[Openvpn-devel] compat/lz4: Update to v1.9.2

Message ID 20201001154658.9798-1-davids@openvpn.net
State Accepted
Headers show
Series [Openvpn-devel] compat/lz4: Update to v1.9.2 | expand

Commit Message

David Sommerseth Oct. 1, 2020, 5:46 a.m. UTC
It's a long while since the bundled lz4 library has received an update.
It pulls in a lot of various fixes and enhancements, some of the changes
fixes compiler warnings and hardens the code a bit too.

Signed-off-by: David Sommerseth <davids@openvpn.net>
---
 src/compat/compat-lz4.c | 2191 ++++++++++++++++++++++++++++-----------
 src/compat/compat-lz4.h |  729 +++++++++----
 2 files changed, 2078 insertions(+), 842 deletions(-)

Comments

Arne Schwabe Oct. 1, 2020, 10:50 p.m. UTC | #1
Am 01.10.20 um 17:46 schrieb David Sommerseth:
> It's a long while since the bundled lz4 library has received an update.
> It pulls in a lot of various fixes and enhancements, some of the changes
> fixes compiler warnings and hardens the code a bit too.
> 

Ack for release/2.5. I double checked that the file is identical (modulo
compat defines) to the upsteam.

I think that this file is outdated so much shows that we do not really
care about compression so much. Also since we included it, lz4 libraries
have been available on all major distribution and not having lz4 is not
as problematic as before. For master I would like to see this
compat-lz4.* removed instead of being updated. For the 2.5.0 release it
is too late to remove, so okay for that.


Acked-By: Arne Schwabe <arne@rfc2549.org>
Илья Шипицин Oct. 2, 2020, 12:03 a.m. UTC | #2
пт, 2 окт. 2020 г. в 13:51, Arne Schwabe <arne@rfc2549.org>:

> Am 01.10.20 um 17:46 schrieb David Sommerseth:
> > It's a long while since the bundled lz4 library has received an update.
> > It pulls in a lot of various fixes and enhancements, some of the changes
> > fixes compiler warnings and hardens the code a bit too.
> >
>
> Ack for release/2.5. I double checked that the file is identical (modulo
> compat defines) to the upsteam.
>
> I think that this file is outdated so much shows that we do not really
> care about compression so much. Also since we included it, lz4 libraries
> have been available on all major distribution and not having lz4 is not
> as problematic as before. For master I would like to see this
> compat-lz4.* removed instead of being updated. For the 2.5.0 release it
> is too late to remove, so okay for that.
>


windows build process uses bundled lz4. if we simply remove it, windows
build silently switch to "no lz4 enabled ... and that is ok"


>
>
> Acked-By: Arne Schwabe <arne@rfc2549.org>
>
> _______________________________________________
> Openvpn-devel mailing list
> Openvpn-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/openvpn-devel
>
<div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">пт, 2 окт. 2020 г. в 13:51, Arne Schwabe &lt;<a href="mailto:arne@rfc2549.org">arne@rfc2549.org</a>&gt;:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Am 01.10.20 um 17:46 schrieb David Sommerseth:<br>
&gt; It&#39;s a long while since the bundled lz4 library has received an update.<br>
&gt; It pulls in a lot of various fixes and enhancements, some of the changes<br>
&gt; fixes compiler warnings and hardens the code a bit too.<br>
&gt; <br>
<br>
Ack for release/2.5. I double checked that the file is identical (modulo<br>
compat defines) to the upsteam.<br>
<br>
I think that this file is outdated so much shows that we do not really<br>
care about compression so much. Also since we included it, lz4 libraries<br>
have been available on all major distribution and not having lz4 is not<br>
as problematic as before. For master I would like to see this<br>
compat-lz4.* removed instead of being updated. For the 2.5.0 release it<br>
is too late to remove, so okay for that.<br></blockquote><div><br></div><div><br></div><div>windows build process uses bundled lz4. if we simply remove it, windows build silently switch to &quot;no lz4 enabled ... and that is ok&quot;<br></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
<br>
<br>
Acked-By: Arne Schwabe &lt;<a href="mailto:arne@rfc2549.org" target="_blank">arne@rfc2549.org</a>&gt;<br>
<br>
_______________________________________________<br>
Openvpn-devel mailing list<br>
<a href="mailto:Openvpn-devel@lists.sourceforge.net" target="_blank">Openvpn-devel@lists.sourceforge.net</a><br>
<a href="https://lists.sourceforge.net/lists/listinfo/openvpn-devel" rel="noreferrer" target="_blank">https://lists.sourceforge.net/lists/listinfo/openvpn-devel</a><br>
</blockquote></div></div>
Arne Schwabe Oct. 2, 2020, 12:11 a.m. UTC | #3
Am 02.10.20 um 12:03 schrieb Илья Шипицин:
> 
> 
> пт, 2 окт. 2020 г. в 13:51, Arne Schwabe <arne@rfc2549.org
> <mailto:arne@rfc2549.org>>:
> 
>     Am 01.10.20 um 17:46 schrieb David Sommerseth:
>     > It's a long while since the bundled lz4 library has received an
>     update.
>     > It pulls in a lot of various fixes and enhancements, some of the
>     changes
>     > fixes compiler warnings and hardens the code a bit too.
>     >
> 
>     Ack for release/2.5. I double checked that the file is identical (modulo
>     compat defines) to the upsteam.
> 
>     I think that this file is outdated so much shows that we do not really
>     care about compression so much. Also since we included it, lz4 libraries
>     have been available on all major distribution and not having lz4 is not
>     as problematic as before. For master I would like to see this
>     compat-lz4.* removed instead of being updated. For the 2.5.0 release it
>     is too late to remove, so okay for that.
> 
> 
> 
> windows build process uses bundled lz4. if we simply remove it, windows
> build silently switch to "no lz4 enabled ... and that is ok"

I think we can find a better solution to include lz4 in the windows
build then putting the dependency in our repository.

Arne
Gert Doering Oct. 2, 2020, 1:06 a.m. UTC | #4
Your patch has been applied to the master, release/2.5 and 2.4 branch.

(Even though I agree with Arne that it sounds like a good plan to
just kick out compat-lz4 in master, given that the times have changed
quite a bit compared to the initial import 6.5 years(!) ago, for
consistency reasons I prefer to have it consistent between master
and the release branches until we kick it out).

Parts of my test rig excercise LZ4 compression, so it's at least
lightly tested (on gentoo, with no system-wide lz4 installed).

commit 0f44a9080530df70410106c244e9efc7f2d8a802 (master)
commit 156db76f6670cb2645aaebf036d91ad4020d8aae (release/2.5)
commit 2c0ed08ab2e340664fda9e790c14bd6f0bb92af5 (release/2.4)
Author: David Sommerseth
Date:   Thu Oct 1 17:46:58 2020 +0200

     compat/lz4: Update to v1.9.2

     Signed-off-by: David Sommerseth <davids@openvpn.net>
     Acked-by: Arne Schwabe <arne@rfc2549.org>
     Message-Id: <20201001154658.9798-1-davids@openvpn.net>
     URL: https://www.mail-archive.com/search?l=mid&q=20201001154658.9798-1-davids@openvpn.net
     Signed-off-by: Gert Doering <gert@greenie.muc.de>


--
kind regards,

Gert Doering

Patch

diff --git a/src/compat/compat-lz4.c b/src/compat/compat-lz4.c
index 723157d2..26a3980b 100644
--- a/src/compat/compat-lz4.c
+++ b/src/compat/compat-lz4.c
@@ -1,5 +1,5 @@ 
 /* This file has been backported by dev-tools/lz4-rebaser.sh
- * from upstream lz4 commit 7bb64ff2b69a9f8367de (v1.7.5)
+ * from upstream lz4 commit fdf2ef5809ca875c4545 (v1.9.2)
  */
 #ifdef HAVE_CONFIG_H
 #include "config.h"
@@ -10,7 +10,7 @@ 
 #ifdef NEED_COMPAT_LZ4
 /*
    LZ4 - Fast LZ compression algorithm
-   Copyright (C) 2011-2016, Yann Collet.
+   Copyright (C) 2011-present, Yann Collet.
 
    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
 
@@ -42,17 +42,16 @@ 
     - LZ4 source repository : https://github.com/lz4/lz4
 */
 
-
 /*-************************************
 *  Tuning parameters
 **************************************/
 /*
- * HEAPMODE :
+ * LZ4_HEAPMODE :
  * Select how default compression functions will allocate memory for their hash table,
  * in memory stack (0:default, fastest), or in memory heap (1:requires malloc()).
  */
-#ifndef HEAPMODE
-#  define HEAPMODE 0
+#ifndef LZ4_HEAPMODE
+#  define LZ4_HEAPMODE 0
 #endif
 
 /*
@@ -73,16 +72,17 @@ 
  * Method 1 : `__packed` statement. It depends on compiler extension (ie, not portable).
  *            This method is safe if your compiler supports it, and *generally* as fast or faster than `memcpy`.
  * Method 2 : direct access. This method is portable but violate C standard.
- *            It can generate buggy code on targets which generate assembly depending on alignment.
+ *            It can generate buggy code on targets which assembly generation depends on alignment.
  *            But in some circumstances, it's the only known way to get the most performance (ie GCC + ARMv6)
  * See https://fastcompression.blogspot.fr/2015/08/accessing-unaligned-memory.html for details.
  * Prefer these methods in priority order (0 > 1 > 2)
  */
-#ifndef LZ4_FORCE_MEMORY_ACCESS   /* can be defined externally, on command line for example */
-#  if defined(__GNUC__) && ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
+#ifndef LZ4_FORCE_MEMORY_ACCESS   /* can be defined externally */
+#  if defined(__GNUC__) && \
+  ( defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || defined(__ARM_ARCH_6K__) \
+  || defined(__ARM_ARCH_6Z__) || defined(__ARM_ARCH_6ZK__) || defined(__ARM_ARCH_6T2__) )
 #    define LZ4_FORCE_MEMORY_ACCESS 2
-#  elif defined(__INTEL_COMPILER) || \
-  (defined(__GNUC__) && ( defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_7R__) || defined(__ARM_ARCH_7M__) || defined(__ARM_ARCH_7S__) ))
+#  elif (defined(__INTEL_COMPILER) && !defined(_WIN32)) || defined(__GNUC__)
 #    define LZ4_FORCE_MEMORY_ACCESS 1
 #  endif
 #endif
@@ -91,14 +91,32 @@ 
  * LZ4_FORCE_SW_BITCOUNT
  * Define this parameter if your target system or compiler does not support hardware bit count
  */
-#if defined(_MSC_VER) && defined(_WIN32_WCE)   /* Visual Studio for Windows CE does not support Hardware bit count */
+#if defined(_MSC_VER) && defined(_WIN32_WCE)   /* Visual Studio for WinCE doesn't support Hardware bit count */
 #  define LZ4_FORCE_SW_BITCOUNT
 #endif
 
 
+
 /*-************************************
 *  Dependency
 **************************************/
+/*
+ * LZ4_SRC_INCLUDED:
+ * Amalgamation flag, whether lz4.c is included
+ */
+#ifndef LZ4_SRC_INCLUDED
+#  define LZ4_SRC_INCLUDED 1
+#endif
+
+#ifndef LZ4_STATIC_LINKING_ONLY
+#define LZ4_STATIC_LINKING_ONLY
+#endif
+
+#ifndef LZ4_DISABLE_DEPRECATE_WARNINGS
+#define LZ4_DISABLE_DEPRECATE_WARNINGS /* due to LZ4_decompress_safe_withPrefix64k */
+#endif
+
+#define LZ4_STATIC_LINKING_ONLY  /* LZ4_DISTANCE_MAX */
 #include "compat-lz4.h"
 /* see also "memory routines" below */
 
@@ -107,42 +125,130 @@ 
 *  Compiler Options
 **************************************/
 #ifdef _MSC_VER    /* Visual Studio */
-#  define FORCE_INLINE static __forceinline
 #  include <intrin.h>
 #  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
 #  pragma warning(disable : 4293)        /* disable: C4293: too large shift (32-bits) */
-#else
-#  if defined(__GNUC__) || defined(__clang__)
-#    define FORCE_INLINE static inline __attribute__((always_inline))
-#  elif defined(__cplusplus) || (defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
-#    define FORCE_INLINE static inline
-#  else
-#    define FORCE_INLINE static
-#  endif
 #endif  /* _MSC_VER */
 
+#ifndef LZ4_FORCE_INLINE
+#  ifdef _MSC_VER    /* Visual Studio */
+#    define LZ4_FORCE_INLINE static __forceinline
+#  else
+#    if defined (__cplusplus) || defined (__STDC_VERSION__) && __STDC_VERSION__ >= 199901L   /* C99 */
+#      ifdef __GNUC__
+#        define LZ4_FORCE_INLINE static inline __attribute__((always_inline))
+#      else
+#        define LZ4_FORCE_INLINE static inline
+#      endif
+#    else
+#      define LZ4_FORCE_INLINE static
+#    endif /* __STDC_VERSION__ */
+#  endif  /* _MSC_VER */
+#endif /* LZ4_FORCE_INLINE */
+
+/* LZ4_FORCE_O2_GCC_PPC64LE and LZ4_FORCE_O2_INLINE_GCC_PPC64LE
+ * gcc on ppc64le generates an unrolled SIMDized loop for LZ4_wildCopy8,
+ * together with a simple 8-byte copy loop as a fall-back path.
+ * However, this optimization hurts the decompression speed by >30%,
+ * because the execution does not go to the optimized loop
+ * for typical compressible data, and all of the preamble checks
+ * before going to the fall-back path become useless overhead.
+ * This optimization happens only with the -O3 flag, and -O2 generates
+ * a simple 8-byte copy loop.
+ * With gcc on ppc64le, all of the LZ4_decompress_* and LZ4_wildCopy8
+ * functions are annotated with __attribute__((optimize("O2"))),
+ * and also LZ4_wildCopy8 is forcibly inlined, so that the O2 attribute
+ * of LZ4_wildCopy8 does not affect the compression speed.
+ */
+#if defined(__PPC64__) && defined(__LITTLE_ENDIAN__) && defined(__GNUC__) && !defined(__clang__)
+#  define LZ4_FORCE_O2_GCC_PPC64LE __attribute__((optimize("O2")))
+#  define LZ4_FORCE_O2_INLINE_GCC_PPC64LE __attribute__((optimize("O2"))) LZ4_FORCE_INLINE
+#else
+#  define LZ4_FORCE_O2_GCC_PPC64LE
+#  define LZ4_FORCE_O2_INLINE_GCC_PPC64LE static
+#endif
+
 #if (defined(__GNUC__) && (__GNUC__ >= 3)) || (defined(__INTEL_COMPILER) && (__INTEL_COMPILER >= 800)) || defined(__clang__)
 #  define expect(expr,value)    (__builtin_expect ((expr),(value)) )
 #else
 #  define expect(expr,value)    (expr)
 #endif
 
+#ifndef likely
 #define likely(expr)     expect((expr) != 0, 1)
+#endif
+#ifndef unlikely
 #define unlikely(expr)   expect((expr) != 0, 0)
+#endif
 
 
 /*-************************************
 *  Memory routines
 **************************************/
 #include <stdlib.h>   /* malloc, calloc, free */
-#define ALLOCATOR(n,s) calloc(n,s)
-#define FREEMEM        free
+#define ALLOC(s)          malloc(s)
+#define ALLOC_AND_ZERO(s) calloc(1,s)
+#define FREEMEM(p)        free(p)
 #include <string.h>   /* memset, memcpy */
-#define MEM_INIT       memset
+#define MEM_INIT(p,v,s)   memset((p),(v),(s))
 
 
 /*-************************************
-*  Basic Types
+*  Common Constants
+**************************************/
+#define MINMATCH 4
+
+#define WILDCOPYLENGTH 8
+#define LASTLITERALS   5   /* see ../doc/lz4_Block_format.md#parsing-restrictions */
+#define MFLIMIT       12   /* see ../doc/lz4_Block_format.md#parsing-restrictions */
+#define MATCH_SAFEGUARD_DISTANCE  ((2*WILDCOPYLENGTH) - MINMATCH)   /* ensure it's possible to write 2 x wildcopyLength without overflowing output buffer */
+#define FASTLOOP_SAFE_DISTANCE 64
+static const int LZ4_minLength = (MFLIMIT+1);
+
+#define KB *(1 <<10)
+#define MB *(1 <<20)
+#define GB *(1U<<30)
+
+#define LZ4_DISTANCE_ABSOLUTE_MAX 65535
+#if (LZ4_DISTANCE_MAX > LZ4_DISTANCE_ABSOLUTE_MAX)   /* max supported by LZ4 format */
+#  error "LZ4_DISTANCE_MAX is too big : must be <= 65535"
+#endif
+
+#define ML_BITS  4
+#define ML_MASK  ((1U<<ML_BITS)-1)
+#define RUN_BITS (8-ML_BITS)
+#define RUN_MASK ((1U<<RUN_BITS)-1)
+
+
+/*-************************************
+*  Error detection
+**************************************/
+#if defined(LZ4_DEBUG) && (LZ4_DEBUG>=1)
+#  include <assert.h>
+#else
+#  ifndef assert
+#    define assert(condition) ((void)0)
+#  endif
+#endif
+
+#define LZ4_STATIC_ASSERT(c)   { enum { LZ4_static_assert = 1/(int)(!!(c)) }; }   /* use after variable declarations */
+
+#if defined(LZ4_DEBUG) && (LZ4_DEBUG>=2)
+#  include <stdio.h>
+static int g_debuglog_enable = 1;
+#  define DEBUGLOG(l, ...) {                                  \
+                if ((g_debuglog_enable) && (l<=LZ4_DEBUG)) {  \
+                    fprintf(stderr, __FILE__ ": ");           \
+                    fprintf(stderr, __VA_ARGS__);             \
+                    fprintf(stderr, " \n");                   \
+            }   }
+#else
+#  define DEBUGLOG(l, ...)      {}    /* disabled */
+#endif
+
+
+/*-************************************
+*  Types
 **************************************/
 #if defined(__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
 # include <stdint.h>
@@ -167,6 +273,13 @@ 
   typedef size_t reg_t;   /* 32-bits in x32 mode */
 #endif
 
+typedef enum {
+    notLimited = 0,
+    limitedOutput = 1,
+    fillOutput = 2
+} limitedOutput_directive;
+
+
 /*-************************************
 *  Reading and writing into memory
 **************************************/
@@ -200,7 +313,7 @@  static reg_t LZ4_read_ARCH(const void* ptr) { return ((const unalign*)ptr)->uArc
 static void LZ4_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
 static void LZ4_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; }
 
-#else  /* safe and portable access through memcpy() */
+#else  /* safe and portable access using memcpy() */
 
 static U16 LZ4_read16(const void* memPtr)
 {
@@ -251,55 +364,113 @@  static void LZ4_writeLE16(void* memPtr, U16 value)
     }
 }
 
-static void LZ4_copy8(void* dst, const void* src)
-{
-    memcpy(dst,src,8);
-}
-
 /* customized variant of memcpy, which can overwrite up to 8 bytes beyond dstEnd */
-static void LZ4_wildCopy(void* dstPtr, const void* srcPtr, void* dstEnd)
+LZ4_FORCE_O2_INLINE_GCC_PPC64LE
+void LZ4_wildCopy8(void* dstPtr, const void* srcPtr, void* dstEnd)
 {
     BYTE* d = (BYTE*)dstPtr;
     const BYTE* s = (const BYTE*)srcPtr;
     BYTE* const e = (BYTE*)dstEnd;
 
-    do { LZ4_copy8(d,s); d+=8; s+=8; } while (d<e);
+    do { memcpy(d,s,8); d+=8; s+=8; } while (d<e);
 }
 
+static const unsigned inc32table[8] = {0, 1, 2,  1,  0,  4, 4, 4};
+static const int      dec64table[8] = {0, 0, 0, -1, -4,  1, 2, 3};
 
-/*-************************************
-*  Common Constants
-**************************************/
-#define MINMATCH 4
 
-#define WILDCOPYLENGTH 8
-#define LASTLITERALS 5
-#define MFLIMIT (WILDCOPYLENGTH+MINMATCH)
-static const int LZ4_minLength = (MFLIMIT+1);
+#ifndef LZ4_FAST_DEC_LOOP
+#  if defined(__i386__) || defined(__x86_64__)
+#    define LZ4_FAST_DEC_LOOP 1
+#  elif defined(__aarch64__) && !defined(__clang__)
+     /* On aarch64, we disable this optimization for clang because on certain
+      * mobile chipsets and clang, it reduces performance. For more information
+      * refer to https://github.com/lz4/lz4/pull/707. */
+#    define LZ4_FAST_DEC_LOOP 1
+#  else
+#    define LZ4_FAST_DEC_LOOP 0
+#  endif
+#endif
 
-#define KB *(1 <<10)
-#define MB *(1 <<20)
-#define GB *(1U<<30)
+#if LZ4_FAST_DEC_LOOP
 
