = ZSTD_count_2segments(ip+MINMATCH, match+MINMATCH, iLimit, dictEnd, prefixStart) + MINMATCH;
-            }
-
-            /* save best solution */
-            if (currentMl > bestLength) {
-                bestLength = currentMl;
-                matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex3;
-                matches[mnum].len = (U32)currentMl;
-                mnum++;
-                if (currentMl > ZSTD_OPT_NUM) goto update;
-                if (ip+currentMl == iLimit) goto update; /* best possible, and avoid read overflow*/
-            }
-        }
-    }
-
-    hashTable[h] = current;   /* Update Hash Table */
-
-    while (nbCompares-- && (matchIndex > windowLow)) {
-        U32* nextPtr = bt + 2*(matchIndex & btMask);
-        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
-        const BYTE* match;
-
-        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
-            match = base + matchIndex;
-            if (match[matchLength] == ip[matchLength]) {
-                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iLimit) +1;
-            }
-        } else {
-            match = dictBase + matchIndex;
-            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart);
-            if (matchIndex+matchLength >= dictLimit)
-                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
-        }
-
-        if (matchLength > bestLength) {
-            if (matchLength > matchEndIdx - matchIndex) matchEndIdx = matchIndex + (U32)matchLength;
-            bestLength = matchLength;
-            matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex;
-            matches[mnum].len = (U32)matchLength;
-            mnum++;
-            if (matchLength > ZSTD_OPT_NUM) break;
-            if (ip+matchLength == iLimit)   /* equal : no way to know if inf or sup */
-                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
-        }
-
-        if (match[matchLength] < ip[matchLength]) {
-            /* match is smaller than current */
-            *smallerPtr = matchIndex;             /* update smaller idx */
-            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-        } else {
-            /* match is larger than current */
-            *largerPtr = matchIndex;
-            commonLengthLarger = matchLength;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-    }   }
-
-    *smallerPtr = *largerPtr = 0;
-
-update:
-    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
-    return mnum;
-}
-
-
-/** Tree updater, providing best match */
-static U32 ZSTD_BtGetAllMatches (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 0, matches, minMatchLen);
-}
-
-
-static U32 ZSTD_BtGetAllMatches_selectMLS (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* ip, const BYTE* const iHighLimit,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    switch(matchLengthSearch)
-    {
-    case 3 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
-    default :
-    case 4 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
-    case 5 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
-    case 7 :
-    case 6 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
-    }
-}
-
-/** Tree updater, providing best match */
-static U32 ZSTD_BtGetAllMatches_extDict (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 1, matches, minMatchLen);
-}
-
-
-static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* ip, const BYTE* const iHighLimit,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    switch(matchLengthSearch)
-    {
-    case 3 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
-    default :
-    case 4 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
-    case 5 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
-    case 7 :
-    case 6 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
-    }
-}
-
-
-/*-*******************************
-*  Optimal parser
-*********************************/
-FORCE_INLINE
-void ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
-                                    const void* src, size_t srcSize, const int ultra)
-{
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base;
-    const BYTE* const prefixStart = base + ctx->dictLimit;
-
-    const U32 maxSearches = 1U << ctx->params.cParams.searchLog;
-    const U32 sufficient_len = ctx->params.cParams.targetLength;
-    const U32 mls = ctx->params.cParams.searchLength;
-    const U32 minMatch = (ctx->params.cParams.searchLength == 3) ? 3 : 4;
-
-    ZSTD_optimal_t* opt = seqStorePtr->priceTable;
-    ZSTD_match_t* matches = seqStorePtr->matchTable;
-    const BYTE* inr;
-    U32 offset, rep[ZSTD_REP_NUM];
-
-    /* init */
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
-    ZSTD_rescaleFreqs(seqStorePtr, (const BYTE*)src, srcSize);
-    ip += (ip==prefixStart);
-    { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=ctx->rep[i]; }
-
-    /* Match Loop */
-    while (ip < ilimit) {
-        U32 cur, match_num, last_pos, litlen, price;
-        U32 u, mlen, best_mlen, best_off, litLength;
-        memset(opt, 0, sizeof(ZSTD_optimal_t));
-        last_pos = 0;
-        litlen = (U32)(ip - anchor);
-
-        /* check repCode */
-        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
-            for (i=(ip == anchor); i<last_i; i++) {
-                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
-                if ( (repCur > 0) && (repCur < (S32)(ip-prefixStart))
-                    && (MEM_readMINMATCH(ip, minMatch) == MEM_readMINMATCH(ip - repCur, minMatch))) {
-                    mlen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repCur, iend) + minMatch;
-                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
-                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
-                        goto _storeSequence;
-                    }
-                    best_off = i - (ip == anchor);
-                    do {
-                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                        if (mlen > last_pos || price < opt[mlen].price)
-                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
-                        mlen--;
-                    } while (mlen >= minMatch);
-        }   }   }
-
-        match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, ip, iend, maxSearches, mls, matches, minMatch);
-
-        if (!last_pos && !match_num) { ip++; continue; }
-
-        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-            best_mlen = matches[match_num-1].len;
-            best_off = matches[match_num-1].off;
-            cur = 0;
-            last_pos = 1;
-            goto _storeSequence;
-        }
-
-        /* set prices using matches at position = 0 */
-        best_mlen = (last_pos) ? last_pos : minMatch;
-        for (u = 0; u < match_num; u++) {
-            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-            best_mlen = matches[u].len;
-            while (mlen <= best_mlen) {
-                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                if (mlen > last_pos || price < opt[mlen].price)
-                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);   /* note : macro modifies last_pos */
-                mlen++;
-        }   }
-
-        if (last_pos < minMatch) { ip++; continue; }
-
-        /* initialize opt[0] */
-        { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
-        opt[0].mlen = 1;
-        opt[0].litlen = litlen;
-
-         /* check further positions */
-        for (cur = 1; cur <= last_pos; cur++) {
-           inr = ip + cur;
-
-           if (opt[cur-1].mlen == 1) {
-                litlen = opt[cur-1].litlen + 1;
-                if (cur > litlen) {
-                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-litlen);
-                } else
-                    price = ZSTD_getLiteralPrice(seqStorePtr, litlen, anchor);
-           } else {
-                litlen = 1;
-                price = opt[cur - 1].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-1);
-           }
-
-           if (cur > last_pos || price <= opt[cur].price)
-                SET_PRICE(cur, 1, 0, litlen, price);
-
-           if (cur == last_pos) break;
-
-           if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
-               continue;
-
-           mlen = opt[cur].mlen;
-           if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
-                opt[cur].rep[2] = opt[cur-mlen].rep[1];
-                opt[cur].rep[1] = opt[cur-mlen].rep[0];
-                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
-           } else {
-                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
-                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
-                opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
-           }
-
-            best_mlen = minMatch;
-            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
-                for (i=(opt[cur].mlen != 1); i<last_i; i++) {  /* check rep */
-                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
-                    if ( (repCur > 0) && (repCur < (S32)(inr-prefixStart))
-                       && (MEM_readMINMATCH(inr, minMatch) == MEM_readMINMATCH(inr - repCur, minMatch))) {
-                       mlen = (U32)ZSTD_count(inr+minMatch, inr+minMatch - repCur, iend) + minMatch;
-
-                       if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
-                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
-                            goto _storeSequence;
-                       }
-
-                       best_off = i - (opt[cur].mlen != 1);