-#define MAXD_LOG 16
-#define MAX_DISTANCE ((1 << MAXD_LOG) - 1)
+LZ4_FORCE_O2_INLINE_GCC_PPC64LE void
+LZ4_memcpy_using_offset_base(BYTE* dstPtr, const BYTE* srcPtr, BYTE* dstEnd, const size_t offset)
+{
+    if (offset < 8) {
+        dstPtr[0] = srcPtr[0];
+        dstPtr[1] = srcPtr[1];
+        dstPtr[2] = srcPtr[2];
+        dstPtr[3] = srcPtr[3];
+        srcPtr += inc32table[offset];
+        memcpy(dstPtr+4, srcPtr, 4);
+        srcPtr -= dec64table[offset];
+        dstPtr += 8;
+    } else {
+        memcpy(dstPtr, srcPtr, 8);
+        dstPtr += 8;
+        srcPtr += 8;
+    }
 
-#define ML_BITS  4
-#define ML_MASK  ((1U<<ML_BITS)-1)
-#define RUN_BITS (8-ML_BITS)
-#define RUN_MASK ((1U<<RUN_BITS)-1)
+    LZ4_wildCopy8(dstPtr, srcPtr, dstEnd);
+}
 
+/* customized variant of memcpy, which can overwrite up to 32 bytes beyond dstEnd
+ * this version copies two times 16 bytes (instead of one time 32 bytes)
+ * because it must be compatible with offsets >= 16. */
+LZ4_FORCE_O2_INLINE_GCC_PPC64LE void
+LZ4_wildCopy32(void* dstPtr, const void* srcPtr, void* dstEnd)
+{
+    BYTE* d = (BYTE*)dstPtr;
+    const BYTE* s = (const BYTE*)srcPtr;
+    BYTE* const e = (BYTE*)dstEnd;
 
-/*-************************************
-*  Common Utils
-**************************************/
-#define LZ4_STATIC_ASSERT(c)    { enum { LZ4_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+    do { memcpy(d,s,16); memcpy(d+16,s+16,16); d+=32; s+=32; } while (d<e);
+}
+
+/* LZ4_memcpy_using_offset()  presumes :
+ * - dstEnd >= dstPtr + MINMATCH
+ * - there is at least 8 bytes available to write after dstEnd */
+LZ4_FORCE_O2_INLINE_GCC_PPC64LE void
+LZ4_memcpy_using_offset(BYTE* dstPtr, const BYTE* srcPtr, BYTE* dstEnd, const size_t offset)
+{
+    BYTE v[8];
+
+    assert(dstEnd >= dstPtr + MINMATCH);
+    LZ4_write32(dstPtr, 0);   /* silence an msan warning when offset==0 */
+
+    switch(offset) {
+    case 1:
+        memset(v, *srcPtr, 8);
+        break;
+    case 2:
+        memcpy(v, srcPtr, 2);
+        memcpy(&v[2], srcPtr, 2);
+        memcpy(&v[4], &v[0], 4);
+        break;
+    case 4:
+        memcpy(v, srcPtr, 4);
+        memcpy(&v[4], srcPtr, 4);
+        break;
+    default:
+        LZ4_memcpy_using_offset_base(dstPtr, srcPtr, dstEnd, offset);
+        return;
+    }
+
+    memcpy(dstPtr, v, 8);
+    dstPtr += 8;
+    while (dstPtr < dstEnd) {
+        memcpy(dstPtr, v, 8);
+        dstPtr += 8;
+    }
+}
+#endif
 
 
 /*-************************************
 *  Common functions
 **************************************/
-static unsigned LZ4_NbCommonBytes (register reg_t val)
+static unsigned LZ4_NbCommonBytes (reg_t val)
 {
     if (LZ4_isLittleEndian()) {
         if (sizeof(val)==8) {
@@ -308,9 +479,16 @@  static unsigned LZ4_NbCommonBytes (register reg_t val)
             _BitScanForward64( &r, (U64)val );
             return (int)(r>>3);
 #       elif (defined(__clang__) || (defined(__GNUC__) && (__GNUC__>=3))) && !defined(LZ4_FORCE_SW_BITCOUNT)
-            return (__builtin_ctzll((U64)val) >> 3);
+            return (unsigned)__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 };
+            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 */ {
@@ -319,23 +497,29 @@  static unsigned LZ4_NbCommonBytes (register reg_t val)
             _BitScanForward( &r, (U32)val );
             return (int)(r>>3);
 #       elif (defined(__clang__) || (defined(__GNUC__) && (__GNUC__>=3))) && !defined(LZ4_FORCE_SW_BITCOUNT)
-            return (__builtin_ctz((U32)val) >> 3);
+            return (unsigned)__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 };
+            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 (sizeof(val)==8) {
+        if (sizeof(val)==8) {   /* 64-bits */
 #       if defined(_MSC_VER) && defined(_WIN64) && !defined(LZ4_FORCE_SW_BITCOUNT)
             unsigned long r = 0;
             _BitScanReverse64( &r, val );
             return (unsigned)(r>>3);
 #       elif (defined(__clang__) || (defined(__GNUC__) && (__GNUC__>=3))) && !defined(LZ4_FORCE_SW_BITCOUNT)
-            return (__builtin_clzll((U64)val) >> 3);
+            return (unsigned)__builtin_clzll((U64)val) >> 3;
 #       else
+            static const U32 by32 = sizeof(val)*4;  /* 32 on 64 bits (goal), 16 on 32 bits.
+                Just to avoid some static analyzer complaining about shift by 32 on 32-bits target.
+                Note that this code path is never triggered in 32-bits mode. */
             unsigned r;
-            if (!(val>>32)) { r=4; } else { r=0; val>>=32; }
+            if (!(val>>by32)) { r=4; } else { r=0; val>>=by32; }
             if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
             r += (!val);
             return r;
@@ -346,7 +530,7 @@  static unsigned LZ4_NbCommonBytes (register reg_t val)
             _BitScanReverse( &r, (unsigned long)val );
             return (unsigned)(r>>3);
 #       elif (defined(__clang__) || (defined(__GNUC__) && (__GNUC__>=3))) && !defined(LZ4_FORCE_SW_BITCOUNT)
-            return (__builtin_clz((U32)val) >> 3);
+            return (unsigned)__builtin_clz((U32)val) >> 3;
 #       else
             unsigned r;
             if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
@@ -358,11 +542,20 @@  static unsigned LZ4_NbCommonBytes (register reg_t val)
 }
 
 #define STEPSIZE sizeof(reg_t)
-static unsigned LZ4_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* pInLimit)
+LZ4_FORCE_INLINE
+unsigned LZ4_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* pInLimit)
 {
     const BYTE* const pStart = pIn;
 
-    while (likely(pIn<pInLimit-(STEPSIZE-1))) {
+    if (likely(pIn < pInLimit-(STEPSIZE-1))) {
+        reg_t const diff = LZ4_read_ARCH(pMatch) ^ LZ4_read_ARCH(pIn);
+        if (!diff) {
+            pIn+=STEPSIZE; pMatch+=STEPSIZE;
+        } else {
+            return LZ4_NbCommonBytes(diff);
+    }   }
+
+    while (likely(pIn < pInLimit-(STEPSIZE-1))) {
         reg_t const diff = LZ4_read_ARCH(pMatch) ^ LZ4_read_ARCH(pIn);
         if (!diff) { pIn+=STEPSIZE; pMatch+=STEPSIZE; continue; }
         pIn += LZ4_NbCommonBytes(diff);
@@ -387,15 +580,34 @@  static const U32 LZ4_skipTrigger = 6;  /* Increase this value ==> compression ru
 /*-************************************
 *  Local Structures and types
 **************************************/
-typedef enum { notLimited = 0, limitedOutput = 1 } limitedOutput_directive;
-typedef enum { byPtr, byU32, byU16 } tableType_t;
+typedef enum { clearedTable = 0, byPtr, byU32, byU16 } tableType_t;
 
-typedef enum { noDict = 0, withPrefix64k, usingExtDict } dict_directive;
+/**
+ * This enum distinguishes several different modes of accessing previous
+ * content in the stream.
+ *
+ * - noDict        : There is no preceding content.
+ * - withPrefix64k : Table entries up to ctx->dictSize before the current blob
+ *                   blob being compressed are valid and refer to the preceding
+ *                   content (of length ctx->dictSize), which is available
+ *                   contiguously preceding in memory the content currently
+ *                   being compressed.
+ * - usingExtDict  : Like withPrefix64k, but the preceding content is somewhere
+ *                   else in memory, starting at ctx->dictionary with length
+ *                   ctx->dictSize.
+ * - usingDictCtx  : Like usingExtDict, but everything concerning the preceding
+ *                   content is in a separate context, pointed to by
+ *                   ctx->dictCtx. ctx->dictionary, ctx->dictSize, and table
+ *                   entries in the current context that refer to positions
+ *                   preceding the beginning of the current compression are
+ *                   ignored. Instead, ctx->dictCtx->dictionary and ctx->dictCtx
+ *                   ->dictSize describe the location and size of the preceding
+ *                   content, and matches are found by looking in the ctx
+ *                   ->dictCtx->hashTable.
+ */
+typedef enum { noDict = 0, withPrefix64k, usingExtDict, usingDictCtx } dict_directive;
 typedef enum { noDictIssue = 0, dictSmall } dictIssue_directive;
 
-typedef enum { endOnOutputSize = 0, endOnInputSize = 1 } endCondition_directive;
-typedef enum { full = 0, partial = 1 } earlyEnd_directive;
-
 
 /*-************************************
 *  Local Utils
@@ -406,6 +618,23 @@  int LZ4_compressBound(int isize)  { return LZ4_COMPRESSBOUND(isize); }
 int LZ4_sizeofState() { return LZ4_STREAMSIZE; }
 
 
+/*-************************************
+*  Internal Definitions used in Tests
+**************************************/
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+int LZ4_compress_forceExtDict (LZ4_stream_t* LZ4_dict, const char* source, char* dest, int srcSize);
+
+int LZ4_decompress_safe_forceExtDict(const char* source, char* dest,
+                                     int compressedSize, int maxOutputSize,
+                                     const void* dictStart, size_t dictSize);
+
+#if defined (__cplusplus)
+}
+#endif
+
 /*-******************************
 *  Compression functions
 ********************************/
@@ -419,102 +648,225 @@  static U32 LZ4_hash4(U32 sequence, tableType_t const tableType)
 
 static U32 LZ4_hash5(U64 sequence, tableType_t const tableType)
 {
-    static const U64 prime5bytes = 889523592379ULL;
-    static const U64 prime8bytes = 11400714785074694791ULL;
     const U32 hashLog = (tableType == byU16) ? LZ4_HASHLOG+1 : LZ4_HASHLOG;
-    if (LZ4_isLittleEndian())
+    if (LZ4_isLittleEndian()) {
+        const U64 prime5bytes = 889523592379ULL;
         return (U32)(((sequence << 24) * prime5bytes) >> (64 - hashLog));
-    else
+    } else {
+        const U64 prime8bytes = 11400714785074694791ULL;
         return (U32)(((sequence >> 24) * prime8bytes) >> (64 - hashLog));
+    }
 }
 
-FORCE_INLINE U32 LZ4_hashPosition(const void* const p, tableType_t const tableType)
+LZ4_FORCE_INLINE U32 LZ4_hashPosition(const void* const p, tableType_t const tableType)
 {
     if ((sizeof(reg_t)==8) && (tableType != byU16)) return LZ4_hash5(LZ4_read_ARCH(p), tableType);
     return LZ4_hash4(LZ4_read32(p), tableType);
 }
 
-static void LZ4_putPositionOnHash(const BYTE* p, U32 h, void* tableBase, tableType_t const tableType, const BYTE* srcBase)
+static void LZ4_clearHash(U32 h, void* tableBase, tableType_t const tableType)
 {
     switch (tableType)
     {
+    default: /* fallthrough */
+    case clearedTable: { /* illegal! */ assert(0); return; }
+    case byPtr: { const BYTE** hashTable = (const BYTE**)tableBase; hashTable[h] = NULL; return; }
+    case byU32: { U32* hashTable = (U32*) tableBase; hashTable[h] = 0; return; }
+    case byU16: { U16* hashTable = (U16*) tableBase; hashTable[h] = 0; return; }
+    }
+}
+
+static void LZ4_putIndexOnHash(U32 idx, U32 h, void* tableBase, tableType_t const tableType)
+{
+    switch (tableType)
+    {
+    default: /* fallthrough */
+    case clearedTable: /* fallthrough */
+    case byPtr: { /* illegal! */ assert(0); return; }
+    case byU32: { U32* hashTable = (U32*) tableBase; hashTable[h] = idx; return; }
+    case byU16: { U16* hashTable = (U16*) tableBase; assert(idx < 65536); hashTable[h] = (U16)idx; return; }
+    }
+}
+
+static void LZ4_putPositionOnHash(const BYTE* p, U32 h,
+                                  void* tableBase, tableType_t const tableType,
+                            const BYTE* srcBase)
+{
+    switch (tableType)
+    {
+    case clearedTable: { /* illegal! */ assert(0); return; }
     case byPtr: { const BYTE** hashTable = (const BYTE**)tableBase; hashTable[h] = p; return; }
     case byU32: { U32* hashTable = (U32*) tableBase; hashTable[h] = (U32)(p-srcBase); return; }
     case byU16: { U16* hashTable = (U16*) tableBase; hashTable[h] = (U16)(p-srcBase); return; }
     }
 }
 
-FORCE_INLINE void LZ4_putPosition(const BYTE* p, void* tableBase, tableType_t tableType, const BYTE* srcBase)
+LZ4_FORCE_INLINE void LZ4_putPosition(const BYTE* p, void* tableBase, tableType_t tableType, const BYTE* srcBase)
 {
     U32 const h = LZ4_hashPosition(p, tableType);
     LZ4_putPositionOnHash(p, h, tableBase, tableType, srcBase);
 }
 
-static const BYTE* LZ4_getPositionOnHash(U32 h, void* tableBase, tableType_t tableType, const BYTE* srcBase)
+/* LZ4_getIndexOnHash() :
+ * Index of match position registered in hash table.
+ * hash position must be calculated by using base+index, or dictBase+index.
+ * Assumption 1 : only valid if tableType == byU32 or byU16.
+ * Assumption 2 : h is presumed valid (within limits of hash table)
+ */
+static U32 LZ4_getIndexOnHash(U32 h, const void* tableBase, tableType_t tableType)
 {
-    if (tableType == byPtr) { const BYTE** hashTable = (const BYTE**) tableBase; return hashTable[h]; }
-    if (tableType == byU32) { const U32* const hashTable = (U32*) tableBase; return hashTable[h] + srcBase; }
-    { const U16* const hashTable = (U16*) tableBase; return hashTable[h] + srcBase; }   /* default, to ensure a return */
+    LZ4_STATIC_ASSERT(LZ4_MEMORY_USAGE > 2);
+    if (tableType == byU32) {
+        const U32* const hashTable = (const U32*) tableBase;
+        assert(h < (1U << (LZ4_MEMORY_USAGE-2)));
+        return hashTable[h];
+    }
+    if (tableType == byU16) {
+        const U16* const hashTable = (const U16*) tableBase;
+        assert(h < (1U << (LZ4_MEMORY_USAGE-1)));
+        return hashTable[h];
+    }
+    assert(0); return 0;  /* forbidden case */
 }
 
-FORCE_INLINE const BYTE* LZ4_getPosition(const BYTE* p, void* tableBase, tableType_t tableType, const BYTE* srcBase)
+static const BYTE* LZ4_getPositionOnHash(U32 h, const void* tableBase, tableType_t tableType, const BYTE* srcBase)
+{
+    if (tableType == byPtr) { const BYTE* const* hashTable = (const BYTE* const*) tableBase; return hashTable[h]; }
+    if (tableType == byU32) { const U32* const hashTable = (const U32*) tableBase; return hashTable[h] + srcBase; }
+    { const U16* const hashTable = (const U16*) tableBase; return hashTable[h] + srcBase; }   /* default, to ensure a return */
+}
+
+LZ4_FORCE_INLINE const BYTE*
+LZ4_getPosition(const BYTE* p,
+                const void* tableBase, tableType_t tableType,
+                const BYTE* srcBase)
 {
     U32 const h = LZ4_hashPosition(p, tableType);
     return LZ4_getPositionOnHash(h, tableBase, tableType, srcBase);
 }
 
+LZ4_FORCE_INLINE void
+LZ4_prepareTable(LZ4_stream_t_internal* const cctx,
+           const int inputSize,
+           const tableType_t tableType) {
+    /* If compression failed during the previous step, then the context
+     * is marked as dirty, therefore, it has to be fully reset.
+     */
+    if (cctx->dirty) {
+        DEBUGLOG(5, "LZ4_prepareTable: Full reset for %p", cctx);
+        MEM_INIT(cctx, 0, sizeof(LZ4_stream_t_internal));
+        return;
+    }
+
+    /* If the table hasn't been used, it's guaranteed to be zeroed out, and is
+     * therefore safe to use no matter what mode we're in. Otherwise, we figure
+     * out if it's safe to leave as is or whether it needs to be reset.
+     */
+    if (cctx->tableType != clearedTable) {
+        assert(inputSize >= 0);
+        if (cctx->tableType != tableType
+          || ((tableType == byU16) && cctx->currentOffset + (unsigned)inputSize >= 0xFFFFU)
+          || ((tableType == byU32) && cctx->currentOffset > 1 GB)
+          || tableType == byPtr
+          || inputSize >= 4 KB)
+        {
+            DEBUGLOG(4, "LZ4_prepareTable: Resetting table in %p", cctx);
+            MEM_INIT(cctx->hashTable, 0, LZ4_HASHTABLESIZE);
+            cctx->currentOffset = 0;
+            cctx->tableType = clearedTable;
+        } else {
+            DEBUGLOG(4, "LZ4_prepareTable: Re-use hash table (no reset)");
+        }
+    }
+
+    /* Adding a gap, so all previous entries are > LZ4_DISTANCE_MAX back, is faster
+     * than compressing without a gap. However, compressing with
+     * currentOffset == 0 is faster still, so we preserve that case.
+     */
+    if (cctx->currentOffset != 0 && tableType == byU32) {
+        DEBUGLOG(5, "LZ4_prepareTable: adding 64KB to currentOffset");
+        cctx->currentOffset += 64 KB;
+    }
+
+    /* Finally, clear history */
+    cctx->dictCtx = NULL;
+    cctx->dictionary = NULL;
+    cctx->dictSize = 0;
+}
 
 /** LZ4_compress_generic() :
     inlined, to ensure branches are decided at compilation time */
-FORCE_INLINE int LZ4_compress_generic(
+LZ4_FORCE_INLINE int LZ4_compress_generic(
                  LZ4_stream_t_internal* const cctx,
                  const char* const source,
                  char* const dest,
                  const int inputSize,
+                 int *inputConsumed, /* only written when outputDirective == fillOutput */
                  const int maxOutputSize,
-                 const limitedOutput_directive outputLimited,
+                 const limitedOutput_directive outputDirective,
                  const tableType_t tableType,
-                 const dict_directive dict,
+                 const dict_directive dictDirective,
                  const dictIssue_directive dictIssue,
-                 const U32 acceleration)
+                 const int acceleration)
 {
+    int result;
     const BYTE* ip = (const BYTE*) source;
-    const BYTE* base;
+
+    U32 const startIndex = cctx->currentOffset;
+    const BYTE* base = (const BYTE*) source - startIndex;
     const BYTE* lowLimit;
-    const BYTE* const lowRefLimit = ip - cctx->dictSize;
-    const BYTE* const dictionary = cctx->dictionary;
-    const BYTE* const dictEnd = dictionary + cctx->dictSize;
-    const ptrdiff_t dictDelta = dictEnd - (const BYTE*)source;
+
+    const LZ4_stream_t_internal* dictCtx = (const LZ4_stream_t_internal*) cctx->dictCtx;
+    const BYTE* const dictionary =
+        dictDirective == usingDictCtx ? dictCtx->dictionary : cctx->dictionary;
+    const U32 dictSize =
+        dictDirective == usingDictCtx ? dictCtx->dictSize : cctx->dictSize;
+    const U32 dictDelta = (dictDirective == usingDictCtx) ? startIndex - dictCtx->currentOffset : 0;   /* make indexes in dictCtx comparable with index in current context */
+
+    int const maybe_extMem = (dictDirective == usingExtDict) || (dictDirective == usingDictCtx);
+    U32 const prefixIdxLimit = startIndex - dictSize;   /* used when dictDirective == dictSmall */
+    const BYTE* const dictEnd = dictionary + dictSize;
     const BYTE* anchor = (const BYTE*) source;
     const BYTE* const iend = ip + inputSize;
-    const BYTE* const mflimit = iend - MFLIMIT;
+    const BYTE* const mflimitPlusOne = iend - MFLIMIT + 1;
     const BYTE* const matchlimit = iend - LASTLITERALS;
 
+    /* the dictCtx currentOffset is indexed on the start of the dictionary,
+     * while a dictionary in the current context precedes the currentOffset */
+    const BYTE* dictBase = (dictDirective == usingDictCtx) ?
+                            dictionary + dictSize - dictCtx->currentOffset :
+                            dictionary + dictSize - startIndex;
+
     BYTE* op = (BYTE*) dest;
     BYTE* const olimit = op + maxOutputSize;
 
+    U32 offset = 0;
     U32 forwardH;
 
-    /* Init conditions */
-    if ((U32)inputSize > (U32)LZ4_MAX_INPUT_SIZE) return 0;   /* Unsupported inputSize, too large (or negative) */
-    switch(dict)
-    {
-    case noDict:
-    default:
-        base = (const BYTE*)source;
-        lowLimit = (const BYTE*)source;
-        break;
-    case withPrefix64k:
-        base = (const BYTE*)source - cctx->currentOffset;
-        lowLimit = (const BYTE*)source - cctx->dictSize;
-        break;
-    case usingExtDict:
-        base = (const BYTE*)source - cctx->currentOffset;
-        lowLimit = (const BYTE*)source;
-        break;
+    DEBUGLOG(5, "LZ4_compress_generic: srcSize=%i, tableType=%u", inputSize, tableType);
+    /* If init conditions are not met, we don't have to mark stream
+     * as having dirty context, since no action was taken yet */
+    if (outputDirective == fillOutput && maxOutputSize < 1) { return 0; } /* Impossible to store anything */
+    if ((U32)inputSize > (U32)LZ4_MAX_INPUT_SIZE) { return 0; }           /* Unsupported inputSize, too large (or negative) */
+    if ((tableType == byU16) && (inputSize>=LZ4_64Klimit)) { return 0; }  /* Size too large (not within 64K limit) */
+    if (tableType==byPtr) assert(dictDirective==noDict);      /* only supported use case with byPtr */
+    assert(acceleration >= 1);
+
+    lowLimit = (const BYTE*)source - (dictDirective == withPrefix64k ? dictSize : 0);
+
+    /* Update context state */
+    if (dictDirective == usingDictCtx) {
+        /* Subsequent linked blocks can't use the dictionary. */
+        /* Instead, they use the block we just compressed. */
+        cctx->dictCtx = NULL;
+        cctx->dictSize = (U32)inputSize;
+    } else {
+        cctx->dictSize += (U32)inputSize;
     }
-    if ((tableType == byU16) && (inputSize>=LZ4_64Klimit)) return 0;   /* Size too large (not within 64K limit) */
-    if (inputSize<LZ4_minLength) goto _last_literals;                  /* Input too small, no compression (all literals) */
+    cctx->currentOffset += (U32)inputSize;
+    cctx->tableType = (U16)tableType;
+
+    if (inputSize<LZ4_minLength) goto _last_literals;        /* Input too small, no compression (all literals) */
 
     /* First Byte */
     LZ4_putPosition(ip, cctx->hashTable, tableType, base);
@@ -522,50 +874,112 @@  FORCE_INLINE int LZ4_compress_generic(
 
     /* Main Loop */
     for ( ; ; ) {
-        ptrdiff_t refDelta = 0;
         const BYTE* match;
         BYTE* token;
+        const BYTE* filledIp;
 
         /* Find a match */
-        {   const BYTE* forwardIp = ip;
-            unsigned step = 1;
-            unsigned searchMatchNb = acceleration << LZ4_skipTrigger;
+        if (tableType == byPtr) {
+            const BYTE* forwardIp = ip;
+            int step = 1;
+            int searchMatchNb = acceleration << LZ4_skipTrigger;
             do {
                 U32 const h = forwardH;
                 ip = forwardIp;
                 forwardIp += step;
                 step = (searchMatchNb++ >> LZ4_skipTrigger);
 
-                if (unlikely(forwardIp > mflimit)) goto _last_literals;
+                if (unlikely(forwardIp > mflimitPlusOne)) goto _last_literals;
+                assert(ip < mflimitPlusOne);
 
                 match = LZ4_getPositionOnHash(h, cctx->hashTable, tableType, base);
-                if (dict==usingExtDict) {
-                    if (match < (const BYTE*)source) {
-                        refDelta = dictDelta;
-                        lowLimit = dictionary;
-                    } else {
-                        refDelta = 0;
-                        lowLimit = (const BYTE*)source;
-                }   }
                 forwardH = LZ4_hashPosition(forwardIp, tableType);
                 LZ4_putPositionOnHash(ip, h, cctx->hashTable, tableType, base);
 
-            } while ( ((dictIssue==dictSmall) ? (match < lowRefLimit) : 0)
-                || ((tableType==byU16) ? 0 : (match + MAX_DISTANCE < ip))
-                || (LZ4_read32(match+refDelta) != LZ4_read32(ip)) );
+            } while ( (match+LZ4_DISTANCE_MAX < ip)
+                   || (LZ4_read32(match) != LZ4_read32(ip)) );
+
+        } else {   /* byU32, byU16 */
+
+            const BYTE* forwardIp = ip;
+            int step = 1;
+            int searchMatchNb = acceleration << LZ4_skipTrigger;
+            do {
+                U32 const h = forwardH;
+                U32 const current = (U32)(forwardIp - base);
+                U32 matchIndex = LZ4_getIndexOnHash(h, cctx->hashTable, tableType);
+                assert(matchIndex <= current);
+                assert(forwardIp - base < (ptrdiff_t)(2 GB - 1));
+                ip = forwardIp;
+                forwardIp += step;
+                step = (searchMatchNb++ >> LZ4_skipTrigger);
+
+                if (unlikely(forwardIp > mflimitPlusOne)) goto _last_literals;
+                assert(ip < mflimitPlusOne);
+
+                if (dictDirective == usingDictCtx) {
+                    if (matchIndex < startIndex) {
+                        /* there was no match, try the dictionary */
+                        assert(tableType == byU32);
+                        matchIndex = LZ4_getIndexOnHash(h, dictCtx->hashTable, byU32);
+                        match = dictBase + matchIndex;
+                        matchIndex += dictDelta;   /* make dictCtx index comparable with current context */
+                        lowLimit = dictionary;
+                    } else {
+                        match = base + matchIndex;
+                        lowLimit = (const BYTE*)source;
+                    }
+                } else if (dictDirective==usingExtDict) {
+                    if (matchIndex < startIndex) {
+                        DEBUGLOG(7, "extDict candidate: matchIndex=%5u  <  startIndex=%5u", matchIndex, startIndex);
+                        assert(startIndex - matchIndex >= MINMATCH);
+                        match = dictBase + matchIndex;
+                        lowLimit = dictionary;
+                    } else {
+                        match = base + matchIndex;
+                        lowLimit = (const BYTE*)source;
+                    }
+                } else {   /* single continuous memory segment */
+                    match = base + matchIndex;
+                }
+                forwardH = LZ4_hashPosition(forwardIp, tableType);
+                LZ4_putIndexOnHash(current, h, cctx->hashTable, tableType);
+
+                DEBUGLOG(7, "candidate at pos=%u  (offset=%u \n", matchIndex, current - matchIndex);
+                if ((dictIssue == dictSmall) && (matchIndex < prefixIdxLimit)) { continue; }    /* match outside of valid area */
+                assert(matchIndex < current);
+                if ( ((tableType != byU16) || (LZ4_DISTANCE_MAX < LZ4_DISTANCE_ABSOLUTE_MAX))
+                  && (matchIndex+LZ4_DISTANCE_MAX < current)) {
+                    continue;
+                } /* too far */
+                assert((current - matchIndex) <= LZ4_DISTANCE_MAX);  /* match now expected within distance */
+
+                if (LZ4_read32(match) == LZ4_read32(ip)) {
+                    if (maybe_extMem) offset = current - matchIndex;
+                    break;   /* match found */
+                }
+
+            } while(1);
         }
 
         /* Catch up */
-        while (((ip>anchor) & (match+refDelta > lowLimit)) && (unlikely(ip[-1]==match[refDelta-1]))) { ip--; match--; }
+        filledIp = ip;
+        while (((ip>anchor) & (match > lowLimit)) && (unlikely(ip[-1]==match[-1]))) { ip--; match--; }
 
         /* Encode Literals */
         {   unsigned const litLength = (unsigned)(ip - anchor);
             token = op++;
-            if ((outputLimited) &&  /* Check output buffer overflow */
-                (unlikely(op + litLength + (2 + 1 + LASTLITERALS) + (litLength/255) > olimit)))
-                return 0;
+            if ((outputDirective == limitedOutput) &&  /* Check output buffer overflow */
+                (unlikely(op + litLength + (2 + 1 + LASTLITERALS) + (litLength/255) > olimit)) ) {
+                return 0;   /* cannot compress within `dst` budget. Stored indexes in hash table are nonetheless fine */
+            }
+            if ((outputDirective == fillOutput) &&
+                (unlikely(op + (litLength+240)/255 /* litlen */ + litLength /* literals */ + 2 /* offset */ + 1 /* token */ + MFLIMIT - MINMATCH /* min last literals so last match is <= end - MFLIMIT */ > olimit))) {
+                op--;
+                goto _last_literals;
+            }
             if (litLength >= RUN_MASK) {
-                int len = (int)litLength-RUN_MASK;
+                int len = (int)(litLength - RUN_MASK);
                 *token = (RUN_MASK<<ML_BITS);
                 for(; len >= 255 ; len-=255) *op++ = 255;
                 *op++ = (BYTE)len;
@@ -573,82 +987,183 @@  FORCE_INLINE int LZ4_compress_generic(
             else *token = (BYTE)(litLength<<ML_BITS);
 
             /* Copy Literals */
-            LZ4_wildCopy(op, anchor, op+litLength);
+            LZ4_wildCopy8(op, anchor, op+litLength);
             op+=litLength;
+            DEBUGLOG(6, "seq.start:%i, literals=%u, match.start:%i",
+                        (int)(anchor-(const BYTE*)source), litLength, (int)(ip-(const BYTE*)source));
         }
 
 _next_match:
+        /* at this stage, the following variables must be correctly set :
+         * - ip : at start of LZ operation
+         * - match : at start of previous pattern occurence; can be within current prefix, or within extDict
+         * - offset : if maybe_ext_memSegment==1 (constant)
+         * - lowLimit : must be == dictionary to mean "match is within extDict"; must be == source otherwise
+         * - token and *token : position to write 4-bits for match length; higher 4-bits for literal length supposed already written
+         */
+
+        if ((outputDirective == fillOutput) &&
+            (op + 2 /* offset */ + 1 /* token */ + MFLIMIT - MINMATCH /* min last literals so last match is <= end - MFLIMIT */ > olimit)) {
+            /* the match was too close to the end, rewind and go to last literals */
+            op = token;
+            goto _last_literals;
+        }
+
         /* Encode Offset */
-        LZ4_writeLE16(op, (U16)(ip-match)); op+=2;
+        if (maybe_extMem) {   /* static test */
+            DEBUGLOG(6, "             with offset=%u  (ext if > %i)", offset, (int)(ip - (const BYTE*)source));
+            assert(offset <= LZ4_DISTANCE_MAX && offset > 0);
+            LZ4_writeLE16(op, (U16)offset); op+=2;
+        } else  {
+            DEBUGLOG(6, "             with offset=%u  (same segment)", (U32)(ip - match));
+            assert(ip-match <= LZ4_DISTANCE_MAX);
+            LZ4_writeLE16(op, (U16)(ip - match)); op+=2;
+        }
 
         /* Encode MatchLength */
         {   unsigned matchCode;
 
-            if ((dict==usingExtDict) && (lowLimit==dictionary)) {
-                const BYTE* limit;
-                match += refDelta;
-                limit = ip + (dictEnd-match);
+            if ( (dictDirective==usingExtDict || dictDirective==usingDictCtx)
+              && (lowLimit==dictionary) /* match within extDict */ ) {
+                const BYTE* limit = ip + (dictEnd-match);
+                assert(dictEnd > match);
                 if (limit > matchlimit) limit = matchlimit;
                 matchCode = LZ4_count(ip+MINMATCH, match+MINMATCH, limit);
-                ip += MINMATCH + matchCode;
+                ip += (size_t)matchCode + MINMATCH;
                 if (ip==limit) {
-                    unsigned const more = LZ4_count(ip, (const BYTE*)source, matchlimit);
+                    unsigned const more = LZ4_count(limit, (const BYTE*)source, matchlimit);
                     matchCode += more;
                     ip += more;
                 }
+                DEBUGLOG(6, "             with matchLength=%u starting in extDict", matchCode+MINMATCH);
             } else {
                 matchCode = LZ4_count(ip+MINMATCH, match+MINMATCH, matchlimit);
-                ip += MINMATCH + matchCode;
+                ip += (size_t)matchCode + MINMATCH;
+                DEBUGLOG(6, "             with matchLength=%u", matchCode+MINMATCH);
             }
 
-            if ( outputLimited &&    /* Check output buffer overflow */
-                (unlikely(op + (1 + LASTLITERALS) + (matchCode>>8) > olimit)) )
-                return 0;
+            if ((outputDirective) &&    /* Check output buffer overflow */
+                (unlikely(op + (1 + LASTLITERALS) + (matchCode+240)/255 > olimit)) ) {
+                if (outputDirective == fillOutput) {
+                    /* Match description too long : reduce it */
+                    U32 newMatchCode = 15 /* in token */ - 1 /* to avoid needing a zero byte */ + ((U32)(olimit - op) - 1 - LASTLITERALS) * 255;
+                    ip -= matchCode - newMatchCode;
+                    assert(newMatchCode < matchCode);
+                    matchCode = newMatchCode;
+                    if (unlikely(ip <= filledIp)) {
+                        /* We have already filled up to filledIp so if ip ends up less than filledIp
+                         * we have positions in the hash table beyond the current position. This is
+                         * a problem if we reuse the hash table. So we have to remove these positions
+                         * from the hash table.
+                         */
+                        const BYTE* ptr;
+                        DEBUGLOG(5, "Clearing %u positions", (U32)(filledIp - ip));
+                        for (ptr = ip; ptr <= filledIp; ++ptr) {
+                            U32 const h = LZ4_hashPosition(ptr, tableType);
+                            LZ4_clearHash(h, cctx->hashTable, tableType);
+                        }
+                    }
+                } else {
+                    assert(outputDirective == limitedOutput);
+                    return 0;   /* cannot compress within `dst` budget. Stored indexes in hash table are nonetheless fine */
+                }
+            }
             if (matchCode >= ML_MASK) {
                 *token += ML_MASK;
                 matchCode -= ML_MASK;
                 LZ4_write32(op, 0xFFFFFFFF);
-                while (matchCode >= 4*255) op+=4, LZ4_write32(op, 0xFFFFFFFF), matchCode -= 4*255;
+                while (matchCode >= 4*255) {
+                    op+=4;
+                    LZ4_write32(op, 0xFFFFFFFF);
+                    matchCode -= 4*255;
+                }
                 op += matchCode / 255;
                 *op++ = (BYTE)(matchCode % 255);
             } else
                 *token += (BYTE)(matchCode);
         }
+        /* Ensure we have enough space for the last literals. */
+        assert(!(outputDirective == fillOutput && op + 1 + LASTLITERALS > olimit));
 
         anchor = ip;
 
         /* Test end of chunk */
-        if (ip > mflimit) break;
+        if (ip >= mflimitPlusOne) break;
 
         /* Fill table */
         LZ4_putPosition(ip-2, cctx->hashTable, tableType, base);
 
         /* Test next position */
-        match = LZ4_getPosition(ip, cctx->hashTable, tableType, base);
-        if (dict==usingExtDict) {
-            if (match < (const BYTE*)source) {
-                refDelta = dictDelta;
-                lowLimit = dictionary;
-            } else {
-                refDelta = 0;
-                lowLimit = (const BYTE*)source;
-        }   }
-        LZ4_putPosition(ip, cctx->hashTable, tableType, base);
-        if ( ((dictIssue==dictSmall) ? (match>=lowRefLimit) : 1)
-            && (match+MAX_DISTANCE>=ip)
-            && (LZ4_read32(match+refDelta)==LZ4_read32(ip)) )
-        { token=op++; *token=0; goto _next_match; }
+        if (tableType == byPtr) {
+
+            match = LZ4_getPosition(ip, cctx->hashTable, tableType, base);
+            LZ4_putPosition(ip, cctx->hashTable, tableType, base);
+            if ( (match+LZ4_DISTANCE_MAX >= ip)
+              && (LZ4_read32(match) == LZ4_read32(ip)) )
+            { token=op++; *token=0; goto _next_match; }
+
+        } else {   /* byU32, byU16 */
+
+            U32 const h = LZ4_hashPosition(ip, tableType);
+            U32 const current = (U32)(ip-base);
+            U32 matchIndex = LZ4_getIndexOnHash(h, cctx->hashTable, tableType);
+            assert(matchIndex < current);
+            if (dictDirective == usingDictCtx) {
+                if (matchIndex < startIndex) {
+                    /* there was no match, try the dictionary */
+                    matchIndex = LZ4_getIndexOnHash(h, dictCtx->hashTable, byU32);
+                    match = dictBase + matchIndex;
+                    lowLimit = dictionary;   /* required for match length counter */
+                    matchIndex += dictDelta;
+                } else {
+                    match = base + matchIndex;
+                    lowLimit = (const BYTE*)source;  /* required for match length counter */
+                }
+            } else if (dictDirective==usingExtDict) {
+                if (matchIndex < startIndex) {
+                    match = dictBase + matchIndex;
+                    lowLimit = dictionary;   /* required for match length counter */
+                } else {
+                    match = base + matchIndex;
+                    lowLimit = (const BYTE*)source;   /* required for match length counter */
+                }
+            } else {   /* single memory segment */
+                match = base + matchIndex;
+            }
+            LZ4_putIndexOnHash(current, h, cctx->hashTable, tableType);
+            assert(matchIndex < current);
+            if ( ((dictIssue==dictSmall) ? (matchIndex >= prefixIdxLimit) : 1)
+              && (((tableType==byU16) && (LZ4_DISTANCE_MAX == LZ4_DISTANCE_ABSOLUTE_MAX)) ? 1 : (matchIndex+LZ4_DISTANCE_MAX >= current))
+              && (LZ4_read32(match) == LZ4_read32(ip)) ) {
+                token=op++;
+                *token=0;
+                if (maybe_extMem) offset = current - matchIndex;
+                DEBUGLOG(6, "seq.start:%i, literals=%u, match.start:%i",
+                            (int)(anchor-(const BYTE*)source), 0, (int)(ip-(const BYTE*)source));
+                goto _next_match;
+            }
+        }
 
         /* Prepare next loop */
         forwardH = LZ4_hashPosition(++ip, tableType);
+
     }
 
 _last_literals:
     /* Encode Last Literals */
-    {   size_t const lastRun = (size_t)(iend - anchor);
-        if ( (outputLimited) &&  /* Check output buffer overflow */
-            ((op - (BYTE*)dest) + lastRun + 1 + ((lastRun+255-RUN_MASK)/255) > (U32)maxOutputSize) )
-            return 0;
+    {   size_t lastRun = (size_t)(iend - anchor);
+        if ( (outputDirective) &&  /* Check output buffer overflow */
+            (op + lastRun + 1 + ((lastRun+255-RUN_MASK)/255) > olimit)) {
+            if (outputDirective == fillOutput) {
+                /* adapt lastRun to fill 'dst' */
+                assert(olimit >= op);
+                lastRun  = (size_t)(olimit-op) - 1;
+                lastRun -= (lastRun+240)/255;
+            } else {
+                assert(outputDirective == limitedOutput);
+                return 0;   /* cannot compress within `dst` budget. Stored indexes in hash table are nonetheless fine */
+            }
+        }
         if (lastRun >= RUN_MASK) {
             size_t accumulator = lastRun - RUN_MASK;
             *op++ = RUN_MASK << ML_BITS;
@@ -658,251 +1173,154 @@  _last_literals:
             *op++ = (BYTE)(lastRun<<ML_BITS);
         }
         memcpy(op, anchor, lastRun);
+        ip = anchor + lastRun;
         op += lastRun;
     }
 