-                       if (mlen > best_mlen) best_mlen = mlen;
-
-                       do {
-                           if (opt[cur].mlen == 1) {
-                                litlen = opt[cur].litlen;
-                                if (cur > litlen) {
-                                    price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
-                                } else
-                                    price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                            } else {
-                                litlen = 0;
-                                price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
-                            }
-
-                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
-                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
-                            mlen--;
-                        } while (mlen >= minMatch);
-            }   }   }
-
-            match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, inr, iend, maxSearches, mls, matches, best_mlen);
-
-            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-                best_mlen = matches[match_num-1].len;
-                best_off = matches[match_num-1].off;
-                last_pos = cur + 1;
-                goto _storeSequence;
-            }
-
-            /* set prices using matches at position = cur */
-            for (u = 0; u < match_num; u++) {
-                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-                best_mlen = matches[u].len;
-
-                while (mlen <= best_mlen) {
-                    if (opt[cur].mlen == 1) {
-                        litlen = opt[cur].litlen;
-                        if (cur > litlen)
-                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
-                        else
-                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                    } else {
-                        litlen = 0;
-                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
-                    }
-
-                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
-                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
-
-                    mlen++;
-        }   }   }
-
-        best_mlen = opt[last_pos].mlen;
-        best_off = opt[last_pos].off;
-        cur = last_pos - best_mlen;
-
-        /* store sequence */
-_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
-        opt[0].mlen = 1;
-
-        while (1) {
-            mlen = opt[cur].mlen;
-            offset = opt[cur].off;
-            opt[cur].mlen = best_mlen;
-            opt[cur].off = best_off;
-            best_mlen = mlen;
-            best_off = offset;
-            if (mlen > cur) break;
-            cur -= mlen;
-        }
-
-        for (u = 0; u <= last_pos;) {
-            u += opt[u].mlen;
-        }
-
-        for (cur=0; cur < last_pos; ) {
-            mlen = opt[cur].mlen;
-            if (mlen == 1) { ip++; cur++; continue; }
-            offset = opt[cur].off;
-            cur += mlen;
-            litLength = (U32)(ip - anchor);
-
-            if (offset > ZSTD_REP_MOVE_OPT) {
-                rep[2] = rep[1];
-                rep[1] = rep[0];
-                rep[0] = offset - ZSTD_REP_MOVE_OPT;
-                offset--;
-            } else {
-                if (offset != 0) {
-                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
-                    if (offset != 1) rep[2] = rep[1];
-                    rep[1] = rep[0];
-                    rep[0] = best_off;
-                }
-                if (litLength==0) offset--;
-            }
-
-            ZSTD_updatePrice(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
-            anchor = ip = ip + mlen;
-    }    }   /* for (cur=0; cur < last_pos; ) */
-
-    /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) ctx->repToConfirm[i] = rep[i]; }
-
-    /* Last Literals */
-    {   size_t const lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
-}
-
-
-FORCE_INLINE
-void ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
-                                     const void* src, size_t srcSize, const int ultra)
-{
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base;
-    const U32 lowestIndex = ctx->lowLimit;
-    const U32 dictLimit = ctx->dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const dictEnd  = dictBase + dictLimit;
-
-    const U32 maxSearches = 1U << ctx->params.cParams.searchLog;
-    const U32 sufficient_len = ctx->params.cParams.targetLength;
-    const U32 mls = ctx->params.cParams.searchLength;
-    const U32 minMatch = (ctx->params.cParams.searchLength == 3) ? 3 : 4;
-
-    ZSTD_optimal_t* opt = seqStorePtr->priceTable;
-    ZSTD_match_t* matches = seqStorePtr->matchTable;
-    const BYTE* inr;
-
-    /* init */
-    U32 offset, rep[ZSTD_REP_NUM];
-    { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=ctx->rep[i]; }
-
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
-    ZSTD_rescaleFreqs(seqStorePtr, (const BYTE*)src, srcSize);
-    ip += (ip==prefixStart);
-
-    /* Match Loop */
-    while (ip < ilimit) {
-        U32 cur, match_num, last_pos, litlen, price;
-        U32 u, mlen, best_mlen, best_off, litLength;
-        U32 current = (U32)(ip-base);
-        memset(opt, 0, sizeof(ZSTD_optimal_t));
-        last_pos = 0;
-        opt[0].litlen = (U32)(ip - anchor);
-
-        /* check repCode */
-        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
-            for (i = (ip==anchor); i<last_i; i++) {
-                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
-                const U32 repIndex = (U32)(current - repCur);
-                const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-                const BYTE* const repMatch = repBase + repIndex;
-                if ( (repCur > 0 && repCur <= (S32)current)
-                   && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
-                   && (MEM_readMINMATCH(ip, minMatch) == MEM_readMINMATCH(repMatch, minMatch)) ) {
-                    /* repcode detected we should take it */
-                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                    mlen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
-
-                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
-                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
-                        goto _storeSequence;
-                    }
-
-                    best_off = i - (ip==anchor);
-                    litlen = opt[0].litlen;
-                    do {
-                        price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                        if (mlen > last_pos || price < opt[mlen].price)
-                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
-                        mlen--;
-                    } while (mlen >= minMatch);
-        }   }   }
-
-        match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, ip, iend, maxSearches, mls, matches, minMatch);  /* first search (depth 0) */
-
-        if (!last_pos && !match_num) { ip++; continue; }
-
-        { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
-        opt[0].mlen = 1;
-
-        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-            best_mlen = matches[match_num-1].len;
-            best_off = matches[match_num-1].off;
-            cur = 0;
-            last_pos = 1;
-            goto _storeSequence;
-        }
-
-        best_mlen = (last_pos) ? last_pos : minMatch;
-
-        /* set prices using matches at position = 0 */
-        for (u = 0; u < match_num; u++) {
-            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-            best_mlen = matches[u].len;
-            litlen = opt[0].litlen;
-            while (mlen <= best_mlen) {
-                price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                if (mlen > last_pos || price < opt[mlen].price)
-                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);
-                mlen++;
-        }   }
-
-        if (last_pos < minMatch) {
-            ip++; continue;
-        }
-
-        /* check further positions */
-        for (cur = 1; cur <= last_pos; cur++) {
-            inr = ip + cur;
-
-            if (opt[cur-1].mlen == 1) {
-                litlen = opt[cur-1].litlen + 1;
-                if (cur > litlen) {
-                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-litlen);
-                } else
-                    price = ZSTD_getLiteralPrice(seqStorePtr, litlen, anchor);
-            } else {
-                litlen = 1;
-                price = opt[cur - 1].price + ZSTD_getLiteralPrice(seqStorePtr, litlen, inr-1);
-            }
-
-            if (cur > last_pos || price <= opt[cur].price)
-                SET_PRICE(cur, 1, 0, litlen, price);
-
-            if (cur == last_pos) break;
-
-            if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
-                continue;
-
-            mlen = opt[cur].mlen;
-            if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
-                opt[cur].rep[2] = opt[cur-mlen].rep[1];
-                opt[cur].rep[1] = opt[cur-mlen].rep[0];
-                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
-            } else {
-                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
-                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
-                opt[cur].rep[0] = ((opt[cur].off==ZSTD_REP_MOVE_OPT) && (mlen != 1)) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
-            }
-
-            best_mlen = minMatch;
-            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