-    /* End */
-    return (int) (((char*)op)-dest);
+    if (outputDirective == fillOutput) {
+        *inputConsumed = (int) (((const char*)ip)-source);
+    }
+    DEBUGLOG(5, "LZ4_compress_generic: compressed %i bytes into %i bytes", inputSize, (int)(((char*)op) - dest));
+    result = (int)(((char*)op) - dest);
+    assert(result > 0);
+    return result;
 }
 
 
 int LZ4_compress_fast_extState(void* state, const char* source, char* dest, int inputSize, int maxOutputSize, int acceleration)
 {
-    LZ4_stream_t_internal* ctx = &((LZ4_stream_t*)state)->internal_donotuse;
-    LZ4_resetStream((LZ4_stream_t*)state);
+    LZ4_stream_t_internal* const ctx = & LZ4_initStream(state, sizeof(LZ4_stream_t)) -> internal_donotuse;
+    assert(ctx != NULL);
     if (acceleration < 1) acceleration = ACCELERATION_DEFAULT;
-
     if (maxOutputSize >= LZ4_compressBound(inputSize)) {
-        if (inputSize < LZ4_64Klimit)
-            return LZ4_compress_generic(ctx, source, dest, inputSize,             0,    notLimited,                        byU16, noDict, noDictIssue, acceleration);
-        else
-            return LZ4_compress_generic(ctx, source, dest, inputSize,             0,    notLimited, (sizeof(void*)==8) ? byU32 : byPtr, noDict, noDictIssue, acceleration);
+        if (inputSize < LZ4_64Klimit) {
+            return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, 0, notLimited, byU16, noDict, noDictIssue, acceleration);
+        } else {
+            const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)source > LZ4_DISTANCE_MAX)) ? byPtr : byU32;
+            return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, 0, notLimited, tableType, noDict, noDictIssue, acceleration);
+        }
     } else {
-        if (inputSize < LZ4_64Klimit)
-            return LZ4_compress_generic(ctx, source, dest, inputSize, maxOutputSize, limitedOutput,                        byU16, noDict, noDictIssue, acceleration);
-        else
-            return LZ4_compress_generic(ctx, source, dest, inputSize, maxOutputSize, limitedOutput, (sizeof(void*)==8) ? byU32 : byPtr, noDict, noDictIssue, acceleration);
+        if (inputSize < LZ4_64Klimit) {
+            return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, byU16, noDict, noDictIssue, acceleration);
+        } else {
+            const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)source > LZ4_DISTANCE_MAX)) ? byPtr : byU32;
+            return LZ4_compress_generic(ctx, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, noDict, noDictIssue, acceleration);
+        }
+    }
+}
+
+/**
+ * LZ4_compress_fast_extState_fastReset() :
+ * A variant of LZ4_compress_fast_extState().
+ *
+ * Using this variant avoids an expensive initialization step. It is only safe
+ * to call if the state buffer is known to be correctly initialized already
+ * (see comment in lz4.h on LZ4_resetStream_fast() for a definition of
+ * "correctly initialized").
+ */
+int LZ4_compress_fast_extState_fastReset(void* state, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration)
+{
+    LZ4_stream_t_internal* ctx = &((LZ4_stream_t*)state)->internal_donotuse;
+    if (acceleration < 1) acceleration = ACCELERATION_DEFAULT;
+
+    if (dstCapacity >= LZ4_compressBound(srcSize)) {
+        if (srcSize < LZ4_64Klimit) {
+            const tableType_t tableType = byU16;
+            LZ4_prepareTable(ctx, srcSize, tableType);
+            if (ctx->currentOffset) {
+                return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, 0, notLimited, tableType, noDict, dictSmall, acceleration);
+            } else {
+                return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, 0, notLimited, tableType, noDict, noDictIssue, acceleration);
+            }
+        } else {
+            const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)src > LZ4_DISTANCE_MAX)) ? byPtr : byU32;
+            LZ4_prepareTable(ctx, srcSize, tableType);
+            return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, 0, notLimited, tableType, noDict, noDictIssue, acceleration);
+        }
+    } else {
+        if (srcSize < LZ4_64Klimit) {
+            const tableType_t tableType = byU16;
+            LZ4_prepareTable(ctx, srcSize, tableType);
+            if (ctx->currentOffset) {
+                return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, dstCapacity, limitedOutput, tableType, noDict, dictSmall, acceleration);
+            } else {
+                return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, dstCapacity, limitedOutput, tableType, noDict, noDictIssue, acceleration);
+            }
+        } else {
+            const tableType_t tableType = ((sizeof(void*)==4) && ((uptrval)src > LZ4_DISTANCE_MAX)) ? byPtr : byU32;
+            LZ4_prepareTable(ctx, srcSize, tableType);
+            return LZ4_compress_generic(ctx, src, dst, srcSize, NULL, dstCapacity, limitedOutput, tableType, noDict, noDictIssue, acceleration);
+        }
     }
 }
 
 
 int LZ4_compress_fast(const char* source, char* dest, int inputSize, int maxOutputSize, int acceleration)
 {
-#if (HEAPMODE)
-    void* ctxPtr = ALLOCATOR(1, sizeof(LZ4_stream_t));   /* malloc-calloc always properly aligned */
+    int result;
+#if (LZ4_HEAPMODE)
+    LZ4_stream_t* ctxPtr = ALLOC(sizeof(LZ4_stream_t));   /* malloc-calloc always properly aligned */
+    if (ctxPtr == NULL) return 0;
 #else
     LZ4_stream_t ctx;
-    void* const ctxPtr = &ctx;
+    LZ4_stream_t* const ctxPtr = &ctx;
 #endif
+    result = LZ4_compress_fast_extState(ctxPtr, source, dest, inputSize, maxOutputSize, acceleration);
 
-    int const result = LZ4_compress_fast_extState(ctxPtr, source, dest, inputSize, maxOutputSize, acceleration);
-
-#if (HEAPMODE)
+#if (LZ4_HEAPMODE)
     FREEMEM(ctxPtr);
 #endif
     return result;
 }
 
 
-int LZ4_compress_default(const char* source, char* dest, int inputSize, int maxOutputSize)
+int LZ4_compress_default(const char* src, char* dst, int srcSize, int maxOutputSize)
 {
-    return LZ4_compress_fast(source, dest, inputSize, maxOutputSize, 1);
+    return LZ4_compress_fast(src, dst, srcSize, maxOutputSize, 1);
 }
 
 
 /* hidden debug function */
 /* strangely enough, gcc generates faster code when this function is uncommented, even if unused */
-int LZ4_compress_fast_force(const char* source, char* dest, int inputSize, int maxOutputSize, int acceleration)
+int LZ4_compress_fast_force(const char* src, char* dst, int srcSize, int dstCapacity, int acceleration)
 {
     LZ4_stream_t ctx;
-    LZ4_resetStream(&ctx);
+    LZ4_initStream(&ctx, sizeof(ctx));
 
-    if (inputSize < LZ4_64Klimit)
-        return LZ4_compress_generic(&ctx.internal_donotuse, source, dest, inputSize, maxOutputSize, limitedOutput, byU16,                        noDict, noDictIssue, acceleration);
-    else
-        return LZ4_compress_generic(&ctx.internal_donotuse, source, dest, inputSize, maxOutputSize, limitedOutput, sizeof(void*)==8 ? byU32 : byPtr, noDict, noDictIssue, acceleration);
-}
-
-
-/*-******************************
-*  *_destSize() variant
-********************************/
-
-static int LZ4_compress_destSize_generic(
-                       LZ4_stream_t_internal* const ctx,
-                 const char* const src,
-                       char* const dst,
-                       int*  const srcSizePtr,
-                 const int targetDstSize,
-                 const tableType_t tableType)
-{
-    const BYTE* ip = (const BYTE*) src;
-    const BYTE* base = (const BYTE*) src;
-    const BYTE* lowLimit = (const BYTE*) src;
-    const BYTE* anchor = ip;
-    const BYTE* const iend = ip + *srcSizePtr;
-    const BYTE* const mflimit = iend - MFLIMIT;
-    const BYTE* const matchlimit = iend - LASTLITERALS;
-
-    BYTE* op = (BYTE*) dst;
-    BYTE* const oend = op + targetDstSize;
-    BYTE* const oMaxLit = op + targetDstSize - 2 /* offset */ - 8 /* because 8+MINMATCH==MFLIMIT */ - 1 /* token */;
-    BYTE* const oMaxMatch = op + targetDstSize - (LASTLITERALS + 1 /* token */);
-    BYTE* const oMaxSeq = oMaxLit - 1 /* token */;
-
-    U32 forwardH;
-
-
-    /* Init conditions */
-    if (targetDstSize < 1) return 0;                                     /* Impossible to store anything */
-    if ((U32)*srcSizePtr > (U32)LZ4_MAX_INPUT_SIZE) return 0;            /* Unsupported input size, too large (or negative) */
-    if ((tableType == byU16) && (*srcSizePtr>=LZ4_64Klimit)) return 0;   /* Size too large (not within 64K limit) */
-    if (*srcSizePtr<LZ4_minLength) goto _last_literals;                  /* Input too small, no compression (all literals) */
-
-    /* First Byte */
-    *srcSizePtr = 0;
-    LZ4_putPosition(ip, ctx->hashTable, tableType, base);
-    ip++; forwardH = LZ4_hashPosition(ip, tableType);
-
-    /* Main Loop */
-    for ( ; ; ) {
-        const BYTE* match;
-        BYTE* token;
-
-        /* Find a match */
-        {   const BYTE* forwardIp = ip;
-            unsigned step = 1;
-            unsigned searchMatchNb = 1 << LZ4_skipTrigger;
-
-            do {
-                U32 h = forwardH;
-                ip = forwardIp;
-                forwardIp += step;
-                step = (searchMatchNb++ >> LZ4_skipTrigger);
-
-                if (unlikely(forwardIp > mflimit)) goto _last_literals;
-
-                match = LZ4_getPositionOnHash(h, ctx->hashTable, tableType, base);
-                forwardH = LZ4_hashPosition(forwardIp, tableType);
-                LZ4_putPositionOnHash(ip, h, ctx->hashTable, tableType, base);
-
-            } while ( ((tableType==byU16) ? 0 : (match + MAX_DISTANCE < ip))
-                || (LZ4_read32(match) != LZ4_read32(ip)) );
-        }
-
-        /* Catch up */
-        while ((ip>anchor) && (match > lowLimit) && (unlikely(ip[-1]==match[-1]))) { ip--; match--; }
-
-        /* Encode Literal length */
-        {   unsigned litLength = (unsigned)(ip - anchor);
-            token = op++;
-            if (op + ((litLength+240)/255) + litLength > oMaxLit) {
-                /* Not enough space for a last match */
-                op--;
-                goto _last_literals;
-            }
-            if (litLength>=RUN_MASK) {
-                unsigned len = litLength - RUN_MASK;
-                *token=(RUN_MASK<<ML_BITS);
-                for(; len >= 255 ; len-=255) *op++ = 255;
-                *op++ = (BYTE)len;
-            }
-            else *token = (BYTE)(litLength<<ML_BITS);
-
-            /* Copy Literals */
-            LZ4_wildCopy(op, anchor, op+litLength);
-            op += litLength;
-        }
-
-_next_match:
-        /* Encode Offset */
-        LZ4_writeLE16(op, (U16)(ip-match)); op+=2;
-
-        /* Encode MatchLength */
-        {   size_t matchLength = LZ4_count(ip+MINMATCH, match+MINMATCH, matchlimit);
-
-            if (op + ((matchLength+240)/255) > oMaxMatch) {
-                /* Match description too long : reduce it */
-                matchLength = (15-1) + (oMaxMatch-op) * 255;
-            }
-            ip += MINMATCH + matchLength;
-
-            if (matchLength>=ML_MASK) {
-                *token += ML_MASK;
-                matchLength -= ML_MASK;
-                while (matchLength >= 255) { matchLength-=255; *op++ = 255; }
-                *op++ = (BYTE)matchLength;
-            }
-            else *token += (BYTE)(matchLength);
-        }
-
-        anchor = ip;
-
-        /* Test end of block */
-        if (ip > mflimit) break;
-        if (op > oMaxSeq) break;
-
-        /* Fill table */
-        LZ4_putPosition(ip-2, ctx->hashTable, tableType, base);
-
-        /* Test next position */
-        match = LZ4_getPosition(ip, ctx->hashTable, tableType, base);
-        LZ4_putPosition(ip, ctx->hashTable, tableType, base);
-        if ( (match+MAX_DISTANCE>=ip)
-            && (LZ4_read32(match)==LZ4_read32(ip)) )
-        { token=op++; *token=0; goto _next_match; }
-
-        /* Prepare next loop */
-        forwardH = LZ4_hashPosition(++ip, tableType);
+    if (srcSize < LZ4_64Klimit) {
+        return LZ4_compress_generic(&ctx.internal_donotuse, src, dst, srcSize, NULL, dstCapacity, limitedOutput, byU16,    noDict, noDictIssue, acceleration);
+    } else {
+        tableType_t const addrMode = (sizeof(void*) > 4) ? byU32 : byPtr;
+        return LZ4_compress_generic(&ctx.internal_donotuse, src, dst, srcSize, NULL, dstCapacity, limitedOutput, addrMode, noDict, noDictIssue, acceleration);
     }
-
-_last_literals:
-    /* Encode Last Literals */
-    {   size_t lastRunSize = (size_t)(iend - anchor);
-        if (op + 1 /* token */ + ((lastRunSize+240)/255) /* litLength */ + lastRunSize /* literals */ > oend) {
-            /* adapt lastRunSize to fill 'dst' */
-            lastRunSize  = (oend-op) - 1;
-            lastRunSize -= (lastRunSize+240)/255;
-        }
-        ip = anchor + lastRunSize;
-
-        if (lastRunSize >= RUN_MASK) {
-            size_t accumulator = lastRunSize - RUN_MASK;
-            *op++ = RUN_MASK << ML_BITS;
-            for(; accumulator >= 255 ; accumulator-=255) *op++ = 255;
-            *op++ = (BYTE) accumulator;
-        } else {
-            *op++ = (BYTE)(lastRunSize<<ML_BITS);
-        }
-        memcpy(op, anchor, lastRunSize);
-        op += lastRunSize;
-    }
-
-    /* End */
-    *srcSizePtr = (int) (((const char*)ip)-src);
-    return (int) (((char*)op)-dst);
 }
 
 
+/* Note!: This function leaves the stream in an unclean/broken state!
+ * It is not safe to subsequently use the same state with a _fastReset() or
+ * _continue() call without resetting it. */
 static int LZ4_compress_destSize_extState (LZ4_stream_t* state, const char* src, char* dst, int* srcSizePtr, int targetDstSize)
 {
-    LZ4_resetStream(state);
+    void* const s = LZ4_initStream(state, sizeof (*state));
+    assert(s != NULL); (void)s;
 
     if (targetDstSize >= LZ4_compressBound(*srcSizePtr)) {  /* compression success is guaranteed */
         return LZ4_compress_fast_extState(state, src, dst, *srcSizePtr, targetDstSize, 1);
     } else {
-        if (*srcSizePtr < LZ4_64Klimit)
-            return LZ4_compress_destSize_generic(&state->internal_donotuse, src, dst, srcSizePtr, targetDstSize, byU16);
-        else
-            return LZ4_compress_destSize_generic(&state->internal_donotuse, src, dst, srcSizePtr, targetDstSize, sizeof(void*)==8 ? byU32 : byPtr);
-    }
+        if (*srcSizePtr < LZ4_64Klimit) {
+            return LZ4_compress_generic(&state->internal_donotuse, src, dst, *srcSizePtr, srcSizePtr, targetDstSize, fillOutput, byU16, noDict, noDictIssue, 1);
+        } else {
+            tableType_t const addrMode = ((sizeof(void*)==4) && ((uptrval)src > LZ4_DISTANCE_MAX)) ? byPtr : byU32;
+            return LZ4_compress_generic(&state->internal_donotuse, src, dst, *srcSizePtr, srcSizePtr, targetDstSize, fillOutput, addrMode, noDict, noDictIssue, 1);
+    }   }
 }
 
 
 int LZ4_compress_destSize(const char* src, char* dst, int* srcSizePtr, int targetDstSize)
 {
-#if (HEAPMODE)
-    LZ4_stream_t* ctx = (LZ4_stream_t*)ALLOCATOR(1, sizeof(LZ4_stream_t));   /* malloc-calloc always properly aligned */
+#if (LZ4_HEAPMODE)
+    LZ4_stream_t* ctx = (LZ4_stream_t*)ALLOC(sizeof(LZ4_stream_t));   /* malloc-calloc always properly aligned */
+    if (ctx == NULL) return 0;
 #else
     LZ4_stream_t ctxBody;
     LZ4_stream_t* ctx = &ctxBody;
@@ -910,7 +1328,7 @@  int LZ4_compress_destSize(const char* src, char* dst, int* srcSizePtr, int targe
 
     int result = LZ4_compress_destSize_extState(ctx, src, dst, srcSizePtr, targetDstSize);
 
-#if (HEAPMODE)
+#if (LZ4_HEAPMODE)
     FREEMEM(ctx);
 #endif
     return result;
@@ -924,19 +1342,54 @@  int LZ4_compress_destSize(const char* src, char* dst, int* srcSizePtr, int targe
 
 LZ4_stream_t* LZ4_createStream(void)
 {
-    LZ4_stream_t* lz4s = (LZ4_stream_t*)ALLOCATOR(8, LZ4_STREAMSIZE_U64);
+    LZ4_stream_t* const lz4s = (LZ4_stream_t*)ALLOC(sizeof(LZ4_stream_t));
     LZ4_STATIC_ASSERT(LZ4_STREAMSIZE >= sizeof(LZ4_stream_t_internal));    /* A compilation error here means LZ4_STREAMSIZE is not large enough */
-    LZ4_resetStream(lz4s);
+    DEBUGLOG(4, "LZ4_createStream %p", lz4s);
+    if (lz4s == NULL) return NULL;
+    LZ4_initStream(lz4s, sizeof(*lz4s));
     return lz4s;
 }
 
+#ifndef _MSC_VER  /* for some reason, Visual fails the aligment test on 32-bit x86 :
+                     it reports an aligment of 8-bytes,
+                     while actually aligning LZ4_stream_t on 4 bytes. */
+static size_t LZ4_stream_t_alignment(void)
+{
+    struct { char c; LZ4_stream_t t; } t_a;
+    return sizeof(t_a) - sizeof(t_a.t);
+}
+#endif
+
+LZ4_stream_t* LZ4_initStream (void* buffer, size_t size)
+{
+    DEBUGLOG(5, "LZ4_initStream");
+    if (buffer == NULL) { return NULL; }
+    if (size < sizeof(LZ4_stream_t)) { return NULL; }
+#ifndef _MSC_VER  /* for some reason, Visual fails the aligment test on 32-bit x86 :
+                     it reports an aligment of 8-bytes,
+                     while actually aligning LZ4_stream_t on 4 bytes. */
+    if (((size_t)buffer) & (LZ4_stream_t_alignment() - 1)) { return NULL; } /* alignment check */
+#endif
+    MEM_INIT(buffer, 0, sizeof(LZ4_stream_t));
+    return (LZ4_stream_t*)buffer;
+}
+
+/* resetStream is now deprecated,
+ * prefer initStream() which is more general */
 void LZ4_resetStream (LZ4_stream_t* LZ4_stream)
 {
+    DEBUGLOG(5, "LZ4_resetStream (ctx:%p)", LZ4_stream);
     MEM_INIT(LZ4_stream, 0, sizeof(LZ4_stream_t));
 }
 
+void LZ4_resetStream_fast(LZ4_stream_t* ctx) {
+    LZ4_prepareTable(&(ctx->internal_donotuse), 0, byU32);
+}
+
 int LZ4_freeStream (LZ4_stream_t* LZ4_stream)
 {
+    if (!LZ4_stream) return 0;   /* support free on NULL */
+    DEBUGLOG(5, "LZ4_freeStream %p", LZ4_stream);
     FREEMEM(LZ4_stream);
     return (0);
 }
@@ -946,43 +1399,88 @@  int LZ4_freeStream (LZ4_stream_t* LZ4_stream)
 int LZ4_loadDict (LZ4_stream_t* LZ4_dict, const char* dictionary, int dictSize)
 {
     LZ4_stream_t_internal* dict = &LZ4_dict->internal_donotuse;
+    const tableType_t tableType = byU32;
     const BYTE* p = (const BYTE*)dictionary;
     const BYTE* const dictEnd = p + dictSize;
     const BYTE* base;
 
-    if ((dict->initCheck) || (dict->currentOffset > 1 GB))  /* Uninitialized structure, or reuse overflow */
-        LZ4_resetStream(LZ4_dict);
+    DEBUGLOG(4, "LZ4_loadDict (%i bytes from %p into %p)", dictSize, dictionary, LZ4_dict);
+
+    /* It's necessary to reset the context,
+     * and not just continue it with prepareTable()
+     * to avoid any risk of generating overflowing matchIndex
+     * when compressing using this dictionary */
+    LZ4_resetStream(LZ4_dict);
+
+    /* We always increment the offset by 64 KB, since, if the dict is longer,
+     * we truncate it to the last 64k, and if it's shorter, we still want to
+     * advance by a whole window length so we can provide the guarantee that
+     * there are only valid offsets in the window, which allows an optimization
+     * in LZ4_compress_fast_continue() where it uses noDictIssue even when the
+     * dictionary isn't a full 64k. */
+    dict->currentOffset += 64 KB;
 
     if (dictSize < (int)HASH_UNIT) {
-        dict->dictionary = NULL;
-        dict->dictSize = 0;
         return 0;
     }
 
     if ((dictEnd - p) > 64 KB) p = dictEnd - 64 KB;
-    dict->currentOffset += 64 KB;
-    base = p - dict->currentOffset;
+    base = dictEnd - dict->currentOffset;
     dict->dictionary = p;
     dict->dictSize = (U32)(dictEnd - p);
-    dict->currentOffset += dict->dictSize;
+    dict->tableType = tableType;
 
     while (p <= dictEnd-HASH_UNIT) {
-        LZ4_putPosition(p, dict->hashTable, byU32, base);
+        LZ4_putPosition(p, dict->hashTable, tableType, base);
         p+=3;
     }
 
-    return dict->dictSize;
+    return (int)dict->dictSize;
 }
 
+void LZ4_attach_dictionary(LZ4_stream_t* workingStream, const LZ4_stream_t* dictionaryStream) {
+    const LZ4_stream_t_internal* dictCtx = dictionaryStream == NULL ? NULL :
+        &(dictionaryStream->internal_donotuse);
 
-static void LZ4_renormDictT(LZ4_stream_t_internal* LZ4_dict, const BYTE* src)
+    DEBUGLOG(4, "LZ4_attach_dictionary (%p, %p, size %u)",
+             workingStream, dictionaryStream,
+             dictCtx != NULL ? dictCtx->dictSize : 0);
+
+    /* Calling LZ4_resetStream_fast() here makes sure that changes will not be
+     * erased by subsequent calls to LZ4_resetStream_fast() in case stream was
+     * marked as having dirty context, e.g. requiring full reset.
+     */
+    LZ4_resetStream_fast(workingStream);
+
+    if (dictCtx != NULL) {
+        /* If the current offset is zero, we will never look in the
+         * external dictionary context, since there is no value a table
+         * entry can take that indicate a miss. In that case, we need
+         * to bump the offset to something non-zero.
+         */
+        if (workingStream->internal_donotuse.currentOffset == 0) {
+            workingStream->internal_donotuse.currentOffset = 64 KB;
+        }
+
+        /* Don't actually attach an empty dictionary.
+         */
+        if (dictCtx->dictSize == 0) {
+            dictCtx = NULL;
+        }
+    }
+    workingStream->internal_donotuse.dictCtx = dictCtx;
+}
+
+
+static void LZ4_renormDictT(LZ4_stream_t_internal* LZ4_dict, int nextSize)
 {
-    if ((LZ4_dict->currentOffset > 0x80000000) ||
-        ((uptrval)LZ4_dict->currentOffset > (uptrval)src)) {   /* address space overflow */
+    assert(nextSize >= 0);
+    if (LZ4_dict->currentOffset + (unsigned)nextSize > 0x80000000) {   /* potential ptrdiff_t overflow (32-bits mode) */
         /* rescale hash table */
         U32 const delta = LZ4_dict->currentOffset - 64 KB;
         const BYTE* dictEnd = LZ4_dict->dictionary + LZ4_dict->dictSize;
         int i;
+        DEBUGLOG(4, "LZ4_renormDictT");
         for (i=0; i<LZ4_HASH_SIZE_U32; i++) {
             if (LZ4_dict->hashTable[i] < delta) LZ4_dict->hashTable[i]=0;
             else LZ4_dict->hashTable[i] -= delta;
@@ -994,17 +1492,30 @@  static void LZ4_renormDictT(LZ4_stream_t_internal* LZ4_dict, const BYTE* src)
 }
 
 
-int LZ4_compress_fast_continue (LZ4_stream_t* LZ4_stream, const char* source, char* dest, int inputSize, int maxOutputSize, int acceleration)
+int LZ4_compress_fast_continue (LZ4_stream_t* LZ4_stream,
+                                const char* source, char* dest,
+                                int inputSize, int maxOutputSize,
+                                int acceleration)
 {
+    const tableType_t tableType = byU32;
     LZ4_stream_t_internal* streamPtr = &LZ4_stream->internal_donotuse;
-    const BYTE* const dictEnd = streamPtr->dictionary + streamPtr->dictSize;
+    const BYTE* dictEnd = streamPtr->dictionary + streamPtr->dictSize;
 
-    const BYTE* smallest = (const BYTE*) source;
-    if (streamPtr->initCheck) return 0;   /* Uninitialized structure detected */
-    if ((streamPtr->dictSize>0) && (smallest>dictEnd)) smallest = dictEnd;
-    LZ4_renormDictT(streamPtr, smallest);
+    DEBUGLOG(5, "LZ4_compress_fast_continue (inputSize=%i)", inputSize);
+
+    if (streamPtr->dirty) { return 0; } /* Uninitialized structure detected */
+    LZ4_renormDictT(streamPtr, inputSize);   /* avoid index overflow */
     if (acceleration < 1) acceleration = ACCELERATION_DEFAULT;
 
+    /* invalidate tiny dictionaries */
+    if ( (streamPtr->dictSize-1 < 4-1)   /* intentional underflow */
+      && (dictEnd != (const BYTE*)source) ) {
+        DEBUGLOG(5, "LZ4_compress_fast_continue: dictSize(%u) at addr:%p is too small", streamPtr->dictSize, streamPtr->dictionary);
+        streamPtr->dictSize = 0;
+        streamPtr->dictionary = (const BYTE*)source;
+        dictEnd = (const BYTE*)source;
+    }
+
     /* Check overlapping input/dictionary space */
     {   const BYTE* sourceEnd = (const BYTE*) source + inputSize;
         if ((sourceEnd > streamPtr->dictionary) && (sourceEnd < dictEnd)) {
@@ -1017,46 +1528,61 @@  int LZ4_compress_fast_continue (LZ4_stream_t* LZ4_stream, const char* source, ch
 
     /* prefix mode : source data follows dictionary */
     if (dictEnd == (const BYTE*)source) {
-        int result;
         if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset))
-            result = LZ4_compress_generic(streamPtr, source, dest, inputSize, maxOutputSize, limitedOutput, byU32, withPrefix64k, dictSmall, acceleration);
+            return LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, withPrefix64k, dictSmall, acceleration);
         else
-            result = LZ4_compress_generic(streamPtr, source, dest, inputSize, maxOutputSize, limitedOutput, byU32, withPrefix64k, noDictIssue, acceleration);
-        streamPtr->dictSize += (U32)inputSize;
-        streamPtr->currentOffset += (U32)inputSize;
-        return result;
+            return LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, withPrefix64k, noDictIssue, acceleration);
     }
 
     /* external dictionary mode */
     {   int result;
-        if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset))
-            result = LZ4_compress_generic(streamPtr, source, dest, inputSize, maxOutputSize, limitedOutput, byU32, usingExtDict, dictSmall, acceleration);
-        else
-            result = LZ4_compress_generic(streamPtr, source, dest, inputSize, maxOutputSize, limitedOutput, byU32, usingExtDict, noDictIssue, acceleration);
+        if (streamPtr->dictCtx) {
+            /* We depend here on the fact that dictCtx'es (produced by
+             * LZ4_loadDict) guarantee that their tables contain no references
+             * to offsets between dictCtx->currentOffset - 64 KB and
+             * dictCtx->currentOffset - dictCtx->dictSize. This makes it safe
+             * to use noDictIssue even when the dict isn't a full 64 KB.
+             */
+            if (inputSize > 4 KB) {
+                /* For compressing large blobs, it is faster to pay the setup
+                 * cost to copy the dictionary's tables into the active context,
+                 * so that the compression loop is only looking into one table.
+                 */
+                memcpy(streamPtr, streamPtr->dictCtx, sizeof(LZ4_stream_t));
+                result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingExtDict, noDictIssue, acceleration);
+            } else {
+                result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingDictCtx, noDictIssue, acceleration);
+            }
+        } else {
+            if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset)) {
+                result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingExtDict, dictSmall, acceleration);
+            } else {
+                result = LZ4_compress_generic(streamPtr, source, dest, inputSize, NULL, maxOutputSize, limitedOutput, tableType, usingExtDict, noDictIssue, acceleration);
+            }
+        }
         streamPtr->dictionary = (const BYTE*)source;
         streamPtr->dictSize = (U32)inputSize;
-        streamPtr->currentOffset += (U32)inputSize;
         return result;
     }
 }
 
 
-/* Hidden debug function, to force external dictionary mode */
-int LZ4_compress_forceExtDict (LZ4_stream_t* LZ4_dict, const char* source, char* dest, int inputSize)
+/* Hidden debug function, to force-test external dictionary mode */
+int LZ4_compress_forceExtDict (LZ4_stream_t* LZ4_dict, const char* source, char* dest, int srcSize)
 {
     LZ4_stream_t_internal* streamPtr = &LZ4_dict->internal_donotuse;
     int result;
-    const BYTE* const dictEnd = streamPtr->dictionary + streamPtr->dictSize;
 
-    const BYTE* smallest = dictEnd;
-    if (smallest > (const BYTE*) source) smallest = (const BYTE*) source;
-    LZ4_renormDictT(streamPtr, smallest);
+    LZ4_renormDictT(streamPtr, srcSize);
 
-    result = LZ4_compress_generic(streamPtr, source, dest, inputSize, 0, notLimited, byU32, usingExtDict, noDictIssue, 1);
+    if ((streamPtr->dictSize < 64 KB) && (streamPtr->dictSize < streamPtr->currentOffset)) {
+        result = LZ4_compress_generic(streamPtr, source, dest, srcSize, NULL, 0, notLimited, byU32, usingExtDict, dictSmall, 1);
+    } else {
+        result = LZ4_compress_generic(streamPtr, source, dest, srcSize, NULL, 0, notLimited, byU32, usingExtDict, noDictIssue, 1);
+    }
 
     streamPtr->dictionary = (const BYTE*)source;
-    streamPtr->dictSize = (U32)inputSize;
-    streamPtr->currentOffset += (U32)inputSize;
+    streamPtr->dictSize = (U32)srcSize;
 
     return result;
 }
@@ -1074,8 +1600,8 @@  int LZ4_saveDict (LZ4_stream_t* LZ4_dict, char* safeBuffer, int dictSize)
     LZ4_stream_t_internal* const dict = &LZ4_dict->internal_donotuse;
     const BYTE* const previousDictEnd = dict->dictionary + dict->dictSize;
 
-    if ((U32)dictSize > 64 KB) dictSize = 64 KB;   /* useless to define a dictionary > 64 KB */
-    if ((U32)dictSize > dict->dictSize) dictSize = dict->dictSize;
+    if ((U32)dictSize > 64 KB) { dictSize = 64 KB; } /* useless to define a dictionary > 64 KB */
+    if ((U32)dictSize > dict->dictSize) { dictSize = (int)dict->dictSize; }
 
     memmove(safeBuffer, previousDictEnd - dictSize, dictSize);
 
@@ -1087,218 +1613,587 @@  int LZ4_saveDict (LZ4_stream_t* LZ4_dict, char* safeBuffer, int dictSize)
 
 
 
-/*-*****************************
-*  Decompression functions
-*******************************/
+/*-*******************************
+ *  Decompression functions
+ ********************************/
+
+typedef enum { endOnOutputSize = 0, endOnInputSize = 1 } endCondition_directive;
+typedef enum { decode_full_block = 0, partial_decode = 1 } earlyEnd_directive;
+
+#undef MIN
+#define MIN(a,b)    ( (a) < (b) ? (a) : (b) )
+
+/* Read the variable-length literal or match length.
+ *
+ * ip - pointer to use as input.
+ * lencheck - end ip.  Return an error if ip advances >= lencheck.
+ * loop_check - check ip >= lencheck in body of loop.  Returns loop_error if so.
+ * initial_check - check ip >= lencheck before start of loop.  Returns initial_error if so.
+ * error (output) - error code.  Should be set to 0 before call.
+ */
+typedef enum { loop_error = -2, initial_error = -1, ok = 0 } variable_length_error;
+LZ4_FORCE_INLINE unsigned
+read_variable_length(const BYTE**ip, const BYTE* lencheck, int loop_check, int initial_check, variable_length_error* error)
+{
+  unsigned length = 0;
+  unsigned s;
+  if (initial_check && unlikely((*ip) >= lencheck)) {    /* overflow detection */
+    *error = initial_error;
+    return length;
+  }
+  do {
+    s = **ip;
+    (*ip)++;
+    length += s;
+    if (loop_check && unlikely((*ip) >= lencheck)) {    /* overflow detection */
+      *error = loop_error;
+      return length;
+    }
+  } while (s==255);
+
+  return length;
+}
+
 /*! LZ4_decompress_generic() :
- *  This generic decompression function cover all use cases.
- *  It shall be instantiated several times, using different sets of directives
- *  Note that it is important this generic function is really inlined,
+ *  This generic decompression function covers all use cases.
+ *  It shall be instantiated several times, using different sets of directives.
+ *  Note that it is important for performance that this function really get inlined,
  *  in order to remove useless branches during compilation optimization.
  */
-FORCE_INLINE int LZ4_decompress_generic(
-                 const char* const source,
-                 char* const dest,
-                 int inputSize,
-                 int outputSize,         /* If endOnInput==endOnInputSize, this value is the max size of Output Buffer. */
+LZ4_FORCE_INLINE int
+LZ4_decompress_generic(
+                 const char* const src,
+                 char* const dst,
+                 int srcSize,
+                 int outputSize,         /* If endOnInput==endOnInputSize, this value is `dstCapacity` */
 
-                 int endOnInput,         /* endOnOutputSize, endOnInputSize */
-                 int partialDecoding,    /* full, partial */
-                 int targetOutputSize,   /* only used if partialDecoding==partial */
-                 int dict,               /* noDict, withPrefix64k, usingExtDict */
-                 const BYTE* const lowPrefix,  /* == dest when no prefix */
+                 endCondition_directive endOnInput,   /* endOnOutputSize, endOnInputSize */
+                 earlyEnd_directive partialDecoding,  /* full, partial */
+                 dict_directive dict,                 /* noDict, withPrefix64k, usingExtDict */
+                 const BYTE* const lowPrefix,  /* always <= dst, == dst when no prefix */
                  const BYTE* const dictStart,  /* only if dict==usingExtDict */
                  const size_t dictSize         /* note : = 0 if noDict */
                  )
 {
-    /* Local Variables */
-    const BYTE* ip = (const BYTE*) source;
-    const BYTE* const iend = ip + inputSize;
+    if (src == NULL) { return -1; }
 
-    BYTE* op = (BYTE*) dest;
-    BYTE* const oend = op + outputSize;
-    BYTE* cpy;
-    BYTE* oexit = op + targetOutputSize;
-    const BYTE* const lowLimit = lowPrefix - dictSize;
+    {   const BYTE* ip = (const BYTE*) src;
+        const BYTE* const iend = ip + srcSize;
 
-    const BYTE* const dictEnd = (const BYTE*)dictStart + dictSize;
-    const unsigned dec32table[] = {0, 1, 2, 1, 4, 4, 4, 4};
-    const int dec64table[] = {0, 0, 0, -1, 0, 1, 2, 3};
+        BYTE* op = (BYTE*) dst;
+        BYTE* const oend = op + outputSize;
+        BYTE* cpy;
 
-    const int safeDecode = (endOnInput==endOnInputSize);
-    const int checkOffset = ((safeDecode) && (dictSize < (int)(64 KB)));
+        const BYTE* const dictEnd = (dictStart == NULL) ? NULL : dictStart + dictSize;
 
+        const int safeDecode = (endOnInput==endOnInputSize);
+        const int checkOffset = ((safeDecode) && (dictSize < (int)(64 KB)));
 
-    /* Special cases */
-    if ((partialDecoding) && (oexit > oend-MFLIMIT)) oexit = oend-MFLIMIT;                        /* targetOutputSize too high => decode everything */
-    if ((endOnInput) && (unlikely(outputSize==0))) return ((inputSize==1) && (*ip==0)) ? 0 : -1;  /* Empty output buffer */
-    if ((!endOnInput) && (unlikely(outputSize==0))) return (*ip==0?1:-1);
 
-    /* Main Loop : decode sequences */
-    while (1) {
-        size_t length;
+        /* Set up the "end" pointers for the shortcut. */
+        const BYTE* const shortiend = iend - (endOnInput ? 14 : 8) /*maxLL*/ - 2 /*offset*/;
+        const BYTE* const shortoend = oend - (endOnInput ? 14 : 8) /*maxLL*/ - 18 /*maxML*/;
+
         const BYTE* match;
         size_t offset;
+        unsigned token;
+        size_t length;
 
-        /* get literal length */
-        unsigned const token = *ip++;
-        if ((length=(token>>ML_BITS)) == RUN_MASK) {
-            unsigned s;
-            do {
-                s = *ip++;
-                length += s;
-            } while ( likely(endOnInput ? ip<iend-RUN_MASK : 1) & (s==255) );
-            if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)(op))) goto _output_error;   /* overflow detection */
-            if ((safeDecode) && unlikely((uptrval)(ip)+length<(uptrval)(ip))) goto _output_error;   /* overflow detection */
+
+        DEBUGLOG(5, "LZ4_decompress_generic (srcSize:%i, dstSize:%i)", srcSize, outputSize);
+
+        /* Special cases */
+        assert(lowPrefix <= op);
+        if ((endOnInput) && (unlikely(outputSize==0))) {
+            /* Empty output buffer */
+            if (partialDecoding) return 0;
+            return ((srcSize==1) && (*ip==0)) ? 0 : -1;
+        }
+        if ((!endOnInput) && (unlikely(outputSize==0))) { return (*ip==0 ? 1 : -1); }
+        if ((endOnInput) && unlikely(srcSize==0)) { return -1; }
+
+	/* Currently the fast loop shows a regression on qualcomm arm chips. */
+#if LZ4_FAST_DEC_LOOP
+        if ((oend - op) < FASTLOOP_SAFE_DISTANCE) {
+            DEBUGLOG(6, "skip fast decode loop");
+            goto safe_decode;
         }
 
-        /* copy literals */
-        cpy = op+length;
-        if ( ((endOnInput) && ((cpy>(partialDecoding?oexit:oend-MFLIMIT)) || (ip+length>iend-(2+1+LASTLITERALS))) )
-            || ((!endOnInput) && (cpy>oend-WILDCOPYLENGTH)) )
-        {
-            if (partialDecoding) {
-                if (cpy > oend) goto _output_error;                           /* Error : write attempt beyond end of output buffer */
-                if ((endOnInput) && (ip+length > iend)) goto _output_error;   /* Error : read attempt beyond end of input buffer */
+        /* Fast loop : decode sequences as long as output < iend-FASTLOOP_SAFE_DISTANCE */
+        while (1) {
+            /* Main fastloop assertion: We can always wildcopy FASTLOOP_SAFE_DISTANCE */
+            assert(oend - op >= FASTLOOP_SAFE_DISTANCE);
+            if (endOnInput) { assert(ip < iend); }
+            token = *ip++;
+            length = token >> ML_BITS;  /* literal length */
+
+            assert(!endOnInput || ip <= iend); /* ip < iend before the increment */
+
+            /* decode literal length */
+            if (length == RUN_MASK) {
+                variable_length_error error = ok;
+                length += read_variable_length(&ip, iend-RUN_MASK, endOnInput, endOnInput, &error);
+                if (error == initial_error) { goto _output_error; }
+                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)(op))) { goto _output_error; } /* overflow detection */
+                if ((safeDecode) && unlikely((uptrval)(ip)+length<(uptrval)(ip))) { goto _output_error; } /* overflow detection */
+
+                /* copy literals */
+                cpy = op+length;
+                LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
+                if (endOnInput) {  /* LZ4_decompress_safe() */
+                    if ((cpy>oend-32) || (ip+length>iend-32)) { goto safe_literal_copy; }
+                    LZ4_wildCopy32(op, ip, cpy);
+                } else {   /* LZ4_decompress_fast() */
+                    if (cpy>oend-8) { goto safe_literal_copy; }
+                    LZ4_wildCopy8(op, ip, cpy); /* LZ4_decompress_fast() cannot copy more than 8 bytes at a time :
+                                                 * it doesn't know input length, and only relies on end-of-block properties */
+                }
+                ip += length; op = cpy;
             } else {
-                if ((!endOnInput) && (cpy != oend)) goto _output_error;       /* Error : block decoding must stop exactly there */
-                if ((endOnInput) && ((ip+length != iend) || (cpy > oend))) goto _output_error;   /* Error : input must be consumed */
+                cpy = op+length;
+                if (endOnInput) {  /* LZ4_decompress_safe() */
+                    DEBUGLOG(7, "copy %u bytes in a 16-bytes stripe", (unsigned)length);
+                    /* We don't need to check oend, since we check it once for each loop below */
+                    if (ip > iend-(16 + 1/*max lit + offset + nextToken*/)) { goto safe_literal_copy; }
+                    /* Literals can only be 14, but hope compilers optimize if we copy by a register size */
+                    memcpy(op, ip, 16);
+                } else {  /* LZ4_decompress_fast() */
+                    /* LZ4_decompress_fast() cannot copy more than 8 bytes at a time :
+                     * it doesn't know input length, and relies on end-of-block properties */
+                    memcpy(op, ip, 8);
+                    if (length > 8) { memcpy(op+8, ip+8, 8); }
+                }
+                ip += length; op = cpy;
             }
-            memcpy(op, ip, length);
-            ip += length;
-            op += length;
-            break;     /* Necessarily EOF, due to parsing restrictions */
-        }
-        LZ4_wildCopy(op, ip, cpy);
-        ip += length; op = cpy;
-
-        /* get offset */
-        offset = LZ4_readLE16(ip); ip+=2;
-        match = op - offset;
-        if ((checkOffset) && (unlikely(match < lowLimit))) goto _output_error;   /* Error : offset outside buffers */
-        LZ4_write32(op, (U32)offset);   /* costs ~1%; silence an msan warning when offset==0 */
-
-        /* get matchlength */
-        length = token & ML_MASK;
-        if (length == ML_MASK) {
-            unsigned s;
-            do {
-                s = *ip++;
-                if ((endOnInput) && (ip > iend-LASTLITERALS)) goto _output_error;
-                length += s;
-            } while (s==255);
-            if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)op)) goto _output_error;   /* overflow detection */
+
+            /* get offset */
+            offset = LZ4_readLE16(ip); ip+=2;
+            match = op - offset;
+            assert(match <= op);
+
+            /* get matchlength */
+            length = token & ML_MASK;
+
+            if (length == ML_MASK) {
+              variable_length_error error = ok;
+              if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) { goto _output_error; } /* Error : offset outside buffers */
+              length += read_variable_length(&ip, iend - LASTLITERALS + 1, endOnInput, 0, &error);
+              if (error != ok) { goto _output_error; }
+                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)op)) { goto _output_error; } /* overflow detection */
+                length += MINMATCH;
+                if (op + length >= oend - FASTLOOP_SAFE_DISTANCE) {
+                    goto safe_match_copy;
+                }
+            } else {
+                length += MINMATCH;
+                if (op + length >= oend - FASTLOOP_SAFE_DISTANCE) {
+                    goto safe_match_copy;
+                }
+
+                /* Fastpath check: Avoids a branch in LZ4_wildCopy32 if true */
+                if ((dict == withPrefix64k) || (match >= lowPrefix)) {
+                    if (offset >= 8) {
+                        assert(match >= lowPrefix);
+                        assert(match <= op);
+                        assert(op + 18 <= oend);
+
+                        memcpy(op, match, 8);
+                        memcpy(op+8, match+8, 8);
+                        memcpy(op+16, match+16, 2);
+                        op += length;
+                        continue;
+            }   }   }
+
+            if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) { goto _output_error; } /* Error : offset outside buffers */
+            /* match starting within external dictionary */
+            if ((dict==usingExtDict) && (match < lowPrefix)) {
+                if (unlikely(op+length > oend-LASTLITERALS)) {
+                    if (partialDecoding) {
+                        length = MIN(length, (size_t)(oend-op));  /* reach end of buffer */
+                    } else {
+                        goto _output_error;  /* end-of-block condition violated */
+                }   }
+
+                if (length <= (size_t)(lowPrefix-match)) {
+                    /* match fits entirely within external dictionary : just copy */
+                    memmove(op, dictEnd - (lowPrefix-match), length);
+                    op += length;
+                } else {
+                    /* match stretches into both external dictionary and current block */
+                    size_t const copySize = (size_t)(lowPrefix - match);
+                    size_t const restSize = length - copySize;
+                    memcpy(op, dictEnd - copySize, copySize);
+                    op += copySize;
+                    if (restSize > (size_t)(op - lowPrefix)) {  /* overlap copy */
+                        BYTE* const endOfMatch = op + restSize;
+                        const BYTE* copyFrom = lowPrefix;
+                        while (op < endOfMatch) { *op++ = *copyFrom++; }
+                    } else {
+                        memcpy(op, lowPrefix, restSize);
+                        op += restSize;
+                }   }
+                continue;
+            }
+
+            /* copy match within block */
+            cpy = op + length;
+
+            assert((op <= oend) && (oend-op >= 32));
+            if (unlikely(offset<16)) {
+                LZ4_memcpy_using_offset(op, match, cpy, offset);
+            } else {
+                LZ4_wildCopy32(op, match, cpy);
+            }
+
+            op = cpy;   /* wildcopy correction */
         }
-        length += MINMATCH;
+    safe_decode:
+#endif
+
+        /* Main Loop : decode remaining sequences where output < FASTLOOP_SAFE_DISTANCE */
+        while (1) {
+            token = *ip++;
+            length = token >> ML_BITS;  /* literal length */
 