-                for (i = (mlen != 1); i<last_i; i++) {
-                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
-                    const U32 repIndex = (U32)(current+cur - repCur);
-                    const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-                    const BYTE* const repMatch = repBase + repIndex;
-                    if ( (repCur > 0 && repCur <= (S32)(current+cur))
-                      && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
-                      && (MEM_readMINMATCH(inr, minMatch) == MEM_readMINMATCH(repMatch, minMatch)) ) {
-                        /* repcode detected */
-                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                        mlen = (U32)ZSTD_count_2segments(inr+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
-
-                        if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
-                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
-                            goto _storeSequence;
-                        }
-
-                        best_off = i - (opt[cur].mlen != 1);
-                        if (mlen > best_mlen) best_mlen = mlen;
-
-                        do {
-                            if (opt[cur].mlen == 1) {
-                                litlen = opt[cur].litlen;
-                                if (cur > litlen) {
-                                    price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
-                                } else
-                                    price = ZSTD_getPrice(seqStorePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                            } else {
-                                litlen = 0;
-                                price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
-                            }
-
-                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
-                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
-                            mlen--;
-                        } while (mlen >= minMatch);
-            }   }   }
-
-            match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, inr, iend, maxSearches, mls, matches, minMatch);
-
-            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-                best_mlen = matches[match_num-1].len;
-                best_off = matches[match_num-1].off;
-                last_pos = cur + 1;
-                goto _storeSequence;
-            }
-
-            /* set prices using matches at position = cur */
-            for (u = 0; u < match_num; u++) {
-                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-                best_mlen = matches[u].len;
-
-                while (mlen <= best_mlen) {
-                    if (opt[cur].mlen == 1) {
-                        litlen = opt[cur].litlen;
-                        if (cur > litlen)
-                            price = opt[cur - litlen].price + ZSTD_getPrice(seqStorePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
-                        else
-                            price = ZSTD_getPrice(seqStorePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                    } else {
-                        litlen = 0;
-                        price = opt[cur].price + ZSTD_getPrice(seqStorePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
-                    }
-
-                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
-                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
-
-                    mlen++;
-        }   }   }   /* for (cur = 1; cur <= last_pos; cur++) */
-
-        best_mlen = opt[last_pos].mlen;
-        best_off = opt[last_pos].off;
-        cur = last_pos - best_mlen;
-
-        /* store sequence */
-_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
-        opt[0].mlen = 1;
-
-        while (1) {
-            mlen = opt[cur].mlen;
-            offset = opt[cur].off;
-            opt[cur].mlen = best_mlen;
-            opt[cur].off = best_off;
-            best_mlen = mlen;
-            best_off = offset;
-            if (mlen > cur) break;
-            cur -= mlen;
-        }
-
-        for (u = 0; u <= last_pos; ) {
-            u += opt[u].mlen;
-        }
-
-        for (cur=0; cur < last_pos; ) {
-            mlen = opt[cur].mlen;
-            if (mlen == 1) { ip++; cur++; continue; }
-            offset = opt[cur].off;
-            cur += mlen;
-            litLength = (U32)(ip - anchor);
-
-            if (offset > ZSTD_REP_MOVE_OPT) {
-                rep[2] = rep[1];
-                rep[1] = rep[0];
-                rep[0] = offset - ZSTD_REP_MOVE_OPT;
-                offset--;
-            } else {
-                if (offset != 0) {
-                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
-                    if (offset != 1) rep[2] = rep[1];
-                    rep[1] = rep[0];
-                    rep[0] = best_off;
-                }
-
-                if (litLength==0) offset--;
-            }
-
-            ZSTD_updatePrice(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
-            anchor = ip = ip + mlen;
-    }    }   /* for (cur=0; cur < last_pos; ) */
-
-    /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) ctx->repToConfirm[i] = rep[i]; }
-
-    /* Last Literals */
-    {   size_t lastLLSize = iend - anchor;
-        memcpy(seqStorePtr->lit, anchor, lastLLSize);
-        seqStorePtr->lit += lastLLSize;
-    }
+#if defined (__cplusplus)
 }
+#endif
 
-#endif  /* ZSTD_OPT_H_91842398743 */
+#endif /* ZSTD_OPT_H */
diff --git a/contrib/zstd/zstdmt_compress.c b/contrib/zstd/zstdmt_compress.c
new file mode 100644
index 000000000..ecb799ab3
--- /dev/null
+++ b/contrib/zstd/zstdmt_compress.c
@@ -0,0 +1,1089 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+
+/* ======   Tuning parameters   ====== */
+#define ZSTDMT_NBTHREADS_MAX 256
+#define ZSTDMT_OVERLAPLOG_DEFAULT 6
+
+
+/* ======   Compiler specifics   ====== */
+#if defined(_MSC_VER)
+#  pragma warning(disable : 4204)   /* disable: C4204: non-constant aggregate initializer */
+#endif
+
+
+/* ======   Dependencies   ====== */
+#include <string.h>      /* memcpy, memset */
+#include "pool.h"        /* threadpool */
+#include "threading.h"   /* mutex */
+#include "zstd_internal.h"  /* MIN, ERROR, ZSTD_*, ZSTD_highbit32 */
+#include "zstdmt_compress.h"
+
+
+/* ======   Debug   ====== */
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
+
+#  include <stdio.h>
+#  include <unistd.h>
+#  include <sys/times.h>
+#  define DEBUGLOGRAW(l, ...) if (l<=ZSTD_DEBUG) { fprintf(stderr, __VA_ARGS__); }
+
+#  define DEBUG_PRINTHEX(l,p,n) {            \
+    unsigned debug_u;                        \
+    for (debug_u=0; debug_u<(n); debug_u++)  \
+        DEBUGLOGRAW(l, "%02X ", ((const unsigned char*)(p))[debug_u]); \
+    DEBUGLOGRAW(l, " \n");                   \
+}
+
+static unsigned long long GetCurrentClockTimeMicroseconds(void)
+{
+   static clock_t _ticksPerSecond = 0;
+   if (_ticksPerSecond <= 0) _ticksPerSecond = sysconf(_SC_CLK_TCK);
+
+   { struct tms junk; clock_t newTicks = (clock_t) times(&junk);
+     return ((((unsigned long long)newTicks)*(1000000))/_ticksPerSecond); }
+}
+
+#define MUTEX_WAIT_TIME_DLEVEL 6
+#define PTHREAD_MUTEX_LOCK(mutex) {               \
+    if (ZSTD_DEBUG>=MUTEX_WAIT_TIME_DLEVEL) {   \
+        unsigned long long const beforeTime = GetCurrentClockTimeMicroseconds(); \
+        pthread_mutex_lock(mutex);                \
+        {   unsigned long long const afterTime = GetCurrentClockTimeMicroseconds(); \
+            unsigned long long const elapsedTime = (afterTime-beforeTime); \
+            if (elapsedTime > 1000) {  /* or whatever threshold you like; I'm using 1 millisecond here */ \
+                DEBUGLOG(MUTEX_WAIT_TIME_DLEVEL, "Thread took %llu microseconds to acquire mutex %s \n", \
+                   elapsedTime, #mutex);          \
+        }   }                                     \
+    } else pthread_mutex_lock(mutex);             \
+}
+
+#else
+
+#  define PTHREAD_MUTEX_LOCK(m) pthread_mutex_lock(m)
+#  define DEBUG_PRINTHEX(l,p,n) {}
+
+#endif
+
+
+/* =====   Buffer Pool   ===== */
+/* a single Buffer Pool can be invoked from multiple threads in parallel */
+
+typedef struct buffer_s {
+    void* start;
+    size_t size;
+} buffer_t;
+
+static const buffer_t g_nullBuffer = { NULL, 0 };
+
+typedef struct ZSTDMT_bufferPool_s {
+    pthread_mutex_t poolMutex;
+    size_t bufferSize;
+    unsigned totalBuffers;
+    unsigned nbBuffers;
+    ZSTD_customMem cMem;
+    buffer_t bTable[1];   /* variable size */
+} ZSTDMT_bufferPool;
+
+static ZSTDMT_bufferPool* ZSTDMT_createBufferPool(unsigned nbThreads, ZSTD_customMem cMem)
+{
+    unsigned const maxNbBuffers = 2*nbThreads + 3;
+    ZSTDMT_bufferPool* const bufPool = (ZSTDMT_bufferPool*)ZSTD_calloc(
+        sizeof(ZSTDMT_bufferPool) + (maxNbBuffers-1) * sizeof(buffer_t), cMem);
+    if (bufPool==NULL) return NULL;
+    if (pthread_mutex_init(&bufPool->poolMutex, NULL)) {
+        ZSTD_free(bufPool, cMem);
+        return NULL;
+    }
+    bufPool->bufferSize = 64 KB;
+    bufPool->totalBuffers = maxNbBuffers;
+    bufPool->nbBuffers = 0;
+    bufPool->cMem = cMem;
+    return bufPool;
+}
+
+static void ZSTDMT_freeBufferPool(ZSTDMT_bufferPool* bufPool)
+{
+    unsigned u;
+    if (!bufPool) return;   /* compatibility with free on NULL */
+    for (u=0; u<bufPool->totalBuffers; u++)
+        ZSTD_free(bufPool->bTable[u].start, bufPool->cMem);
+    pthread_mutex_destroy(&bufPool->poolMutex);
+    ZSTD_free(bufPool, bufPool->cMem);
+}
+
+/* only works at initialization, not during compression */