-        /* check external dictionary */
-        if ((dict==usingExtDict) && (match < lowPrefix)) {
-            if (unlikely(op+length > oend-LASTLITERALS)) goto _output_error;   /* doesn't respect parsing restriction */
+            assert(!endOnInput || ip <= iend); /* ip < iend before the increment */
+
+            /* A two-stage shortcut for the most common case:
+             * 1) If the literal length is 0..14, and there is enough space,
+             * enter the shortcut and copy 16 bytes on behalf of the literals
+             * (in the fast mode, only 8 bytes can be safely copied this way).
+             * 2) Further if the match length is 4..18, copy 18 bytes in a similar
+             * manner; but we ensure that there's enough space in the output for
+             * those 18 bytes earlier, upon entering the shortcut (in other words,
+             * there is a combined check for both stages).
+             */
+            if ( (endOnInput ? length != RUN_MASK : length <= 8)
+                /* strictly "less than" on input, to re-enter the loop with at least one byte */
+              && likely((endOnInput ? ip < shortiend : 1) & (op <= shortoend)) ) {
+                /* Copy the literals */
+                memcpy(op, ip, endOnInput ? 16 : 8);
+                op += length; ip += length;
+
+                /* The second stage: prepare for match copying, decode full info.
+                 * If it doesn't work out, the info won't be wasted. */
+                length = token & ML_MASK; /* match length */
+                offset = LZ4_readLE16(ip); ip += 2;
+                match = op - offset;
+                assert(match <= op); /* check overflow */
+
+                /* Do not deal with overlapping matches. */
+                if ( (length != ML_MASK)
+                  && (offset >= 8)
+                  && (dict==withPrefix64k || match >= lowPrefix) ) {
+                    /* Copy the match. */
+                    memcpy(op + 0, match + 0, 8);
+                    memcpy(op + 8, match + 8, 8);
+                    memcpy(op +16, match +16, 2);
+                    op += length + MINMATCH;
+                    /* Both stages worked, load the next token. */
+                    continue;
+                }
+
+                /* The second stage didn't work out, but the info is ready.
+                 * Propel it right to the point of match copying. */
+                goto _copy_match;
+            }
+
+            /* decode literal length */
+            if (length == RUN_MASK) {
+                variable_length_error error = ok;
+                length += read_variable_length(&ip, iend-RUN_MASK, endOnInput, endOnInput, &error);
+                if (error == initial_error) { goto _output_error; }
+                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)(op))) { goto _output_error; } /* overflow detection */
+                if ((safeDecode) && unlikely((uptrval)(ip)+length<(uptrval)(ip))) { goto _output_error; } /* overflow detection */
+            }
 
-            if (length <= (size_t)(lowPrefix-match)) {
-                /* match can be copied as a single segment from external dictionary */
-                memmove(op, dictEnd - (lowPrefix-match), length);
+            /* copy literals */
+            cpy = op+length;
+#if LZ4_FAST_DEC_LOOP
+        safe_literal_copy:
+#endif
+            LZ4_STATIC_ASSERT(MFLIMIT >= WILDCOPYLENGTH);
+            if ( ((endOnInput) && ((cpy>oend-MFLIMIT) || (ip+length>iend-(2+1+LASTLITERALS))) )
+              || ((!endOnInput) && (cpy>oend-WILDCOPYLENGTH)) )
+            {
+                /* We've either hit the input parsing restriction or the output parsing restriction.
+                 * If we've hit the input parsing condition then this must be the last sequence.
+                 * If we've hit the output parsing condition then we are either using partialDecoding
+                 * or we've hit the output parsing condition.
+                 */
+                if (partialDecoding) {
+                    /* Since we are partial decoding we may be in this block because of the output parsing
+                     * restriction, which is not valid since the output buffer is allowed to be undersized.
+                     */
+                    assert(endOnInput);
+                    /* If we're in this block because of the input parsing condition, then we must be on the
+                     * last sequence (or invalid), so we must check that we exactly consume the input.
+                     */
+                    if ((ip+length>iend-(2+1+LASTLITERALS)) && (ip+length != iend)) { goto _output_error; }
+                    assert(ip+length <= iend);
+                    /* We are finishing in the middle of a literals segment.
+                     * Break after the copy.
+                     */
+                    if (cpy > oend) {
+                        cpy = oend;
+                        assert(op<=oend);
+                        length = (size_t)(oend-op);
+                    }
+                    assert(ip+length <= iend);
+                } else {
+                    /* We must be on the last sequence because of the parsing limitations so check
+                     * that we exactly regenerate the original size (must be exact when !endOnInput).
+                     */
+                    if ((!endOnInput) && (cpy != oend)) { goto _output_error; }
+                     /* We must be on the last sequence (or invalid) because of the parsing limitations
+                      * so check that we exactly consume the input and don't overrun the output buffer.
+                      */
+                    if ((endOnInput) && ((ip+length != iend) || (cpy > oend))) { goto _output_error; }
+                }
+                memmove(op, ip, length);  /* supports overlapping memory regions, which only matters for in-place decompression scenarios */
+                ip += length;
                 op += length;
+                /* Necessarily EOF when !partialDecoding. When partialDecoding
+                 * it is EOF if we've either filled the output buffer or hit
+                 * the input parsing restriction.
+                 */
+                if (!partialDecoding || (cpy == oend) || (ip == iend)) {
+                    break;
+                }
             } else {
-                /* match encompass external dictionary and current block */
-                size_t const copySize = (size_t)(lowPrefix-match);
-                size_t const restSize = length - copySize;
-                memcpy(op, dictEnd - copySize, copySize);
-                op += copySize;
-                if (restSize > (size_t)(op-lowPrefix)) {  /* overlap copy */
-                    BYTE* const endOfMatch = op + restSize;
-                    const BYTE* copyFrom = lowPrefix;
-                    while (op < endOfMatch) *op++ = *copyFrom++;
+                LZ4_wildCopy8(op, ip, cpy);   /* may overwrite up to WILDCOPYLENGTH beyond cpy */
+                ip += length; op = cpy;
+            }
+
+            /* get offset */
+            offset = LZ4_readLE16(ip); ip+=2;
+            match = op - offset;
+
+            /* get matchlength */
+            length = token & ML_MASK;
+
+    _copy_match:
+            if (length == ML_MASK) {
+              variable_length_error error = ok;
+              length += read_variable_length(&ip, iend - LASTLITERALS + 1, endOnInput, 0, &error);
+              if (error != ok) goto _output_error;
+                if ((safeDecode) && unlikely((uptrval)(op)+length<(uptrval)op)) goto _output_error;   /* overflow detection */
+            }
+            length += MINMATCH;
+
+#if LZ4_FAST_DEC_LOOP
+        safe_match_copy:
+#endif
+            if ((checkOffset) && (unlikely(match + dictSize < lowPrefix))) goto _output_error;   /* Error : offset outside buffers */
+            /* match starting within external dictionary */
+            if ((dict==usingExtDict) && (match < lowPrefix)) {
+                if (unlikely(op+length > oend-LASTLITERALS)) {
+                    if (partialDecoding) length = MIN(length, (size_t)(oend-op));
+                    else goto _output_error;   /* doesn't respect parsing restriction */
+                }
+
+                if (length <= (size_t)(lowPrefix-match)) {
+                    /* match fits entirely within external dictionary : just copy */
+                    memmove(op, dictEnd - (lowPrefix-match), length);
+                    op += length;
                 } else {
-                    memcpy(op, lowPrefix, restSize);
-                    op += restSize;
-            }   }
-            continue;
-        }
+                    /* match stretches into both external dictionary and current block */
+                    size_t const copySize = (size_t)(lowPrefix - match);
+                    size_t const restSize = length - copySize;
+                    memcpy(op, dictEnd - copySize, copySize);
+                    op += copySize;
+                    if (restSize > (size_t)(op - lowPrefix)) {  /* overlap copy */
+                        BYTE* const endOfMatch = op + restSize;
+                        const BYTE* copyFrom = lowPrefix;
+                        while (op < endOfMatch) *op++ = *copyFrom++;
+                    } else {
+                        memcpy(op, lowPrefix, restSize);
+                        op += restSize;
+                }   }
+                continue;
+            }
+            assert(match >= lowPrefix);
+
+            /* copy match within block */
+            cpy = op + length;
 
-        /* copy match within block */
-        cpy = op + length;
-        if (unlikely(offset<8)) {
-            const int dec64 = dec64table[offset];
-            op[0] = match[0];
-            op[1] = match[1];
-            op[2] = match[2];
-            op[3] = match[3];
-            match += dec32table[offset];
-            memcpy(op+4, match, 4);
-            match -= dec64;
-        } else { LZ4_copy8(op, match); match+=8; }
-        op += 8;
-
-        if (unlikely(cpy>oend-12)) {
-            BYTE* const oCopyLimit = oend-(WILDCOPYLENGTH-1);
-            if (cpy > oend-LASTLITERALS) goto _output_error;    /* Error : last LASTLITERALS bytes must be literals (uncompressed) */
-            if (op < oCopyLimit) {
-                LZ4_wildCopy(op, match, oCopyLimit);
-                match += oCopyLimit - op;
-                op = oCopyLimit;
+            /* partialDecoding : may end anywhere within the block */
+            assert(op<=oend);
+            if (partialDecoding && (cpy > oend-MATCH_SAFEGUARD_DISTANCE)) {
+                size_t const mlen = MIN(length, (size_t)(oend-op));
+                const BYTE* const matchEnd = match + mlen;
+                BYTE* const copyEnd = op + mlen;
+                if (matchEnd > op) {   /* overlap copy */
+                    while (op < copyEnd) { *op++ = *match++; }
+                } else {
+                    memcpy(op, match, mlen);
+                }
+                op = copyEnd;
+                if (op == oend) { break; }
+                continue;
             }
-            while (op<cpy) *op++ = *match++;
-        } else {
-            LZ4_copy8(op, match);
-            if (length>16) LZ4_wildCopy(op+8, match+8, cpy);
+
+            if (unlikely(offset<8)) {
+                LZ4_write32(op, 0);   /* silence msan warning when offset==0 */
+                op[0] = match[0];
+                op[1] = match[1];
+                op[2] = match[2];
+                op[3] = match[3];
+                match += inc32table[offset];
+                memcpy(op+4, match, 4);
+                match -= dec64table[offset];
+            } else {
+                memcpy(op, match, 8);
+                match += 8;
+            }
+            op += 8;
+
+            if (unlikely(cpy > oend-MATCH_SAFEGUARD_DISTANCE)) {
+                BYTE* const oCopyLimit = oend - (WILDCOPYLENGTH-1);
+                if (cpy > oend-LASTLITERALS) { goto _output_error; } /* Error : last LASTLITERALS bytes must be literals (uncompressed) */
+                if (op < oCopyLimit) {
+                    LZ4_wildCopy8(op, match, oCopyLimit);
+                    match += oCopyLimit - op;
+                    op = oCopyLimit;
+                }
+                while (op < cpy) { *op++ = *match++; }
+            } else {
+                memcpy(op, match, 8);
+                if (length > 16)  { LZ4_wildCopy8(op+8, match+8, cpy); }
+            }
+            op = cpy;   /* wildcopy correction */
         }
-        op=cpy;   /* correction */
+
+        /* end of decoding */
+        if (endOnInput) {
+           return (int) (((char*)op)-dst);     /* Nb of output bytes decoded */
+       } else {
+           return (int) (((const char*)ip)-src);   /* Nb of input bytes read */
+       }
+
+        /* Overflow error detected */
+    _output_error:
+        return (int) (-(((const char*)ip)-src))-1;
     }
-
-    /* end of decoding */
-    if (endOnInput)
-       return (int) (((char*)op)-dest);     /* Nb of output bytes decoded */
-    else
-       return (int) (((const char*)ip)-source);   /* Nb of input bytes read */
-
-    /* Overflow error detected */
-_output_error:
-    return (int) (-(((const char*)ip)-source))-1;
 }
 
 
+/*===== Instantiate the API decoding functions. =====*/
+
+LZ4_FORCE_O2_GCC_PPC64LE
 int LZ4_decompress_safe(const char* source, char* dest, int compressedSize, int maxDecompressedSize)
 {
-    return LZ4_decompress_generic(source, dest, compressedSize, maxDecompressedSize, endOnInputSize, full, 0, noDict, (BYTE*)dest, NULL, 0);
+    return LZ4_decompress_generic(source, dest, compressedSize, maxDecompressedSize,
+                                  endOnInputSize, decode_full_block, noDict,
+                                  (BYTE*)dest, NULL, 0);
 }
 
-int LZ4_decompress_safe_partial(const char* source, char* dest, int compressedSize, int targetOutputSize, int maxDecompressedSize)
+LZ4_FORCE_O2_GCC_PPC64LE
+int LZ4_decompress_safe_partial(const char* src, char* dst, int compressedSize, int targetOutputSize, int dstCapacity)
 {
-    return LZ4_decompress_generic(source, dest, compressedSize, maxDecompressedSize, endOnInputSize, partial, targetOutputSize, noDict, (BYTE*)dest, NULL, 0);
+    dstCapacity = MIN(targetOutputSize, dstCapacity);
+    return LZ4_decompress_generic(src, dst, compressedSize, dstCapacity,
+                                  endOnInputSize, partial_decode,
+                                  noDict, (BYTE*)dst, NULL, 0);
 }
 
+LZ4_FORCE_O2_GCC_PPC64LE
 int LZ4_decompress_fast(const char* source, char* dest, int originalSize)
 {
-    return LZ4_decompress_generic(source, dest, 0, originalSize, endOnOutputSize, full, 0, withPrefix64k, (BYTE*)(dest - 64 KB), NULL, 64 KB);
+    return LZ4_decompress_generic(source, dest, 0, originalSize,
+                                  endOnOutputSize, decode_full_block, withPrefix64k,
+                                  (BYTE*)dest - 64 KB, NULL, 0);
 }
 
+/*===== Instantiate a few more decoding cases, used more than once. =====*/
+
+LZ4_FORCE_O2_GCC_PPC64LE /* Exported, an obsolete API function. */
+int LZ4_decompress_safe_withPrefix64k(const char* source, char* dest, int compressedSize, int maxOutputSize)
+{
+    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize,
+                                  endOnInputSize, decode_full_block, withPrefix64k,
+                                  (BYTE*)dest - 64 KB, NULL, 0);
+}
+
+/* Another obsolete API function, paired with the previous one. */
+int LZ4_decompress_fast_withPrefix64k(const char* source, char* dest, int originalSize)
+{
+    /* LZ4_decompress_fast doesn't validate match offsets,
+     * and thus serves well with any prefixed dictionary. */
+    return LZ4_decompress_fast(source, dest, originalSize);
+}
+
+LZ4_FORCE_O2_GCC_PPC64LE
+static int LZ4_decompress_safe_withSmallPrefix(const char* source, char* dest, int compressedSize, int maxOutputSize,
+                                               size_t prefixSize)
+{
+    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize,
+                                  endOnInputSize, decode_full_block, noDict,
+                                  (BYTE*)dest-prefixSize, NULL, 0);
+}
+
+LZ4_FORCE_O2_GCC_PPC64LE
+int LZ4_decompress_safe_forceExtDict(const char* source, char* dest,
+                                     int compressedSize, int maxOutputSize,
+                                     const void* dictStart, size_t dictSize)
+{
+    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize,
+                                  endOnInputSize, decode_full_block, usingExtDict,
+                                  (BYTE*)dest, (const BYTE*)dictStart, dictSize);
+}
+
+LZ4_FORCE_O2_GCC_PPC64LE
+static int LZ4_decompress_fast_extDict(const char* source, char* dest, int originalSize,
+                                       const void* dictStart, size_t dictSize)
+{
+    return LZ4_decompress_generic(source, dest, 0, originalSize,
+                                  endOnOutputSize, decode_full_block, usingExtDict,
+                                  (BYTE*)dest, (const BYTE*)dictStart, dictSize);
+}
+
+/* The "double dictionary" mode, for use with e.g. ring buffers: the first part
+ * of the dictionary is passed as prefix, and the second via dictStart + dictSize.
+ * These routines are used only once, in LZ4_decompress_*_continue().
+ */
+LZ4_FORCE_INLINE
+int LZ4_decompress_safe_doubleDict(const char* source, char* dest, int compressedSize, int maxOutputSize,
+                                   size_t prefixSize, const void* dictStart, size_t dictSize)
+{
+    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize,
+                                  endOnInputSize, decode_full_block, usingExtDict,
+                                  (BYTE*)dest-prefixSize, (const BYTE*)dictStart, dictSize);
+}
+
+LZ4_FORCE_INLINE
+int LZ4_decompress_fast_doubleDict(const char* source, char* dest, int originalSize,
+                                   size_t prefixSize, const void* dictStart, size_t dictSize)
+{
+    return LZ4_decompress_generic(source, dest, 0, originalSize,
+                                  endOnOutputSize, decode_full_block, usingExtDict,
+                                  (BYTE*)dest-prefixSize, (const BYTE*)dictStart, dictSize);
+}
 
 /*===== streaming decompression functions =====*/
 
-/*
- * If you prefer dynamic allocation methods,
- * LZ4_createStreamDecode()
- * provides a pointer (void*) towards an initialized LZ4_streamDecode_t structure.
- */
 LZ4_streamDecode_t* LZ4_createStreamDecode(void)
 {
-    LZ4_streamDecode_t* lz4s = (LZ4_streamDecode_t*) ALLOCATOR(1, sizeof(LZ4_streamDecode_t));
+    LZ4_streamDecode_t* lz4s = (LZ4_streamDecode_t*) ALLOC_AND_ZERO(sizeof(LZ4_streamDecode_t));
+    LZ4_STATIC_ASSERT(LZ4_STREAMDECODESIZE >= sizeof(LZ4_streamDecode_t_internal));    /* A compilation error here means LZ4_STREAMDECODESIZE is not large enough */
     return lz4s;
 }
 
 int LZ4_freeStreamDecode (LZ4_streamDecode_t* LZ4_stream)
 {
+    if (LZ4_stream == NULL) { return 0; }  /* support free on NULL */
     FREEMEM(LZ4_stream);
     return 0;
 }
 
-/*!
- * LZ4_setStreamDecode() :
- * Use this function to instruct where to find the dictionary.
- * This function is not necessary if previous data is still available where it was decoded.
- * Loading a size of 0 is allowed (same effect as no dictionary).
- * Return : 1 if OK, 0 if error
+/*! LZ4_setStreamDecode() :
+ *  Use this function to instruct where to find the dictionary.
+ *  This function is not necessary if previous data is still available where it was decoded.
+ *  Loading a size of 0 is allowed (same effect as no dictionary).
+ * @return : 1 if OK, 0 if error
  */
 int LZ4_setStreamDecode (LZ4_streamDecode_t* LZ4_streamDecode, const char* dictionary, int dictSize)
 {
@@ -1310,6 +2205,25 @@  int LZ4_setStreamDecode (LZ4_streamDecode_t* LZ4_streamDecode, const char* dicti
     return 1;
 }
 
+/*! LZ4_decoderRingBufferSize() :
+ *  when setting a ring buffer for streaming decompression (optional scenario),
+ *  provides the minimum size of this ring buffer
+ *  to be compatible with any source respecting maxBlockSize condition.
+ *  Note : in a ring buffer scenario,
+ *  blocks are presumed decompressed next to each other.
+ *  When not enough space remains for next block (remainingSize < maxBlockSize),
+ *  decoding resumes from beginning of ring buffer.
+ * @return : minimum ring buffer size,
+ *           or 0 if there is an error (invalid maxBlockSize).
+ */
+int LZ4_decoderRingBufferSize(int maxBlockSize)
+{
+    if (maxBlockSize < 0) return 0;
+    if (maxBlockSize > LZ4_MAX_INPUT_SIZE) return 0;
+    if (maxBlockSize < 16) maxBlockSize = 16;
+    return LZ4_DECODER_RING_BUFFER_SIZE(maxBlockSize);
+}
+
 /*
 *_continue() :
     These decoding functions allow decompression of multiple blocks in "streaming" mode.
@@ -1317,52 +2231,75 @@  int LZ4_setStreamDecode (LZ4_streamDecode_t* LZ4_streamDecode, const char* dicti
     If it's not possible, save the relevant part of decoded data into a safe buffer,
     and indicate where it stands using LZ4_setStreamDecode()
 */
+LZ4_FORCE_O2_GCC_PPC64LE
 int LZ4_decompress_safe_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* source, char* dest, int compressedSize, int maxOutputSize)
 {
     LZ4_streamDecode_t_internal* lz4sd = &LZ4_streamDecode->internal_donotuse;
     int result;
 
-    if (lz4sd->prefixEnd == (BYTE*)dest) {
-        result = LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize,
-                                        endOnInputSize, full, 0,
-                                        usingExtDict, lz4sd->prefixEnd - lz4sd->prefixSize, lz4sd->externalDict, lz4sd->extDictSize);
+    if (lz4sd->prefixSize == 0) {
+        /* The first call, no dictionary yet. */
+        assert(lz4sd->extDictSize == 0);
+        result = LZ4_decompress_safe(source, dest, compressedSize, maxOutputSize);
         if (result <= 0) return result;
-        lz4sd->prefixSize += result;
+        lz4sd->prefixSize = (size_t)result;
+        lz4sd->prefixEnd = (BYTE*)dest + result;
+    } else if (lz4sd->prefixEnd == (BYTE*)dest) {
+        /* They're rolling the current segment. */
+        if (lz4sd->prefixSize >= 64 KB - 1)
+            result = LZ4_decompress_safe_withPrefix64k(source, dest, compressedSize, maxOutputSize);
+        else if (lz4sd->extDictSize == 0)
+            result = LZ4_decompress_safe_withSmallPrefix(source, dest, compressedSize, maxOutputSize,
+                                                         lz4sd->prefixSize);
+        else
+            result = LZ4_decompress_safe_doubleDict(source, dest, compressedSize, maxOutputSize,
+                                                    lz4sd->prefixSize, lz4sd->externalDict, lz4sd->extDictSize);
+        if (result <= 0) return result;
+        lz4sd->prefixSize += (size_t)result;
         lz4sd->prefixEnd  += result;
     } else {
+        /* The buffer wraps around, or they're switching to another buffer. */
         lz4sd->extDictSize = lz4sd->prefixSize;
         lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
-        result = LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize,
-                                        endOnInputSize, full, 0,
-                                        usingExtDict, (BYTE*)dest, lz4sd->externalDict, lz4sd->extDictSize);
+        result = LZ4_decompress_safe_forceExtDict(source, dest, compressedSize, maxOutputSize,
+                                                  lz4sd->externalDict, lz4sd->extDictSize);
         if (result <= 0) return result;
-        lz4sd->prefixSize = result;
+        lz4sd->prefixSize = (size_t)result;
         lz4sd->prefixEnd  = (BYTE*)dest + result;
     }
 
     return result;
 }
 
+LZ4_FORCE_O2_GCC_PPC64LE
 int LZ4_decompress_fast_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* source, char* dest, int originalSize)
 {
     LZ4_streamDecode_t_internal* lz4sd = &LZ4_streamDecode->internal_donotuse;
     int result;
+    assert(originalSize >= 0);
 
-    if (lz4sd->prefixEnd == (BYTE*)dest) {
-        result = LZ4_decompress_generic(source, dest, 0, originalSize,
-                                        endOnOutputSize, full, 0,
-                                        usingExtDict, lz4sd->prefixEnd - lz4sd->prefixSize, lz4sd->externalDict, lz4sd->extDictSize);
+    if (lz4sd->prefixSize == 0) {
+        assert(lz4sd->extDictSize == 0);
+        result = LZ4_decompress_fast(source, dest, originalSize);
         if (result <= 0) return result;
-        lz4sd->prefixSize += originalSize;
+        lz4sd->prefixSize = (size_t)originalSize;
+        lz4sd->prefixEnd = (BYTE*)dest + originalSize;
+    } else if (lz4sd->prefixEnd == (BYTE*)dest) {
+        if (lz4sd->prefixSize >= 64 KB - 1 || lz4sd->extDictSize == 0)
+            result = LZ4_decompress_fast(source, dest, originalSize);
+        else
+            result = LZ4_decompress_fast_doubleDict(source, dest, originalSize,
+                                                    lz4sd->prefixSize, lz4sd->externalDict, lz4sd->extDictSize);
+        if (result <= 0) return result;
+        lz4sd->prefixSize += (size_t)originalSize;
         lz4sd->prefixEnd  += originalSize;
     } else {
         lz4sd->extDictSize = lz4sd->prefixSize;
         lz4sd->externalDict = lz4sd->prefixEnd - lz4sd->extDictSize;
-        result = LZ4_decompress_generic(source, dest, 0, originalSize,
-                                        endOnOutputSize, full, 0,
-                                        usingExtDict, (BYTE*)dest, lz4sd->externalDict, lz4sd->extDictSize);
+        result = LZ4_decompress_fast_extDict(source, dest, originalSize,
+                                             lz4sd->externalDict, lz4sd->extDictSize);
         if (result <= 0) return result;
-        lz4sd->prefixSize = originalSize;
+        lz4sd->prefixSize = (size_t)originalSize;
         lz4sd->prefixEnd  = (BYTE*)dest + originalSize;
     }
 
@@ -1377,32 +2314,27 @@  Advanced decoding functions :
     the dictionary must be explicitly provided within parameters
 */
 
-FORCE_INLINE int LZ4_decompress_usingDict_generic(const char* source, char* dest, int compressedSize, int maxOutputSize, int safe, const char* dictStart, int dictSize)
+int LZ4_decompress_safe_usingDict(const char* source, char* dest, int compressedSize, int maxOutputSize, const char* dictStart, int dictSize)
 {
     if (dictSize==0)
-        return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, safe, full, 0, noDict, (BYTE*)dest, NULL, 0);
+        return LZ4_decompress_safe(source, dest, compressedSize, maxOutputSize);
     if (dictStart+dictSize == dest) {
-        if (dictSize >= (int)(64 KB - 1))
-            return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, safe, full, 0, withPrefix64k, (BYTE*)dest-64 KB, NULL, 0);
-        return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, safe, full, 0, noDict, (BYTE*)dest-dictSize, NULL, 0);
+        if (dictSize >= 64 KB - 1) {
+            return LZ4_decompress_safe_withPrefix64k(source, dest, compressedSize, maxOutputSize);
+        }
+        assert(dictSize >= 0);
+        return LZ4_decompress_safe_withSmallPrefix(source, dest, compressedSize, maxOutputSize, (size_t)dictSize);
     }
-    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, safe, full, 0, usingExtDict, (BYTE*)dest, (const BYTE*)dictStart, dictSize);
-}
-
-int LZ4_decompress_safe_usingDict(const char* source, char* dest, int compressedSize, int maxOutputSize, const char* dictStart, int dictSize)
-{
-    return LZ4_decompress_usingDict_generic(source, dest, compressedSize, maxOutputSize, 1, dictStart, dictSize);
+    assert(dictSize >= 0);
+    return LZ4_decompress_safe_forceExtDict(source, dest, compressedSize, maxOutputSize, dictStart, (size_t)dictSize);
 }
 
 int LZ4_decompress_fast_usingDict(const char* source, char* dest, int originalSize, const char* dictStart, int dictSize)
 {
-    return LZ4_decompress_usingDict_generic(source, dest, 0, originalSize, 0, dictStart, dictSize);
-}
-
-/* debug function */
-int LZ4_decompress_safe_forceExtDict(const char* source, char* dest, int compressedSize, int maxOutputSize, const char* dictStart, int dictSize)
-{
-    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, endOnInputSize, full, 0, usingExtDict, (BYTE*)dest, (const BYTE*)dictStart, dictSize);
+    if (dictSize==0 || dictStart+dictSize == dest)
+        return LZ4_decompress_fast(source, dest, originalSize);
+    assert(dictSize >= 0);
+    return LZ4_decompress_fast_extDict(source, dest, originalSize, dictStart, (size_t)dictSize);
 }
 
 
@@ -1410,64 +2342,67 @@  int LZ4_decompress_safe_forceExtDict(const char* source, char* dest, int compres
 *  Obsolete Functions
 ***************************************************/
 /* obsolete compression functions */
-int LZ4_compress_limitedOutput(const char* source, char* dest, int inputSize, int maxOutputSize) { return LZ4_compress_default(source, dest, inputSize, maxOutputSize); }
-int LZ4_compress(const char* source, char* dest, int inputSize) { return LZ4_compress_default(source, dest, inputSize, LZ4_compressBound(inputSize)); }
-int LZ4_compress_limitedOutput_withState (void* state, const char* src, char* dst, int srcSize, int dstSize) { return LZ4_compress_fast_extState(state, src, dst, srcSize, dstSize, 1); }
-int LZ4_compress_withState (void* state, const char* src, char* dst, int srcSize) { return LZ4_compress_fast_extState(state, src, dst, srcSize, LZ4_compressBound(srcSize), 1); }
-int LZ4_compress_limitedOutput_continue (LZ4_stream_t* LZ4_stream, const char* src, char* dst, int srcSize, int maxDstSize) { return LZ4_compress_fast_continue(LZ4_stream, src, dst, srcSize, maxDstSize, 1); }
-int LZ4_compress_continue (LZ4_stream_t* LZ4_stream, const char* source, char* dest, int inputSize) { return LZ4_compress_fast_continue(LZ4_stream, source, dest, inputSize, LZ4_compressBound(inputSize), 1); }
+int LZ4_compress_limitedOutput(const char* source, char* dest, int inputSize, int maxOutputSize)
+{
+    return LZ4_compress_default(source, dest, inputSize, maxOutputSize);
+}
+int LZ4_compress(const char* src, char* dest, int srcSize)
+{
+    return LZ4_compress_default(src, dest, srcSize, LZ4_compressBound(srcSize));
+}
+int LZ4_compress_limitedOutput_withState (void* state, const char* src, char* dst, int srcSize, int dstSize)
+{
+    return LZ4_compress_fast_extState(state, src, dst, srcSize, dstSize, 1);
+}
+int LZ4_compress_withState (void* state, const char* src, char* dst, int srcSize)
+{
+    return LZ4_compress_fast_extState(state, src, dst, srcSize, LZ4_compressBound(srcSize), 1);
+}
+int LZ4_compress_limitedOutput_continue (LZ4_stream_t* LZ4_stream, const char* src, char* dst, int srcSize, int dstCapacity)
+{
+    return LZ4_compress_fast_continue(LZ4_stream, src, dst, srcSize, dstCapacity, 1);
+}
+int LZ4_compress_continue (LZ4_stream_t* LZ4_stream, const char* source, char* dest, int inputSize)
+{
+    return LZ4_compress_fast_continue(LZ4_stream, source, dest, inputSize, LZ4_compressBound(inputSize), 1);
+}
 
 /*
-These function names are deprecated and should no longer be used.
+These decompression functions are deprecated and should no longer be used.
 They are only provided here for compatibility with older user programs.
 - LZ4_uncompress is totally equivalent to LZ4_decompress_fast
 - LZ4_uncompress_unknownOutputSize is totally equivalent to LZ4_decompress_safe
 */
-int LZ4_uncompress (const char* source, char* dest, int outputSize) { return LZ4_decompress_fast(source, dest, outputSize); }
-int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize) { return LZ4_decompress_safe(source, dest, isize, maxOutputSize); }
-
+int LZ4_uncompress (const char* source, char* dest, int outputSize)
+{
+    return LZ4_decompress_fast(source, dest, outputSize);
+}
+int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize)
+{
+    return LZ4_decompress_safe(source, dest, isize, maxOutputSize);
+}
 
 /* Obsolete Streaming functions */
 
 int LZ4_sizeofStreamState() { return LZ4_STREAMSIZE; }
 
-static void LZ4_init(LZ4_stream_t* lz4ds, BYTE* base)
-{
-    MEM_INIT(lz4ds, 0, sizeof(LZ4_stream_t));
-    lz4ds->internal_donotuse.bufferStart = base;
-}
-
 int LZ4_resetStreamState(void* state, char* inputBuffer)
 {
-    if ((((uptrval)state) & 3) != 0) return 1;   /* Error : pointer is not aligned on 4-bytes boundary */
-    LZ4_init((LZ4_stream_t*)state, (BYTE*)inputBuffer);
+    (void)inputBuffer;
+    LZ4_resetStream((LZ4_stream_t*)state);
     return 0;
 }
 
 void* LZ4_create (char* inputBuffer)
 {
-    LZ4_stream_t* lz4ds = (LZ4_stream_t*)ALLOCATOR(8, sizeof(LZ4_stream_t));
-    LZ4_init (lz4ds, (BYTE*)inputBuffer);
-    return lz4ds;
+    (void)inputBuffer;
+    return LZ4_createStream();
 }
 
-char* LZ4_slideInputBuffer (void* LZ4_Data)
+char* LZ4_slideInputBuffer (void* state)
 {
-    LZ4_stream_t_internal* ctx = &((LZ4_stream_t*)LZ4_Data)->internal_donotuse;
-    int dictSize = LZ4_saveDict((LZ4_stream_t*)LZ4_Data, (char*)ctx->bufferStart, 64 KB);
-    return (char*)(ctx->bufferStart + dictSize);
-}
-
-/* Obsolete streaming decompression functions */
-
-int LZ4_decompress_safe_withPrefix64k(const char* source, char* dest, int compressedSize, int maxOutputSize)
-{
-    return LZ4_decompress_generic(source, dest, compressedSize, maxOutputSize, endOnInputSize, full, 0, withPrefix64k, (BYTE*)dest - 64 KB, NULL, 64 KB);
-}
-
-int LZ4_decompress_fast_withPrefix64k(const char* source, char* dest, int originalSize)
-{
-    return LZ4_decompress_generic(source, dest, 0, originalSize, endOnOutputSize, full, 0, withPrefix64k, (BYTE*)dest - 64 KB, NULL, 64 KB);
+    /* avoid const char * -> char * conversion warning */
+    return (char *)(uptrval)((LZ4_stream_t*)state)->internal_donotuse.dictionary;
 }
 
 #endif   /* LZ4_COMMONDEFS_ONLY */
diff --git a/src/compat/compat-lz4.h b/src/compat/compat-lz4.h
index 0aae19c9..32108e23 100644
--- a/src/compat/compat-lz4.h
+++ b/src/compat/compat-lz4.h
@@ -1,7 +1,7 @@ 
 /*
  *  LZ4 - Fast LZ compression algorithm
  *  Header File
- *  Copyright (C) 2011-2016, Yann Collet.
+ *  Copyright (C) 2011-present, Yann Collet.
 
    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
 
@@ -32,13 +32,13 @@ 
     - LZ4 homepage : http://www.lz4.org
     - LZ4 source repository : https://github.com/lz4/lz4
 */
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
 #ifndef LZ4_H_2983827168210
 #define LZ4_H_2983827168210
 
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
 /* --- Dependency --- */
 #include <stddef.h>   /* size_t */
 
@@ -46,24 +46,31 @@  extern "C" {
 /**
   Introduction
 
-  LZ4 is lossless compression algorithm, providing compression speed at 400 MB/s per core,
+  LZ4 is lossless compression algorithm, providing compression speed >500 MB/s per core,
   scalable with multi-cores CPU. It features an extremely fast decoder, with speed in
   multiple GB/s per core, typically reaching RAM speed limits on multi-core systems.
 
   The LZ4 compression library provides in-memory compression and decompression functions.
+  It gives full buffer control to user.
   Compression can be done in:
     - a single step (described as Simple Functions)
     - a single step, reusing a context (described in Advanced Functions)
     - unbounded multiple steps (described as Streaming compression)
 
-  lz4.h provides block compression functions. It gives full buffer control to user.
-  Decompressing an lz4-compressed block also requires metadata (such as compressed size).
-  Each application is free to encode such metadata in whichever way it wants.
+  lz4.h generates and decodes LZ4-compressed blocks (doc/lz4_Block_format.md).
+  Decompressing such a compressed block requires additional metadata.
+  Exact metadata depends on exact decompression function.
+  For the typical case of LZ4_decompress_safe(),
+  metadata includes block's compressed size, and maximum bound of decompressed size.
+  Each application is free to encode and pass such metadata in whichever way it wants.
 