+static size_t ZSTDMT_sizeof_bufferPool(ZSTDMT_bufferPool* bufPool)
+{
+    size_t const poolSize = sizeof(*bufPool)
+                            + (bufPool->totalBuffers - 1) * sizeof(buffer_t);
+    unsigned u;
+    size_t totalBufferSize = 0;
+    pthread_mutex_lock(&bufPool->poolMutex);
+    for (u=0; u<bufPool->totalBuffers; u++)
+        totalBufferSize += bufPool->bTable[u].size;
+    pthread_mutex_unlock(&bufPool->poolMutex);
+
+    return poolSize + totalBufferSize;
+}
+
+static void ZSTDMT_setBufferSize(ZSTDMT_bufferPool* bufPool, size_t bSize)
+{
+    bufPool->bufferSize = bSize;
+}
+
+/** ZSTDMT_getBuffer() :
+ *  assumption : bufPool must be valid */
+static buffer_t ZSTDMT_getBuffer(ZSTDMT_bufferPool* bufPool)
+{
+    size_t const bSize = bufPool->bufferSize;
+    DEBUGLOG(5, "ZSTDMT_getBuffer");
+    pthread_mutex_lock(&bufPool->poolMutex);
+    if (bufPool->nbBuffers) {   /* try to use an existing buffer */
+        buffer_t const buf = bufPool->bTable[--(bufPool->nbBuffers)];
+        size_t const availBufferSize = buf.size;
+        if ((availBufferSize >= bSize) & (availBufferSize <= 10*bSize)) {
+            /* large enough, but not too much */
+            pthread_mutex_unlock(&bufPool->poolMutex);
+            return buf;
+        }
+        /* size conditions not respected : scratch this buffer, create new one */
+        DEBUGLOG(5, "existing buffer does not meet size conditions => freeing");
+        ZSTD_free(buf.start, bufPool->cMem);
+    }
+    pthread_mutex_unlock(&bufPool->poolMutex);
+    /* create new buffer */
+    DEBUGLOG(5, "create a new buffer");
+    {   buffer_t buffer;
+        void* const start = ZSTD_malloc(bSize, bufPool->cMem);
+        buffer.start = start;   /* note : start can be NULL if malloc fails ! */
+        buffer.size = (start==NULL) ? 0 : bSize;
+        return buffer;
+    }
+}
+
+/* store buffer for later re-use, up to pool capacity */
+static void ZSTDMT_releaseBuffer(ZSTDMT_bufferPool* bufPool, buffer_t buf)
+{
+    if (buf.start == NULL) return;   /* compatible with release on NULL */
+    DEBUGLOG(5, "ZSTDMT_releaseBuffer");
+    pthread_mutex_lock(&bufPool->poolMutex);
+    if (bufPool->nbBuffers < bufPool->totalBuffers) {
+        bufPool->bTable[bufPool->nbBuffers++] = buf;  /* stored for later use */
+        pthread_mutex_unlock(&bufPool->poolMutex);
+        return;
+    }
+    pthread_mutex_unlock(&bufPool->poolMutex);
+    /* Reached bufferPool capacity (should not happen) */
+    DEBUGLOG(5, "buffer pool capacity reached => freeing ");
+    ZSTD_free(buf.start, bufPool->cMem);
+}
+
+/* Sets parameters relevant to the compression job, initializing others to
+ * default values. Notably, nbThreads should probably be zero. */
+static ZSTD_CCtx_params ZSTDMT_makeJobCCtxParams(ZSTD_CCtx_params const params)
+{
+    ZSTD_CCtx_params jobParams;
+    memset(&jobParams, 0, sizeof(jobParams));
+
+    jobParams.cParams = params.cParams;
+    jobParams.fParams = params.fParams;
+    jobParams.compressionLevel = params.compressionLevel;
+
+    jobParams.ldmParams = params.ldmParams;
+    return jobParams;
+}
+
+/* =====   CCtx Pool   ===== */
+/* a single CCtx Pool can be invoked from multiple threads in parallel */
+
+typedef struct {
+    pthread_mutex_t poolMutex;
+    unsigned totalCCtx;
+    unsigned availCCtx;
+    ZSTD_customMem cMem;
+    ZSTD_CCtx* cctx[1];   /* variable size */
+} ZSTDMT_CCtxPool;
+
+/* note : all CCtx borrowed from the pool should be released back to the pool _before_ freeing the pool */
+static void ZSTDMT_freeCCtxPool(ZSTDMT_CCtxPool* pool)
+{
+    unsigned u;
+    for (u=0; u<pool->totalCCtx; u++)
+        ZSTD_freeCCtx(pool->cctx[u]);  /* note : compatible with free on NULL */
+    pthread_mutex_destroy(&pool->poolMutex);
+    ZSTD_free(pool, pool->cMem);
+}
+
+/* ZSTDMT_createCCtxPool() :
+ * implies nbThreads >= 1 , checked by caller ZSTDMT_createCCtx() */
+static ZSTDMT_CCtxPool* ZSTDMT_createCCtxPool(unsigned nbThreads,
+                                              ZSTD_customMem cMem)
+{
+    ZSTDMT_CCtxPool* const cctxPool = (ZSTDMT_CCtxPool*) ZSTD_calloc(
+        sizeof(ZSTDMT_CCtxPool) + (nbThreads-1)*sizeof(ZSTD_CCtx*), cMem);
+    if (!cctxPool) return NULL;
+    if (pthread_mutex_init(&cctxPool->poolMutex, NULL)) {
+        ZSTD_free(cctxPool, cMem);
+        return NULL;
+    }
+    cctxPool->cMem = cMem;
+    cctxPool->totalCCtx = nbThreads;
+    cctxPool->availCCtx = 1;   /* at least one cctx for single-thread mode */
+    cctxPool->cctx[0] = ZSTD_createCCtx_advanced(cMem);
+    if (!cctxPool->cctx[0]) { ZSTDMT_freeCCtxPool(cctxPool); return NULL; }
+    DEBUGLOG(3, "cctxPool created, with %u threads", nbThreads);
+    return cctxPool;
+}
+
+/* only works during initialization phase, not during compression */
+static size_t ZSTDMT_sizeof_CCtxPool(ZSTDMT_CCtxPool* cctxPool)
+{
+    pthread_mutex_lock(&cctxPool->poolMutex);
+    {   unsigned const nbThreads = cctxPool->totalCCtx;
+        size_t const poolSize = sizeof(*cctxPool)
+                                + (nbThreads-1)*sizeof(ZSTD_CCtx*);
+        unsigned u;
+        size_t totalCCtxSize = 0;
+        for (u=0; u<nbThreads; u++) {
+            totalCCtxSize += ZSTD_sizeof_CCtx(cctxPool->cctx[u]);
+        }
+        pthread_mutex_unlock(&cctxPool->poolMutex);
+        return poolSize + totalCCtxSize;
+    }
+}
+
+static ZSTD_CCtx* ZSTDMT_getCCtx(ZSTDMT_CCtxPool* cctxPool)
+{
+    DEBUGLOG(5, "ZSTDMT_getCCtx");
+    pthread_mutex_lock(&cctxPool->poolMutex);
+    if (cctxPool->availCCtx) {
+        cctxPool->availCCtx--;
+        {   ZSTD_CCtx* const cctx = cctxPool->cctx[cctxPool->availCCtx];
+            pthread_mutex_unlock(&cctxPool->poolMutex);
+            return cctx;
+    }   }
+    pthread_mutex_unlock(&cctxPool->poolMutex);
+    DEBUGLOG(5, "create one more CCtx");
+    return ZSTD_createCCtx_advanced(cctxPool->cMem);   /* note : can be NULL, when creation fails ! */
+}
+
+static void ZSTDMT_releaseCCtx(ZSTDMT_CCtxPool* pool, ZSTD_CCtx* cctx)
+{
+    if (cctx==NULL) return;   /* compatibility with release on NULL */
+    pthread_mutex_lock(&pool->poolMutex);
+    if (pool->availCCtx < pool->totalCCtx)
+        pool->cctx[pool->availCCtx++] = cctx;
+    else {
+        /* pool overflow : should not happen, since totalCCtx==nbThreads */
+        DEBUGLOG(5, "CCtx pool overflow : free cctx");
+        ZSTD_freeCCtx(cctx);
+    }
+    pthread_mutex_unlock(&pool->poolMutex);
+}
+
+
+/* =====   Thread worker   ===== */
+
+typedef struct {
+    buffer_t src;
+    const void* srcStart;
+    size_t   dictSize;
+    size_t   srcSize;
+    buffer_t dstBuff;
+    size_t   cSize;
+    size_t   dstFlushed;
+    unsigned firstChunk;
+    unsigned lastChunk;
+    unsigned jobCompleted;
+    unsigned jobScanned;
+    pthread_mutex_t* jobCompleted_mutex;
+    pthread_cond_t* jobCompleted_cond;
+    ZSTD_CCtx_params params;
+    const ZSTD_CDict* cdict;
+    ZSTDMT_CCtxPool* cctxPool;
+    ZSTDMT_bufferPool* bufPool;
+    unsigned long long fullFrameSize;
+} ZSTDMT_jobDescription;
+
+/* ZSTDMT_compressChunk() : POOL_function type */
+void ZSTDMT_compressChunk(void* jobDescription)
+{
+    ZSTDMT_jobDescription* const job = (ZSTDMT_jobDescription*)jobDescription;
+    ZSTD_CCtx* cctx = ZSTDMT_getCCtx(job->cctxPool);
+    const void* const src = (const char*)job->srcStart + job->dictSize;
+    buffer_t dstBuff = job->dstBuff;
+    DEBUGLOG(5, "job (first:%u) (last:%u) : dictSize %u, srcSize %u",
+                 job->firstChunk, job->lastChunk, (U32)job->dictSize, (U32)job->srcSize);
+
+    if (cctx==NULL) {
+        job->cSize = ERROR(memory_allocation);
+        goto _endJob;
+    }
+
+    if (dstBuff.start == NULL) {
+        dstBuff = ZSTDMT_getBuffer(job->bufPool);
+        if (dstBuff.start==NULL) {
+            job->cSize = ERROR(memory_allocation);
+            goto _endJob;
+        }
+        job->dstBuff = dstBuff;
+    }
+
+    if (job->cdict) {  /* should only happen for first segment */
+        size_t const initError = ZSTD_compressBegin_usingCDict_advanced(cctx, job->cdict, job->params.fParams, job->fullFrameSize);
+        DEBUGLOG(5, "using CDict");
+        if (ZSTD_isError(initError)) { job->cSize = initError; goto _endJob; }
+    } else {  /* srcStart points at reloaded section */
+        if (!job->firstChunk) job->params.fParams.contentSizeFlag = 0;  /* ensure no srcSize control */
+        { ZSTD_CCtx_params jobParams = job->params;
+          size_t const forceWindowError =
+              ZSTD_CCtxParam_setParameter(&jobParams, ZSTD_p_forceMaxWindow, !job->firstChunk);
+          /* Force loading dictionary in "content-only" mode (no header analysis) */
+          size_t const initError = ZSTD_compressBegin_advanced_internal(cctx, job->srcStart, job->dictSize, ZSTD_dm_rawContent, jobParams, job->fullFrameSize);
+            if (ZSTD_isError(initError) || ZSTD_isError(forceWindowError)) {
+                job->cSize = initError;
+                goto _endJob;
+            }
+    }   }
+    if (!job->firstChunk) {  /* flush and overwrite frame header when it's not first segment */
+        size_t const hSize = ZSTD_compressContinue(cctx, dstBuff.start, dstBuff.size, src, 0);
+        if (ZSTD_isError(hSize)) { job->cSize = hSize; goto _endJob; }
+        ZSTD_invalidateRepCodes(cctx);
+    }
+
+    DEBUGLOG(5, "Compressing : ");
+    DEBUG_PRINTHEX(4, job->srcStart, 12);
+    job->cSize = (job->lastChunk) ?
+                 ZSTD_compressEnd     (cctx, dstBuff.start, dstBuff.size, src, job->srcSize) :
+                 ZSTD_compressContinue(cctx, dstBuff.start, dstBuff.size, src, job->srcSize);
+    DEBUGLOG(5, "compressed %u bytes into %u bytes   (first:%u) (last:%u)",
+                (unsigned)job->srcSize, (unsigned)job->cSize, job->firstChunk, job->lastChunk);
+    DEBUGLOG(5, "dstBuff.size : %u ; => %s", (U32)dstBuff.size, ZSTD_getErrorName(job->cSize));
+
+_endJob:
+    ZSTDMT_releaseCCtx(job->cctxPool, cctx);
+    ZSTDMT_releaseBuffer(job->bufPool, job->src);
+    job->src = g_nullBuffer; job->srcStart = NULL;
+    PTHREAD_MUTEX_LOCK(job->jobCompleted_mutex);
+    job->jobCompleted = 1;
+    job->jobScanned = 0;
+    pthread_cond_signal(job->jobCompleted_cond);
+    pthread_mutex_unlock(job->jobCompleted_mutex);
+}
+
+
+/* ------------------------------------------ */
+/* =====   Multi-threaded compression   ===== */
+/* ------------------------------------------ */
+
+typedef struct {
+    buffer_t buffer;
+    size_t filled;
+} inBuff_t;
+
+struct ZSTDMT_CCtx_s {
+    POOL_ctx* factory;
+    ZSTDMT_jobDescription* jobs;
+    ZSTDMT_bufferPool* bufPool;
+    ZSTDMT_CCtxPool* cctxPool;
+    pthread_mutex_t jobCompleted_mutex;
+    pthread_cond_t jobCompleted_cond;
+    size_t targetSectionSize;
+    size_t inBuffSize;
+    size_t dictSize;
+    size_t targetDictSize;
+    inBuff_t inBuff;
+    ZSTD_CCtx_params params;
+    XXH64_state_t xxhState;
+    unsigned jobIDMask;
+    unsigned doneJobID;
+    unsigned nextJobID;
+    unsigned frameEnded;
+    unsigned allJobsCompleted;
+    unsigned long long frameContentSize;
+    ZSTD_customMem cMem;
+    ZSTD_CDict* cdictLocal;
+    const ZSTD_CDict* cdict;
+};
+
+static ZSTDMT_jobDescription* ZSTDMT_allocJobsTable(U32* nbJobsPtr, ZSTD_customMem cMem)
+{
+    U32 const nbJobsLog2 = ZSTD_highbit32(*nbJobsPtr) + 1;
+    U32 const nbJobs = 1 << nbJobsLog2;
+    *nbJobsPtr = nbJobs;
+    return (ZSTDMT_jobDescription*) ZSTD_calloc(
+                            nbJobs * sizeof(ZSTDMT_jobDescription), cMem);
+}
+
+/* Internal only */
+size_t ZSTDMT_initializeCCtxParameters(ZSTD_CCtx_params* params, unsigned nbThreads)
+{
+    params->nbThreads = nbThreads;
+    params->overlapSizeLog = ZSTDMT_OVERLAPLOG_DEFAULT;
+    params->jobSize = 0;
+    return 0;
+}
+
+ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbThreads, ZSTD_customMem cMem)
+{
+    ZSTDMT_CCtx* mtctx;
+    U32 nbJobs = nbThreads + 2;
+    DEBUGLOG(3, "ZSTDMT_createCCtx_advanced");
+
+    if (nbThreads < 1) return NULL;
+    nbThreads = MIN(nbThreads , ZSTDMT_NBTHREADS_MAX);
+    if ((cMem.customAlloc!=NULL) ^ (cMem.customFree!=NULL))
+        /* invalid custom allocator */
+        return NULL;
+
+    mtctx = (ZSTDMT_CCtx*) ZSTD_calloc(sizeof(ZSTDMT_CCtx), cMem);
+    if (!mtctx) return NULL;
+    ZSTDMT_initializeCCtxParameters(&mtctx->params, nbThreads);
+    mtctx->cMem = cMem;
+    mtctx->allJobsCompleted = 1;
+    mtctx->factory = POOL_create_advanced(nbThreads, 0, cMem);
+    mtctx->jobs = ZSTDMT_allocJobsTable(&nbJobs, cMem);
+    mtctx->jobIDMask = nbJobs - 1;
+    mtctx->bufPool = ZSTDMT_createBufferPool(nbThreads, cMem);
+    mtctx->cctxPool = ZSTDMT_createCCtxPool(nbThreads, cMem);
+    if (!mtctx->factory | !mtctx->jobs | !mtctx->bufPool | !mtctx->cctxPool) {
+        ZSTDMT_freeCCtx(mtctx);
+        return NULL;
+    }
+    if (pthread_mutex_init(&mtctx->jobCompleted_mutex, NULL)) {
+        ZSTDMT_freeCCtx(mtctx);
+        return NULL;
+    }
+    if (pthread_cond_init(&mtctx->jobCompleted_cond, NULL)) {
+        ZSTDMT_freeCCtx(mtctx);
+        return NULL;
+    }
+    DEBUGLOG(3, "mt_cctx created, for %u threads", nbThreads);
+    return mtctx;
+}
+
+ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbThreads)
+{
+    return ZSTDMT_createCCtx_advanced(nbThreads, ZSTD_defaultCMem);
+}
+
+/* ZSTDMT_releaseAllJobResources() :
+ * note : ensure all workers are killed first ! */
+static void ZSTDMT_releaseAllJobResources(ZSTDMT_CCtx* mtctx)
+{
+    unsigned jobID;
+    DEBUGLOG(3, "ZSTDMT_releaseAllJobResources");
+    for (jobID=0; jobID <= mtctx->jobIDMask; jobID++) {
+        ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[jobID].dstBuff);
+        mtctx->jobs[jobID].dstBuff = g_nullBuffer;
+        ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[jobID].src);
+        mtctx->jobs[jobID].src = g_nullBuffer;
+    }
+    memset(mtctx->jobs, 0, (mtctx->jobIDMask+1)*sizeof(ZSTDMT_jobDescription));
+    ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->inBuff.buffer);
+    mtctx->inBuff.buffer = g_nullBuffer;
+    mtctx->allJobsCompleted = 1;
+}
+
+size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx)
+{
+    if (mtctx==NULL) return 0;   /* compatible with free on NULL */
+    POOL_free(mtctx->factory);
+    if (!mtctx->allJobsCompleted) ZSTDMT_releaseAllJobResources(mtctx); /* stop workers first */
+    ZSTDMT_freeBufferPool(mtctx->bufPool);  /* release job resources into pools first */
+    ZSTD_free(mtctx->jobs, mtctx->cMem);
+    ZSTDMT_freeCCtxPool(mtctx->cctxPool);
+    ZSTD_freeCDict(mtctx->cdictLocal);
+    pthread_mutex_destroy(&mtctx->jobCompleted_mutex);
+    pthread_cond_destroy(&mtctx->jobCompleted_cond);
+    ZSTD_free(mtctx, mtctx->cMem);
+    return 0;
+}
+
+size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx)
+{
+    if (mtctx == NULL) return 0;   /* supports sizeof NULL */
+    return sizeof(*mtctx)
+            + POOL_sizeof(mtctx->factory)
+            + ZSTDMT_sizeof_bufferPool(mtctx->bufPool)
+            + (mtctx->jobIDMask+1) * sizeof(ZSTDMT_jobDescription)
+            + ZSTDMT_sizeof_CCtxPool(mtctx->cctxPool)
+            + ZSTD_sizeof_CDict(mtctx->cdictLocal);
+}
+
+/* Internal only */
+size_t ZSTDMT_CCtxParam_setMTCtxParameter(
+    ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, unsigned value) {
+    switch(parameter)
+    {
+    case ZSTDMT_p_sectionSize :
+        params->jobSize = value;
+        return 0;
+    case ZSTDMT_p_overlapSectionLog :
+        DEBUGLOG(4, "ZSTDMT_p_overlapSectionLog : %u", value);
+        params->overlapSizeLog = (value >= 9) ? 9 : value;
+        return 0;
+    default :
+        return ERROR(parameter_unsupported);
+    }
+}
+
+size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned value)
+{
+    switch(parameter)
+    {
+    case ZSTDMT_p_sectionSize :
+        return ZSTDMT_CCtxParam_setMTCtxParameter(&mtctx->params, parameter, value);
+    case ZSTDMT_p_overlapSectionLog :
+        return ZSTDMT_CCtxParam_setMTCtxParameter(&mtctx->params, parameter, value);
+    default :
+        return ERROR(parameter_unsupported);
+    }
+}
+
+/* ------------------------------------------ */
+/* =====   Multi-threaded compression   ===== */
+/* ------------------------------------------ */
+
+static unsigned computeNbChunks(size_t srcSize, unsigned windowLog, unsigned nbThreads) {
+    size_t const chunkSizeTarget = (size_t)1 << (windowLog + 2);
+    size_t const chunkMaxSize = chunkSizeTarget << 2;
+    size_t const passSizeMax = chunkMaxSize * nbThreads;
+    unsigned const multiplier = (unsigned)(srcSize / passSizeMax) + 1;
+    unsigned const nbChunksLarge = multiplier * nbThreads;
+    unsigned const nbChunksMax = (unsigned)(srcSize / chunkSizeTarget) + 1;
+    unsigned const nbChunksSmall = MIN(nbChunksMax, nbThreads);
+    return (multiplier>1) ? nbChunksLarge : nbChunksSmall;
+}
+
+static size_t ZSTDMT_compress_advanced_internal(
+                ZSTDMT_CCtx* mtctx,
+                void* dst, size_t dstCapacity,
+          const void* src, size_t srcSize,
+          const ZSTD_CDict* cdict,
+                ZSTD_CCtx_params const params)
+{
+    ZSTD_CCtx_params const jobParams = ZSTDMT_makeJobCCtxParams(params);
+    unsigned const overlapRLog = (params.overlapSizeLog>9) ? 0 : 9-params.overlapSizeLog;
+    size_t const overlapSize = (overlapRLog>=9) ? 0 : (size_t)1 << (params.cParams.windowLog - overlapRLog);
+    unsigned nbChunks = computeNbChunks(srcSize, params.cParams.windowLog, params.nbThreads);
+    size_t const proposedChunkSize = (srcSize + (nbChunks-1)) / nbChunks;
+    size_t const avgChunkSize = ((proposedChunkSize & 0x1FFFF) < 0x7FFF) ? proposedChunkSize + 0xFFFF : proposedChunkSize;   /* avoid too small last block */
+    const char* const srcStart = (const char*)src;
+    size_t remainingSrcSize = srcSize;
+    unsigned const compressWithinDst = (dstCapacity >= ZSTD_compressBound(srcSize)) ? nbChunks : (unsigned)(dstCapacity / ZSTD_compressBound(avgChunkSize));  /* presumes avgChunkSize >= 256 KB, which should be the case */
+    size_t frameStartPos = 0, dstBufferPos = 0;
+    XXH64_state_t xxh64;
+    assert(jobParams.nbThreads == 0);
+    assert(mtctx->cctxPool->totalCCtx == params.nbThreads);
+
+    DEBUGLOG(4, "nbChunks  : %2u   (chunkSize : %u bytes)   ", nbChunks, (U32)avgChunkSize);
+    if (nbChunks==1) {   /* fallback to single-thread mode */
+        ZSTD_CCtx* const cctx = mtctx->cctxPool->cctx[0];
+        if (cdict) return ZSTD_compress_usingCDict_advanced(cctx, dst, dstCapacity, src, srcSize, cdict, jobParams.fParams);
+        return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, NULL, 0, jobParams);
+    }
+    assert(avgChunkSize >= 256 KB);  /* condition for ZSTD_compressBound(A) + ZSTD_compressBound(B) <= ZSTD_compressBound(A+B), which is required for compressWithinDst */
+    ZSTDMT_setBufferSize(mtctx->bufPool, ZSTD_compressBound(avgChunkSize) );
+    XXH64_reset(&xxh64, 0);
+
+    if (nbChunks > mtctx->jobIDMask+1) {  /* enlarge job table */
+        U32 nbJobs = nbChunks;
+        ZSTD_free(mtctx->jobs, mtctx->cMem);
+        mtctx->jobIDMask = 0;
+        mtctx->jobs = ZSTDMT_allocJobsTable(&nbJobs, mtctx->cMem);
+        if (mtctx->jobs==NULL) return ERROR(memory_allocation);
+        mtctx->jobIDMask = nbJobs - 1;
+    }
+
+    {   unsigned u;
+        for (u=0; u<nbChunks; u++) {
+            size_t const chunkSize = MIN(remainingSrcSize, avgChunkSize);
+            size_t const dstBufferCapacity = ZSTD_compressBound(chunkSize);
+            buffer_t const dstAsBuffer = { (char*)dst + dstBufferPos, dstBufferCapacity };
+            buffer_t const dstBuffer = u < compressWithinDst ? dstAsBuffer : g_nullBuffer;
+            size_t dictSize = u ? overlapSize : 0;
+
+            mtctx->jobs[u].src = g_nullBuffer;
+            mtctx->jobs[u].srcStart = srcStart + frameStartPos - dictSize;
+            mtctx->jobs[u].dictSize = dictSize;
+            mtctx->jobs[u].srcSize = chunkSize;
+            mtctx->jobs[u].cdict = mtctx->nextJobID==0 ? cdict : NULL;
+            mtctx->jobs[u].fullFrameSize = srcSize;
+            mtctx->jobs[u].params = jobParams;
+            /* do not calculate checksum within sections, but write it in header for first section */
+            if (u!=0) mtctx->jobs[u].params.fParams.checksumFlag = 0;
+            mtctx->jobs[u].dstBuff = dstBuffer;
+            mtctx->jobs[u].cctxPool = mtctx->cctxPool;
+            mtctx->jobs[u].bufPool = mtctx->bufPool;
+            mtctx->jobs[u].firstChunk = (u==0);
+            mtctx->jobs[u].lastChunk = (u==nbChunks-1);
+            mtctx->jobs[u].jobCompleted = 0;
+            mtctx->jobs[u].jobCompleted_mutex = &mtctx->jobCompleted_mutex;
+            mtctx->jobs[u].jobCompleted_cond = &mtctx->jobCompleted_cond;
+
+            if (params.fParams.checksumFlag) {
+                XXH64_update(&xxh64, srcStart + frameStartPos, chunkSize);
+            }
+
+            DEBUGLOG(5, "posting job %u   (%u bytes)", u, (U32)chunkSize);
+            DEBUG_PRINTHEX(6, mtctx->jobs[u].srcStart, 12);
+            POOL_add(mtctx->factory, ZSTDMT_compressChunk, &mtctx->jobs[u]);
+
+            frameStartPos += chunkSize;
+            dstBufferPos += dstBufferCapacity;
+            remainingSrcSize -= chunkSize;
+    }   }
+
+    /* collect result */
+    {   size_t error = 0, dstPos = 0;
+        unsigned chunkID;
+        for (chunkID=0; chunkID<nbChunks; chunkID++) {
+            DEBUGLOG(5, "waiting for chunk %u ", chunkID);
+            PTHREAD_MUTEX_LOCK(&mtctx->jobCompleted_mutex);
+            while (mtctx->jobs[chunkID].jobCompleted==0) {
+                DEBUGLOG(5, "waiting for jobCompleted signal from chunk %u", chunkID);
+                pthread_cond_wait(&mtctx->jobCompleted_cond, &mtctx->jobCompleted_mutex);
+            }
+            pthread_mutex_unlock(&mtctx->jobCompleted_mutex);
+            DEBUGLOG(5, "ready to write chunk %u ", chunkID);
+
+            mtctx->jobs[chunkID].srcStart = NULL;
+            {   size_t const cSize = mtctx->jobs[chunkID].cSize;
+                if (ZSTD_isError(cSize)) error = cSize;
+                if ((!error) && (dstPos + cSize > dstCapacity)) error = ERROR(dstSize_tooSmall);
+                if (chunkID) {   /* note : chunk 0 is written directly at dst, which is correct position */
+                    if (!error)
+                        memmove((char*)dst + dstPos, mtctx->jobs[chunkID].dstBuff.start, cSize);  /* may overlap when chunk compressed within dst */
+                    if (chunkID >= compressWithinDst) {  /* chunk compressed into its own buffer, which must be released */
+                        DEBUGLOG(5, "releasing buffer %u>=%u", chunkID, compressWithinDst);
+                        ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[chunkID].dstBuff);
+                    }
+                    mtctx->jobs[chunkID].dstBuff = g_nullBuffer;
+                }
+                dstPos += cSize ;
+            }
+        }  /* for (chunkID=0; chunkID<nbChunks; chunkID++) */
+
+        DEBUGLOG(4, "checksumFlag : %u ", params.fParams.checksumFlag);
+        if (params.fParams.checksumFlag) {
+            U32 const checksum = (U32)XXH64_digest(&xxh64);
+            if (dstPos + 4 > dstCapacity) {
+                error = ERROR(dstSize_tooSmall);
+            } else {
+                DEBUGLOG(4, "writing checksum : %08X \n", checksum);
+                MEM_writeLE32((char*)dst + dstPos, checksum);
+                dstPos += 4;
+        }   }
+
+        if (!error) DEBUGLOG(4, "compressed size : %u  ", (U32)dstPos);
+        return error ? error : dstPos;
+    }
+}
+
+size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
+                               void* dst, size_t dstCapacity,
+                         const void* src, size_t srcSize,
+                         const ZSTD_CDict* cdict,
+                               ZSTD_parameters const params,
+                               unsigned overlapLog)
+{
+    ZSTD_CCtx_params cctxParams = mtctx->params;
+    cctxParams.cParams = params.cParams;
+    cctxParams.fParams = params.fParams;
+    cctxParams.overlapSizeLog = overlapLog;
+    return ZSTDMT_compress_advanced_internal(mtctx,
+                                             dst, dstCapacity,
+                                             src, srcSize,
+                                             cdict, cctxParams);
+}
+
+
+size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
+                           void* dst, size_t dstCapacity,
+                     const void* src, size_t srcSize,
+                           int compressionLevel)
+{
+    U32 const overlapLog = (compressionLevel >= ZSTD_maxCLevel()) ? 9 : ZSTDMT_OVERLAPLOG_DEFAULT;
+    ZSTD_parameters params = ZSTD_getParams(compressionLevel, srcSize, 0);
+    params.fParams.contentSizeFlag = 1;
+    return ZSTDMT_compress_advanced(mtctx, dst, dstCapacity, src, srcSize, NULL, params, overlapLog);
+}
+
+
+/* ====================================== */
+/* =======      Streaming API     ======= */
+/* ====================================== */
+
+static void ZSTDMT_waitForAllJobsCompleted(ZSTDMT_CCtx* zcs)
+{
+    DEBUGLOG(4, "ZSTDMT_waitForAllJobsCompleted");
+    while (zcs->doneJobID < zcs->nextJobID) {
+        unsigned const jobID = zcs->doneJobID & zcs->jobIDMask;
+        PTHREAD_MUTEX_LOCK(&zcs->jobCompleted_mutex);
+        while (zcs->jobs[jobID].jobCompleted==0) {
+            DEBUGLOG(5, "waiting for jobCompleted signal from chunk %u", zcs->doneJobID);   /* we want to block when waiting for data to flush */
+            pthread_cond_wait(&zcs->jobCompleted_cond, &zcs->jobCompleted_mutex);
+        }
+        pthread_mutex_unlock(&zcs->jobCompleted_mutex);
+        zcs->doneJobID++;
+    }
+}
+
+size_t ZSTDMT_initCStream_internal(
+        ZSTDMT_CCtx* zcs,
+        const void* dict, size_t dictSize, ZSTD_dictMode_e dictMode,
+        const ZSTD_CDict* cdict, ZSTD_CCtx_params params,
+        unsigned long long pledgedSrcSize)
+{
+    DEBUGLOG(4, "ZSTDMT_initCStream_internal");
+    /* params are supposed to be fully validated at this point */
+    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+    assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
+    assert(zcs->cctxPool->totalCCtx == params.nbThreads);
+
+    if (params.nbThreads==1) {
+        ZSTD_CCtx_params const singleThreadParams = ZSTDMT_makeJobCCtxParams(params);
+        DEBUGLOG(4, "single thread mode");
+        assert(singleThreadParams.nbThreads == 0);
+        return ZSTD_initCStream_internal(zcs->cctxPool->cctx[0],
+                                         dict, dictSize, cdict,
+                                         singleThreadParams, pledgedSrcSize);
+    }
+
+    if (zcs->allJobsCompleted == 0) {   /* previous compression not correctly finished */
+        ZSTDMT_waitForAllJobsCompleted(zcs);
+        ZSTDMT_releaseAllJobResources(zcs);
+        zcs->allJobsCompleted = 1;
+    }
+
+    zcs->params = params;
+    zcs->frameContentSize = pledgedSrcSize;
+    if (dict) {
+        DEBUGLOG(4,"cdictLocal: %08X", (U32)(size_t)zcs->cdictLocal);
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = ZSTD_createCDict_advanced(dict, dictSize,
+                                                    ZSTD_dlm_byCopy, dictMode, /* note : a loadPrefix becomes an internal CDict */
+                                                    params.cParams, zcs->cMem);
+        zcs->cdict = zcs->cdictLocal;
+        if (zcs->cdictLocal == NULL) return ERROR(memory_allocation);
+    } else {
+        DEBUGLOG(4,"cdictLocal: %08X", (U32)(size_t)zcs->cdictLocal);
+        ZSTD_freeCDict(zcs->cdictLocal);
+        zcs->cdictLocal = NULL;
+        zcs->cdict = cdict;
+    }
+
+    zcs->targetDictSize = (params.overlapSizeLog==0) ? 0 : (size_t)1 << (params.cParams.windowLog - (9 - params.overlapSizeLog));
+    DEBUGLOG(4, "overlapLog : %u ", params.overlapSizeLog);
+    DEBUGLOG(4, "overlap Size : %u KB", (U32)(zcs->targetDictSize>>10));
+    zcs->targetSectionSize = params.jobSize ? params.jobSize : (size_t)1 << (params.cParams.windowLog + 2);
+    zcs->targetSectionSize = MAX(ZSTDMT_SECTION_SIZE_MIN, zcs->targetSectionSize);
+    zcs->targetSectionSize = MAX(zcs->targetDictSize, zcs->targetSectionSize);
+    DEBUGLOG(4, "Section Size : %u KB", (U32)(zcs->targetSectionSize>>10));
+    zcs->inBuffSize = zcs->targetDictSize + zcs->targetSectionSize;
+    ZSTDMT_setBufferSize(zcs->bufPool, MAX(zcs->inBuffSize, ZSTD_compressBound(zcs->targetSectionSize)) );
+    zcs->inBuff.buffer = g_nullBuffer;
+    zcs->dictSize = 0;
+    zcs->doneJobID = 0;
+    zcs->nextJobID = 0;
+    zcs->frameEnded = 0;
+    zcs->allJobsCompleted = 0;
+    if (params.fParams.checksumFlag) XXH64_reset(&zcs->xxhState, 0);
+    return 0;
+}
+
+size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
+                             const void* dict, size_t dictSize,
+                                   ZSTD_parameters params,
+                                   unsigned long long pledgedSrcSize)
+{
+    ZSTD_CCtx_params cctxParams = mtctx->params;
+    DEBUGLOG(5, "ZSTDMT_initCStream_advanced");
+    cctxParams.cParams = params.cParams;
+    cctxParams.fParams = params.fParams;
+    return ZSTDMT_initCStream_internal(mtctx, dict, dictSize, ZSTD_dm_auto, NULL,
+                                       cctxParams, pledgedSrcSize);
+}
+
+size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
+                               const ZSTD_CDict* cdict,
+                                     ZSTD_frameParameters fParams,
+                                     unsigned long long pledgedSrcSize)
+{
+    ZSTD_CCtx_params cctxParams = mtctx->params;
+    cctxParams.cParams = ZSTD_getCParamsFromCDict(cdict);
+    cctxParams.fParams = fParams;
+    if (cdict==NULL) return ERROR(dictionary_wrong);   /* method incompatible with NULL cdict */
+    return ZSTDMT_initCStream_internal(mtctx, NULL, 0 /*dictSize*/, ZSTD_dm_auto, cdict,
+                                       cctxParams, pledgedSrcSize);
+}
+
+
+/* ZSTDMT_resetCStream() :
+ * pledgedSrcSize is optional and can be zero == unknown */
+size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* zcs, unsigned long long pledgedSrcSize)
+{
+    if (zcs->params.nbThreads==1)
+        return ZSTD_resetCStream(zcs->cctxPool->cctx[0], pledgedSrcSize);
+    return ZSTDMT_initCStream_internal(zcs, NULL, 0, ZSTD_dm_auto, 0, zcs->params,
+                                       pledgedSrcSize);
+}
+
+size_t ZSTDMT_initCStream(ZSTDMT_CCtx* zcs, int compressionLevel) {
+    ZSTD_parameters const params = ZSTD_getParams(compressionLevel, 0, 0);
+    ZSTD_CCtx_params cctxParams = zcs->params;
+    cctxParams.cParams = params.cParams;
+    cctxParams.fParams = params.fParams;
+    return ZSTDMT_initCStream_internal(zcs, NULL, 0, ZSTD_dm_auto, NULL, cctxParams, 0);
+}
+
+
+static size_t ZSTDMT_createCompressionJob(ZSTDMT_CCtx* zcs, size_t srcSize, unsigned endFrame)
+{
+    unsigned const jobID = zcs->nextJobID & zcs->jobIDMask;
+
+    DEBUGLOG(4, "preparing job %u to compress %u bytes with %u preload ",
+                zcs->nextJobID, (U32)srcSize, (U32)zcs->dictSize);
+    zcs->jobs[jobID].src = zcs->inBuff.buffer;
+    zcs->jobs[jobID].srcStart = zcs->inBuff.buffer.start;
+    zcs->jobs[jobID].srcSize = srcSize;
+    zcs->jobs[jobID].dictSize = zcs->dictSize;
+    assert(zcs->inBuff.filled >= srcSize + zcs->dictSize);
+    zcs->jobs[jobID].params = zcs->params;
+    /* do not calculate checksum within sections, but write it in header for first section */
+    if (zcs->nextJobID) zcs->jobs[jobID].params.fParams.checksumFlag = 0;
+    zcs->jobs[jobID].cdict = zcs->nextJobID==0 ? zcs->cdict : NULL;
+    zcs->jobs[jobID].fullFrameSize = zcs->frameContentSize;
+    zcs->jobs[jobID].dstBuff = g_nullBuffer;
+    zcs->jobs[jobID].cctxPool = zcs->cctxPool;
+    zcs->jobs[jobID].bufPool = zcs->bufPool;
+    zcs->jobs[jobID].firstChunk = (zcs->nextJobID==0);
+    zcs->jobs[jobID].lastChunk = endFrame;
+    zcs->jobs[jobID].jobCompleted = 0;
+    zcs->jobs[jobID].dstFlushed = 0;
+    zcs->jobs[jobID].jobCompleted_mutex = &zcs->jobCompleted_mutex;
+    zcs->jobs[jobID].jobCompleted_cond = &zcs->jobCompleted_cond;
+
+    if (zcs->params.fParams.checksumFlag)
+        XXH64_update(&zcs->xxhState, (const char*)zcs->inBuff.buffer.start + zcs->dictSize, srcSize);
+
+    /* get a new buffer for next input */
+    if (!endFrame) {
+        size_t const newDictSize = MIN(srcSize + zcs->dictSize, zcs->targetDictSize);
+        zcs->inBuff.buffer = ZSTDMT_getBuffer(zcs->bufPool);
+        if (zcs->inBuff.buffer.start == NULL) {   /* not enough memory to allocate next input buffer */
+            zcs->jobs[jobID].jobCompleted = 1;
+            zcs->nextJobID++;
+            ZSTDMT_waitForAllJobsCompleted(zcs);
+            ZSTDMT_releaseAllJobResources(zcs);
+            return ERROR(memory_allocation);
+        }
+        zcs->inBuff.filled -= srcSize + zcs->dictSize - newDictSize;
+        memmove(zcs->inBuff.buffer.start,
+            (const char*)zcs->jobs[jobID].srcStart + zcs->dictSize + srcSize - newDictSize,
+            zcs->inBuff.filled);
+        zcs->dictSize = newDictSize;
+    } else {   /* if (endFrame==1) */
+        zcs->inBuff.buffer = g_nullBuffer;
+        zcs->inBuff.filled = 0;
+        zcs->dictSize = 0;
+        zcs->frameEnded = 1;
+        if (zcs->nextJobID == 0) {
+            /* single chunk exception : checksum is calculated directly within worker thread */
+            zcs->params.fParams.checksumFlag = 0;
+    }   }
+
+    DEBUGLOG(4, "posting job %u : %u bytes  (end:%u) (note : doneJob = %u=>%u)",
+                zcs->nextJobID,
+                (U32)zcs->jobs[jobID].srcSize,
+                zcs->jobs[jobID].lastChunk,
+                zcs->doneJobID,
+                zcs->doneJobID & zcs->jobIDMask);
+    POOL_add(zcs->factory, ZSTDMT_compressChunk, &zcs->jobs[jobID]);   /* this call is blocking when thread worker pool is exhausted */
+    zcs->nextJobID++;
+    return 0;
+}
+
+
+/* ZSTDMT_flushNextJob() :
+ * output : will be updated with amount of data flushed .
+ * blockToFlush : if >0, the function will block and wait if there is no data available to flush .
+ * @return : amount of data remaining within internal buffer, 1 if unknown but > 0, 0 if no more, or an error code */
+static size_t ZSTDMT_flushNextJob(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output, unsigned blockToFlush)
+{
+    unsigned const wJobID = zcs->doneJobID & zcs->jobIDMask;
+    if (zcs->doneJobID == zcs->nextJobID) return 0;   /* all flushed ! */
+    PTHREAD_MUTEX_LOCK(&zcs->jobCompleted_mutex);
+    while (zcs->jobs[wJobID].jobCompleted==0) {
+        DEBUGLOG(5, "waiting for jobCompleted signal from job %u", zcs->doneJobID);
+        if (!blockToFlush) { pthread_mutex_unlock(&zcs->jobCompleted_mutex); return 0; }  /* nothing ready to be flushed => skip */
+        pthread_cond_wait(&zcs->jobCompleted_cond, &zcs->jobCompleted_mutex);  /* block when nothing available to flush */
+    }
+    pthread_mutex_unlock(&zcs->jobCompleted_mutex);
+    /* compression job completed : output can be flushed */
+    {   ZSTDMT_jobDescription job = zcs->jobs[wJobID];
+        if (!job.jobScanned) {
+            if (ZSTD_isError(job.cSize)) {
+                DEBUGLOG(5, "compression error detected ");
+                ZSTDMT_waitForAllJobsCompleted(zcs);
+                ZSTDMT_releaseAllJobResources(zcs);
+                return job.cSize;
+            }
+            DEBUGLOG(5, "zcs->params.fParams.checksumFlag : %u ", zcs->params.fParams.checksumFlag);
+            if (zcs->params.fParams.checksumFlag) {
+                if (zcs->frameEnded && (zcs->doneJobID+1 == zcs->nextJobID)) {  /* write checksum at end of last section */
+                    U32 const checksum = (U32)XXH64_digest(&zcs->xxhState);
+                    DEBUGLOG(5, "writing checksum : %08X \n", checksum);
+                    MEM_writeLE32((char*)job.dstBuff.start + job.cSize, checksum);
+                    job.cSize += 4;
+                    zcs->jobs[wJobID].cSize += 4;
+            }   }
+            zcs->jobs[wJobID].jobScanned = 1;
+        }
+        {   size_t const toWrite = MIN(job.cSize - job.dstFlushed, output->size - output->pos);
+            DEBUGLOG(5, "Flushing %u bytes from job %u ", (U32)toWrite, zcs->doneJobID);
+            memcpy((char*)output->dst + output->pos, (const char*)job.dstBuff.start + job.dstFlushed, toWrite);
+            output->pos += toWrite;
+            job.dstFlushed += toWrite;
+        }
+        if (job.dstFlushed == job.cSize) {   /* output buffer fully flushed => move to next one */
+            ZSTDMT_releaseBuffer(zcs->bufPool, job.dstBuff);
+            zcs->jobs[wJobID].dstBuff = g_nullBuffer;
+            zcs->jobs[wJobID].jobCompleted = 0;
+            zcs->doneJobID++;
+        } else {
+            zcs->jobs[wJobID].dstFlushed = job.dstFlushed;
+        }
+        /* return value : how many bytes left in buffer ; fake it to 1 if unknown but >0 */
+        if (job.cSize > job.dstFlushed) return (job.cSize - job.dstFlushed);
+        if (zcs->doneJobID < zcs->nextJobID) return 1;   /* still some buffer to flush */
+        zcs->allJobsCompleted = zcs->frameEnded;   /* frame completed and entirely flushed */
+        return 0;   /* everything flushed */
+}   }
+
+
+/** ZSTDMT_compressStream_generic() :
+ *  internal use only
+ *  assumption : output and input are valid (pos <= size)
+ * @return : minimum amount of data remaining to flush, 0 if none */
+size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
+                                     ZSTD_outBuffer* output,
+                                     ZSTD_inBuffer* input,
+                                     ZSTD_EndDirective endOp)
+{
+    size_t const newJobThreshold = mtctx->dictSize + mtctx->targetSectionSize;
+    assert(output->pos <= output->size);
+    assert(input->pos  <= input->size);
+    if ((mtctx->frameEnded) && (endOp==ZSTD_e_continue)) {
+        /* current frame being ended. Only flush/end are allowed. Or start new frame with init */
+        return ERROR(stage_wrong);
+    }
+    if (mtctx->params.nbThreads==1) {  /* delegate to single-thread (synchronous) */
+        return ZSTD_compressStream_generic(mtctx->cctxPool->cctx[0], output, input, endOp);
+    }
+
+    /* single-pass shortcut (note : this is synchronous-mode) */
+    if ( (mtctx->nextJobID==0)      /* just started */
+      && (mtctx->inBuff.filled==0)  /* nothing buffered */
+      && (endOp==ZSTD_e_end)        /* end order */
+      && (output->size - output->pos >= ZSTD_compressBound(input->size - input->pos)) ) { /* enough room */
+        size_t const cSize = ZSTDMT_compress_advanced_internal(mtctx,
+                (char*)output->dst + output->pos, output->size - output->pos,
+                (const char*)input->src + input->pos, input->size - input->pos,
+                mtctx->cdict, mtctx->params);
+        if (ZSTD_isError(cSize)) return cSize;
+        input->pos = input->size;
+        output->pos += cSize;
+        ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->inBuff.buffer);  /* was allocated in initStream */
+        mtctx->allJobsCompleted = 1;
+        mtctx->frameEnded = 1;
+        return 0;
+    }
+
+    /* fill input buffer */
+    if (input->size > input->pos) {   /* support NULL input */
+        if (mtctx->inBuff.buffer.start == NULL) {
+            mtctx->inBuff.buffer = ZSTDMT_getBuffer(mtctx->bufPool);
+            if (mtctx->inBuff.buffer.start == NULL) return ERROR(memory_allocation);
+            mtctx->inBuff.filled = 0;
+        }
+        {   size_t const toLoad = MIN(input->size - input->pos, mtctx->inBuffSize - mtctx->inBuff.filled);
+            DEBUGLOG(5, "inBuff:%08X;  inBuffSize=%u;  ToCopy=%u", (U32)(size_t)mtctx->inBuff.buffer.start, (U32)mtctx->inBuffSize, (U32)toLoad);
+            memcpy((char*)mtctx->inBuff.buffer.start + mtctx->inBuff.filled, (const char*)input->src + input->pos, toLoad);
+            input->pos += toLoad;
+            mtctx->inBuff.filled += toLoad;
+    }   }
+
+    if ( (mtctx->inBuff.filled >= newJobThreshold)  /* filled enough : let's compress */
+      && (mtctx->nextJobID <= mtctx->doneJobID + mtctx->jobIDMask) ) {   /* avoid overwriting job round buffer */
+        CHECK_F( ZSTDMT_createCompressionJob(mtctx, mtctx->targetSectionSize, 0 /* endFrame */) );
+    }
+
+    /* check for potential compressed data ready to be flushed */
+    CHECK_F( ZSTDMT_flushNextJob(mtctx, output, (mtctx->inBuff.filled == mtctx->inBuffSize) /* blockToFlush */) ); /* block if it wasn't possible to create new job due to saturation */
+
+    if (input->pos < input->size)  /* input not consumed : do not flush yet */
+        endOp = ZSTD_e_continue;
+
+    switch(endOp)
+    {
+        case ZSTD_e_flush:
+            return ZSTDMT_flushStream(mtctx, output);
+        case ZSTD_e_end:
+            return ZSTDMT_endStream(mtctx, output);
+        case ZSTD_e_continue:
+            return 1;
+        default:
+            return ERROR(GENERIC);   /* invalid endDirective */
+    }
+}
+
+
+size_t ZSTDMT_compressStream(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output, ZSTD_inBuffer* input)
+{
+    CHECK_F( ZSTDMT_compressStream_generic(zcs, output, input, ZSTD_e_continue) );
+
+    /* recommended next input size : fill current input buffer */
+    return zcs->inBuffSize - zcs->inBuff.filled;   /* note : could be zero when input buffer is fully filled and no more availability to create new job */
+}
+
+
+static size_t ZSTDMT_flushStream_internal(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output, unsigned endFrame)
+{
+    size_t const srcSize = zcs->inBuff.filled - zcs->dictSize;
+
+    if ( ((srcSize > 0) || (endFrame && !zcs->frameEnded))
+       && (zcs->nextJobID <= zcs->doneJobID + zcs->jobIDMask) ) {
+        CHECK_F( ZSTDMT_createCompressionJob(zcs, srcSize, endFrame) );
+    }
+
+    /* check if there is any data available to flush */
+    return ZSTDMT_flushNextJob(zcs, output, 1 /* blockToFlush */);
+}
+
+
+size_t ZSTDMT_flushStream(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output)
+{
+    DEBUGLOG(5, "ZSTDMT_flushStream");
+    if (zcs->params.nbThreads==1)
+        return ZSTD_flushStream(zcs->cctxPool->cctx[0], output);
+    return ZSTDMT_flushStream_internal(zcs, output, 0 /* endFrame */);
+}
+
+size_t ZSTDMT_endStream(ZSTDMT_CCtx* zcs, ZSTD_outBuffer* output)
+{
+    DEBUGLOG(4, "ZSTDMT_endStream");
+    if (zcs->params.nbThreads==1)
+        return ZSTD_endStream(zcs->cctxPool->cctx[0], output);
+    return ZSTDMT_flushStream_internal(zcs, output, 1 /* endFrame */);
+}
diff --git a/contrib/zstd/zstdmt_compress.h b/contrib/zstd/zstdmt_compress.h
new file mode 100644
index 000000000..5b3ddc9d1
--- /dev/null
+++ b/contrib/zstd/zstdmt_compress.h
@@ -0,0 +1,134 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+ #ifndef ZSTDMT_COMPRESS_H
+ #define ZSTDMT_COMPRESS_H
+
+ #if defined (__cplusplus)
+ extern "C" {
+ #endif
+
+
+/* Note : This is an internal API.
+ *        Some methods are still exposed (ZSTDLIB_API),
+ *        because it used to be the only way to invoke MT compression.
+ *        Now, it's recommended to use ZSTD_compress_generic() instead.
+ *        These methods will stop being exposed in a future version */
+
+/* ===   Dependencies   === */
+#include <stddef.h>                /* size_t */
+#ifndef ZSTD_STATIC_LINKING_ONLY
+#define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_parameters */
+#endif
+#include "zstd.h"            /* ZSTD_inBuffer, ZSTD_outBuffer, ZSTDLIB_API */
+
+
+/* ===   Memory management   === */
+typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
+ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbThreads);
+ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbThreads,
+                                                    ZSTD_customMem cMem);
+ZSTDLIB_API size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
+
+ZSTDLIB_API size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
+
+
+/* ===   Simple buffer-to-butter one-pass function   === */
+
+ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
+                                       void* dst, size_t dstCapacity,
+                                 const void* src, size_t srcSize,
+                                       int compressionLevel);
+
+
+
+/* ===   Streaming functions   === */
+
+ZSTDLIB_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
+ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize);    /**< pledgedSrcSize is optional and can be zero == unknown */
+
+ZSTDLIB_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
+
+ZSTDLIB_API size_t ZSTDMT_flushStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);   /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
+ZSTDLIB_API size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);     /**< @return : 0 == all flushed; >0 : still some data to be flushed; or an error code (ZSTD_isError()) */
+
+
+/* ===   Advanced functions and parameters  === */
+
+#ifndef ZSTDMT_SECTION_SIZE_MIN
+#  define ZSTDMT_SECTION_SIZE_MIN (1U << 20)   /* 1 MB - Minimum size of each compression job */
+#endif
+
+ZSTDLIB_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
+                                           void* dst, size_t dstCapacity,
+                                     const void* src, size_t srcSize,
+                                     const ZSTD_CDict* cdict,
+                                           ZSTD_parameters const params,
+                                           unsigned overlapLog);
+
+ZSTDLIB_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
+                                        const void* dict, size_t dictSize,   /* dict can be released after init, a local copy is preserved within zcs */
+                                        ZSTD_parameters params,
+                                        unsigned long long pledgedSrcSize);  /* pledgedSrcSize is optional and can be zero == unknown */
+
+ZSTDLIB_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
+                                        const ZSTD_CDict* cdict,
+                                        ZSTD_frameParameters fparams,
+                                        unsigned long long pledgedSrcSize);  /* note : zero means empty */
+
+/* ZSTDMT_parameter :
+ * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
+typedef enum {
+    ZSTDMT_p_sectionSize,        /* size of input "section". Each section is compressed in parallel. 0 means default, which is dynamically determined within compression functions */
+    ZSTDMT_p_overlapSectionLog   /* Log of overlapped section; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window */
+} ZSTDMT_parameter;
+
+/* ZSTDMT_setMTCtxParameter() :
+ * allow setting individual parameters, one at a time, among a list of enums defined in ZSTDMT_parameter.
+ * The function must be called typically after ZSTD_createCCtx().
+ * Parameters not explicitly reset by ZSTDMT_init*() remain the same in consecutive compression sessions.
+ * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
+ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned value);
+
+
+/*! ZSTDMT_compressStream_generic() :
+ *  Combines ZSTDMT_compressStream() with ZSTDMT_flushStream() or ZSTDMT_endStream()
+ *  depending on flush directive.
+ * @return : minimum amount of data still to be flushed
+ *           0 if fully flushed
+ *           or an error code */
+ZSTDLIB_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
+                                                ZSTD_outBuffer* output,
+                                                ZSTD_inBuffer* input,
+                                                ZSTD_EndDirective endOp);
+
+
+/* ===   Private definitions; never ever use directly  === */
+
+size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, unsigned value);
+
+size_t ZSTDMT_initializeCCtxParameters(ZSTD_CCtx_params* params, unsigned nbThreads);
+
+/*! ZSTDMT_initCStream_internal() :
+ *  Private use only. Init streaming operation.
+ *  expects params to be valid.
+ *  must receive dict, or cdict, or none, but not both.
+ *  @return : 0, or an error code */
+size_t ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
+                    const void* dict, size_t dictSize, ZSTD_dictMode_e dictMode,
+                    const ZSTD_CDict* cdict,
+                    ZSTD_CCtx_params params, unsigned long long pledgedSrcSize);
+
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif   /* ZSTDMT_COMPRESS_H */