-  An additional format, called LZ4 frame specification (doc/lz4_Frame_format.md),
-  take care of encoding standard metadata alongside LZ4-compressed blocks.
-  If your application requires interoperability, it's recommended to use it.
-  A library is provided to take care of it, see lz4frame.h.
+  lz4.h only handle blocks, it can not generate Frames.
+
+  Blocks are different from Frames (doc/lz4_Frame_format.md).
+  Frames bundle both blocks and metadata in a specified manner.
+  Embedding metadata is required for compressed data to be self-contained and portable.
+  Frame format is delivered through a companion API, declared in lz4frame.h.
+  The `lz4` CLI can only manage frames.
 */
 
 /*^***************************************************************
@@ -72,20 +79,28 @@  extern "C" {
 /*
 *  LZ4_DLL_EXPORT :
 *  Enable exporting of functions when building a Windows DLL
+*  LZ4LIB_VISIBILITY :
+*  Control library symbols visibility.
 */
+#ifndef LZ4LIB_VISIBILITY
+#  if defined(__GNUC__) && (__GNUC__ >= 4)
+#    define LZ4LIB_VISIBILITY __attribute__ ((visibility ("default")))
+#  else
+#    define LZ4LIB_VISIBILITY
+#  endif
+#endif
 #if defined(LZ4_DLL_EXPORT) && (LZ4_DLL_EXPORT==1)
-#  define LZ4LIB_API __declspec(dllexport)
+#  define LZ4LIB_API __declspec(dllexport) LZ4LIB_VISIBILITY
 #elif defined(LZ4_DLL_IMPORT) && (LZ4_DLL_IMPORT==1)
-#  define LZ4LIB_API __declspec(dllimport) /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
+#  define LZ4LIB_API __declspec(dllimport) LZ4LIB_VISIBILITY /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
 #else
-#  define LZ4LIB_API
+#  define LZ4LIB_API LZ4LIB_VISIBILITY
 #endif
 
-
-/*========== Version =========== */
+/*------   Version   ------*/
 #define LZ4_VERSION_MAJOR    1    /* for breaking interface changes  */
-#define LZ4_VERSION_MINOR    7    /* for new (non-breaking) interface capabilities */
-#define LZ4_VERSION_RELEASE  5    /* for tweaks, bug-fixes, or development */
+#define LZ4_VERSION_MINOR    9    /* for new (non-breaking) interface capabilities */
+#define LZ4_VERSION_RELEASE  2    /* for tweaks, bug-fixes, or development */
 
 #define LZ4_VERSION_NUMBER (LZ4_VERSION_MAJOR *100*100 + LZ4_VERSION_MINOR *100 + LZ4_VERSION_RELEASE)
 
@@ -94,8 +109,8 @@  extern "C" {
 #define LZ4_EXPAND_AND_QUOTE(str) LZ4_QUOTE(str)
 #define LZ4_VERSION_STRING LZ4_EXPAND_AND_QUOTE(LZ4_LIB_VERSION)
 
-LZ4LIB_API int LZ4_versionNumber (void);
-LZ4LIB_API const char* LZ4_versionString (void);
+LZ4LIB_API int LZ4_versionNumber (void);  /**< library version number; useful to check dll version */
+LZ4LIB_API const char* LZ4_versionString (void);   /**< library version string; useful to check dll version */
 
 
 /*-************************************
@@ -104,41 +119,49 @@  LZ4LIB_API const char* LZ4_versionString (void);
 /*!
  * LZ4_MEMORY_USAGE :
  * Memory usage formula : N->2^N Bytes (examples : 10 -> 1KB; 12 -> 4KB ; 16 -> 64KB; 20 -> 1MB; etc.)
- * Increasing memory usage improves compression ratio
- * Reduced memory usage can improve speed, due to cache effect
+ * Increasing memory usage improves compression ratio.
+ * Reduced memory usage may improve speed, thanks to better cache locality.
  * Default value is 14, for 16KB, which nicely fits into Intel x86 L1 cache
  */
-#define LZ4_MEMORY_USAGE 14
+#ifndef LZ4_MEMORY_USAGE
+# define LZ4_MEMORY_USAGE 14
+#endif
 
 
 /*-************************************
 *  Simple Functions
 **************************************/
 /*! LZ4_compress_default() :
-    Compresses 'sourceSize' bytes from buffer 'source'
-    into already allocated 'dest' buffer of size 'maxDestSize'.
-    Compression is guaranteed to succeed if 'maxDestSize' >= LZ4_compressBound(sourceSize).
-    It also runs faster, so it's a recommended setting.
-    If the function cannot compress 'source' into a more limited 'dest' budget,
-    compression stops *immediately*, and the function result is zero.
-    As a consequence, 'dest' content is not valid.
-    This function never writes outside 'dest' buffer, nor read outside 'source' buffer.
-        sourceSize  : Max supported value is LZ4_MAX_INPUT_VALUE
-        maxDestSize : full or partial size of buffer 'dest' (which must be already allocated)
-        return : the number of bytes written into buffer 'dest' (necessarily <= maxOutputSize)
-              or 0 if compression fails */
-LZ4LIB_API int LZ4_compress_default(const char* source, char* dest, int sourceSize, int maxDestSize);
+ *  Compresses 'srcSize' bytes from buffer 'src'
+ *  into already allocated 'dst' buffer of size 'dstCapacity'.
+ *  Compression is guaranteed to succeed if 'dstCapacity' >= LZ4_compressBound(srcSize).
+ *  It also runs faster, so it's a recommended setting.
+ *  If the function cannot compress 'src' into a more limited 'dst' budget,
+ *  compression stops *immediately*, and the function result is zero.
+ *  In which case, 'dst' content is undefined (invalid).
+ *      srcSize : max supported value is LZ4_MAX_INPUT_SIZE.
+ *      dstCapacity : size of buffer 'dst' (which must be already allocated)
+ *     @return  : the number of bytes written into buffer 'dst' (necessarily <= dstCapacity)
+ *                or 0 if compression fails
+ * Note : This function is protected against buffer overflow scenarios (never writes outside 'dst' buffer, nor read outside 'source' buffer).
+ */
+LZ4LIB_API int LZ4_compress_default(const char* src, char* dst, int srcSize, int dstCapacity);
 
 /*! LZ4_decompress_safe() :
-    compressedSize : is the precise full size of the compressed block.
-    maxDecompressedSize : is the size of destination buffer, which must be already allocated.
-    return : the number of bytes decompressed into destination buffer (necessarily <= maxDecompressedSize)
-             If destination buffer is not large enough, decoding will stop and output an error code (<0).
-             If the source stream is detected malformed, the function will stop decoding and return a negative result.
-             This function is protected against buffer overflow exploits, including malicious data packets.
-             It never writes outside output buffer, nor reads outside input buffer.
-*/
-LZ4LIB_API int LZ4_decompress_safe (const char* source, char* dest, int compressedSize, int maxDecompressedSize);
+ *  compressedSize : is the exact complete size of the compressed block.
+ *  dstCapacity : is the size of destination buffer (which must be already allocated), presumed an upper bound of decompressed size.
+ * @return : the number of bytes decompressed into destination buffer (necessarily <= dstCapacity)
+ *           If destination buffer is not large enough, decoding will stop and output an error code (negative value).
+ *           If the source stream is detected malformed, the function will stop decoding and return a negative result.
+ * Note 1 : This function is protected against malicious data packets :
+ *          it will never writes outside 'dst' buffer, nor read outside 'source' buffer,
+ *          even if the compressed block is maliciously modified to order the decoder to do these actions.
+ *          In such case, the decoder stops immediately, and considers the compressed block malformed.
+ * Note 2 : compressedSize and dstCapacity must be provided to the function, the compressed block does not contain them.
+ *          The implementation is free to send / store / derive this information in whichever way is most beneficial.
+ *          If there is a need for a different format which bundles together both compressed data and its metadata, consider looking at lz4frame.h instead.
+ */
+LZ4LIB_API int LZ4_decompress_safe (const char* src, char* dst, int compressedSize, int dstCapacity);
 
 
 /*-************************************
@@ -147,184 +170,389 @@  LZ4LIB_API int LZ4_decompress_safe (const char* source, char* dest, int compress
 #define LZ4_MAX_INPUT_SIZE        0x7E000000   /* 2 113 929 216 bytes */
 #define LZ4_COMPRESSBOUND(isize)  ((unsigned)(isize) > (unsigned)LZ4_MAX_INPUT_SIZE ? 0 : (isize) + ((isize)/255) + 16)
 
-/*!
-LZ4_compressBound() :
+/*! LZ4_compressBound() :
     Provides the maximum size that LZ4 compression may output in a "worst case" scenario (input data not compressible)
     This function is primarily useful for memory allocation purposes (destination buffer size).
     Macro LZ4_COMPRESSBOUND() is also provided for compilation-time evaluation (stack memory allocation for example).
-    Note that LZ4_compress_default() compress faster when dest buffer size is >= LZ4_compressBound(srcSize)
+    Note that LZ4_compress_default() compresses faster when dstCapacity is >= LZ4_compressBound(srcSize)
         inputSize  : max supported value is LZ4_MAX_INPUT_SIZE
         return : maximum output size in a "worst case" scenario
-              or 0, if input size is too large ( > LZ4_MAX_INPUT_SIZE)
+              or 0, if input size is incorrect (too large or negative)
 */
 LZ4LIB_API int LZ4_compressBound(int inputSize);
 
-/*!
-LZ4_compress_fast() :
-    Same as LZ4_compress_default(), but allows to select an "acceleration" factor.
+/*! LZ4_compress_fast() :
+    Same as LZ4_compress_default(), but allows selection of "acceleration" factor.
     The larger the acceleration value, the faster the algorithm, but also the lesser the compression.
     It's a trade-off. It can be fine tuned, with each successive value providing roughly +~3% to speed.
     An acceleration value of "1" is the same as regular LZ4_compress_default()
-    Values <= 0 will be replaced by ACCELERATION_DEFAULT (see lz4.c), which is 1.
+    Values <= 0 will be replaced by ACCELERATION_DEFAULT (currently == 1, see lz4.c).
 */
-LZ4LIB_API int LZ4_compress_fast (const char* source, char* dest, int sourceSize, int maxDestSize, int acceleration);
+LZ4LIB_API int LZ4_compress_fast (const char* src, char* dst, int srcSize, int dstCapacity, int acceleration);
 
 
-/*!
-LZ4_compress_fast_extState() :
-    Same compression function, just using an externally allocated memory space to store compression state.
-    Use LZ4_sizeofState() to know how much memory must be allocated,
-    and allocate it on 8-bytes boundaries (using malloc() typically).
-    Then, provide it as 'void* state' to compression function.
-*/
+/*! LZ4_compress_fast_extState() :
+ *  Same as LZ4_compress_fast(), using an externally allocated memory space for its state.
+ *  Use LZ4_sizeofState() to know how much memory must be allocated,
+ *  and allocate it on 8-bytes boundaries (using `malloc()` typically).
+ *  Then, provide this buffer as `void* state` to compression function.
+ */
 LZ4LIB_API int LZ4_sizeofState(void);
-LZ4LIB_API int LZ4_compress_fast_extState (void* state, const char* source, char* dest, int inputSize, int maxDestSize, int acceleration);
+LZ4LIB_API int LZ4_compress_fast_extState (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration);
 
 
-/*!
-LZ4_compress_destSize() :
-    Reverse the logic, by compressing as much data as possible from 'source' buffer
-    into already allocated buffer 'dest' of size 'targetDestSize'.
-    This function either compresses the entire 'source' content into 'dest' if it's large enough,
-    or fill 'dest' buffer completely with as much data as possible from 'source'.
-        *sourceSizePtr : will be modified to indicate how many bytes where read from 'source' to fill 'dest'.
-                         New value is necessarily <= old value.
-        return : Nb bytes written into 'dest' (necessarily <= targetDestSize)
-              or 0 if compression fails
+/*! LZ4_compress_destSize() :
+ *  Reverse the logic : compresses as much data as possible from 'src' buffer
+ *  into already allocated buffer 'dst', of size >= 'targetDestSize'.
+ *  This function either compresses the entire 'src' content into 'dst' if it's large enough,
+ *  or fill 'dst' buffer completely with as much data as possible from 'src'.
+ *  note: acceleration parameter is fixed to "default".
+ *
+ * *srcSizePtr : will be modified to indicate how many bytes where read from 'src' to fill 'dst'.
+ *               New value is necessarily <= input value.
+ * @return : Nb bytes written into 'dst' (necessarily <= targetDestSize)
+ *           or 0 if compression fails.
 */
-LZ4LIB_API int LZ4_compress_destSize (const char* source, char* dest, int* sourceSizePtr, int targetDestSize);
+LZ4LIB_API int LZ4_compress_destSize (const char* src, char* dst, int* srcSizePtr, int targetDstSize);
 
 
-/*!
-LZ4_decompress_fast() :
-    originalSize : is the original and therefore uncompressed size
-    return : the number of bytes read from the source buffer (in other words, the compressed size)
-             If the source stream is detected malformed, the function will stop decoding and return a negative result.
-             Destination buffer must be already allocated. Its size must be a minimum of 'originalSize' bytes.
-    note : This function fully respect memory boundaries for properly formed compressed data.
-           It is a bit faster than LZ4_decompress_safe().
-           However, it does not provide any protection against intentionally modified data stream (malicious input).
-           Use this function in trusted environment only (data to decode comes from a trusted source).
-*/
-LZ4LIB_API int LZ4_decompress_fast (const char* source, char* dest, int originalSize);
-
-/*!
-LZ4_decompress_safe_partial() :
-    This function decompress a compressed block of size 'compressedSize' at position 'source'
-    into destination buffer 'dest' of size 'maxDecompressedSize'.
-    The function tries to stop decompressing operation as soon as 'targetOutputSize' has been reached,
-    reducing decompression time.
-    return : the number of bytes decoded in the destination buffer (necessarily <= maxDecompressedSize)
-       Note : this number can be < 'targetOutputSize' should the compressed block to decode be smaller.
-             Always control how many bytes were decoded.
-             If the source stream is detected malformed, the function will stop decoding and return a negative result.
-             This function never writes outside of output buffer, and never reads outside of input buffer. It is therefore protected against malicious data packets
-*/
-LZ4LIB_API int LZ4_decompress_safe_partial (const char* source, char* dest, int compressedSize, int targetOutputSize, int maxDecompressedSize);
+/*! LZ4_decompress_safe_partial() :
+ *  Decompress an LZ4 compressed block, of size 'srcSize' at position 'src',
+ *  into destination buffer 'dst' of size 'dstCapacity'.
+ *  Up to 'targetOutputSize' bytes will be decoded.
+ *  The function stops decoding on reaching this objective,
+ *  which can boost performance when only the beginning of a block is required.
+ *
+ * @return : the number of bytes decoded in `dst` (necessarily <= dstCapacity)
+ *           If source stream is detected malformed, function returns a negative result.
+ *
+ *  Note : @return can be < targetOutputSize, if compressed block contains less data.
+ *
+ *  Note 2 : this function features 2 parameters, targetOutputSize and dstCapacity,
+ *           and expects targetOutputSize <= dstCapacity.
+ *           It effectively stops decoding on reaching targetOutputSize,
+ *           so dstCapacity is kind of redundant.
+ *           This is because in a previous version of this function,
+ *           decoding operation would not "break" a sequence in the middle.
+ *           As a consequence, there was no guarantee that decoding would stop at exactly targetOutputSize,
+ *           it could write more bytes, though only up to dstCapacity.
+ *           Some "margin" used to be required for this operation to work properly.
+ *           This is no longer necessary.
+ *           The function nonetheless keeps its signature, in an effort to not break API.
+ */
+LZ4LIB_API int LZ4_decompress_safe_partial (const char* src, char* dst, int srcSize, int targetOutputSize, int dstCapacity);
 
 
 /*-*********************************************
 *  Streaming Compression Functions
 ***********************************************/
-typedef union LZ4_stream_u LZ4_stream_t;   /* incomplete type (defined later) */
+typedef union LZ4_stream_u LZ4_stream_t;  /* incomplete type (defined later) */
 
-/*! LZ4_createStream() and LZ4_freeStream() :
- *  LZ4_createStream() will allocate and initialize an `LZ4_stream_t` structure.
- *  LZ4_freeStream() releases its memory.
- */
 LZ4LIB_API LZ4_stream_t* LZ4_createStream(void);
 LZ4LIB_API int           LZ4_freeStream (LZ4_stream_t* streamPtr);
 
-/*! LZ4_resetStream() :
- *  An LZ4_stream_t structure can be allocated once and re-used multiple times.
- *  Use this function to init an allocated `LZ4_stream_t` structure and start a new compression.
+/*! LZ4_resetStream_fast() : v1.9.0+
+ *  Use this to prepare an LZ4_stream_t for a new chain of dependent blocks
+ *  (e.g., LZ4_compress_fast_continue()).
+ *
+ *  An LZ4_stream_t must be initialized once before usage.
+ *  This is automatically done when created by LZ4_createStream().
+ *  However, should the LZ4_stream_t be simply declared on stack (for example),
+ *  it's necessary to initialize it first, using LZ4_initStream().
+ *
+ *  After init, start any new stream with LZ4_resetStream_fast().
+ *  A same LZ4_stream_t can be re-used multiple times consecutively
+ *  and compress multiple streams,
+ *  provided that it starts each new stream with LZ4_resetStream_fast().
+ *
+ *  LZ4_resetStream_fast() is much faster than LZ4_initStream(),
+ *  but is not compatible with memory regions containing garbage data.
+ *
+ *  Note: it's only useful to call LZ4_resetStream_fast()
+ *        in the context of streaming compression.
+ *        The *extState* functions perform their own resets.
+ *        Invoking LZ4_resetStream_fast() before is redundant, and even counterproductive.
  */
-LZ4LIB_API void LZ4_resetStream (LZ4_stream_t* streamPtr);
+LZ4LIB_API void LZ4_resetStream_fast (LZ4_stream_t* streamPtr);
 
 /*! LZ4_loadDict() :
- *  Use this function to load a static dictionary into LZ4_stream.
- *  Any previous data will be forgotten, only 'dictionary' will remain in memory.
- *  Loading a size of 0 is allowed.
- *  Return : dictionary size, in bytes (necessarily <= 64 KB)
+ *  Use this function to reference a static dictionary into LZ4_stream_t.
+ *  The dictionary must remain available during compression.
+ *  LZ4_loadDict() triggers a reset, so any previous data will be forgotten.
+ *  The same dictionary will have to be loaded on decompression side for successful decoding.
+ *  Dictionary are useful for better compression of small data (KB range).
+ *  While LZ4 accept any input as dictionary,
+ *  results are generally better when using Zstandard's Dictionary Builder.
+ *  Loading a size of 0 is allowed, and is the same as reset.
+ * @return : loaded dictionary size, in bytes (necessarily <= 64 KB)
  */
 LZ4LIB_API int LZ4_loadDict (LZ4_stream_t* streamPtr, const char* dictionary, int dictSize);
 
 /*! LZ4_compress_fast_continue() :
- *  Compress buffer content 'src', using data from previously compressed blocks as dictionary to improve compression ratio.
- *  Important : Previous data blocks are assumed to still be present and unmodified !
- *  'dst' buffer must be already allocated.
- *  If maxDstSize >= LZ4_compressBound(srcSize), compression is guaranteed to succeed, and runs faster.
- *  If not, and if compressed data cannot fit into 'dst' buffer size, compression stops, and function returns a zero.
+ *  Compress 'src' content using data from previously compressed blocks, for better compression ratio.
+ * 'dst' buffer must be already allocated.
+ *  If dstCapacity >= LZ4_compressBound(srcSize), compression is guaranteed to succeed, and runs faster.
+ *
+ * @return : size of compressed block
+ *           or 0 if there is an error (typically, cannot fit into 'dst').
+ *
+ *  Note 1 : Each invocation to LZ4_compress_fast_continue() generates a new block.
+ *           Each block has precise boundaries.
+ *           Each block must be decompressed separately, calling LZ4_decompress_*() with relevant metadata.
+ *           It's not possible to append blocks together and expect a single invocation of LZ4_decompress_*() to decompress them together.
+ *
+ *  Note 2 : The previous 64KB of source data is __assumed__ to remain present, unmodified, at same address in memory !
+ *
+ *  Note 3 : When input is structured as a double-buffer, each buffer can have any size, including < 64 KB.
+ *           Make sure that buffers are separated, by at least one byte.
+ *           This construction ensures that each block only depends on previous block.
+ *
+ *  Note 4 : If input buffer is a ring-buffer, it can have any size, including < 64 KB.
+ *
+ *  Note 5 : After an error, the stream status is undefined (invalid), it can only be reset or freed.
  */
-LZ4LIB_API int LZ4_compress_fast_continue (LZ4_stream_t* streamPtr, const char* src, char* dst, int srcSize, int maxDstSize, int acceleration);
+LZ4LIB_API int LZ4_compress_fast_continue (LZ4_stream_t* streamPtr, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration);
 
 /*! LZ4_saveDict() :
- *  If previously compressed data block is not guaranteed to remain available at its memory location,
+ *  If last 64KB data cannot be guaranteed to remain available at its current memory location,
  *  save it into a safer place (char* safeBuffer).
- *  Note : you don't need to call LZ4_loadDict() afterwards,
- *         dictionary is immediately usable, you can therefore call LZ4_compress_fast_continue().
- *  Return : saved dictionary size in bytes (necessarily <= dictSize), or 0 if error.
+ *  This is schematically equivalent to a memcpy() followed by LZ4_loadDict(),
+ *  but is much faster, because LZ4_saveDict() doesn't need to rebuild tables.
+ * @return : saved dictionary size in bytes (necessarily <= maxDictSize), or 0 if error.
  */
-LZ4LIB_API int LZ4_saveDict (LZ4_stream_t* streamPtr, char* safeBuffer, int dictSize);
+LZ4LIB_API int LZ4_saveDict (LZ4_stream_t* streamPtr, char* safeBuffer, int maxDictSize);
 
 
 /*-**********************************************
 *  Streaming Decompression Functions
 *  Bufferless synchronous API
 ************************************************/
-typedef union LZ4_streamDecode_u LZ4_streamDecode_t;   /* incomplete type (defined later) */
+typedef union LZ4_streamDecode_u LZ4_streamDecode_t;   /* tracking context */
 
-/* creation / destruction of streaming decompression tracking structure */
+/*! LZ4_createStreamDecode() and LZ4_freeStreamDecode() :
+ *  creation / destruction of streaming decompression tracking context.
+ *  A tracking context can be re-used multiple times.
+ */
 LZ4LIB_API LZ4_streamDecode_t* LZ4_createStreamDecode(void);
 LZ4LIB_API int                 LZ4_freeStreamDecode (LZ4_streamDecode_t* LZ4_stream);
 
 /*! LZ4_setStreamDecode() :
- *  Use this function to instruct where to find the dictionary.
- *  Setting a size of 0 is allowed (same effect as reset).
- *  @return : 1 if OK, 0 if error
+ *  An LZ4_streamDecode_t context can be allocated once and re-used multiple times.
+ *  Use this function to start decompression of a new stream of blocks.
+ *  A dictionary can optionally be set. Use NULL or size 0 for a reset order.
+ *  Dictionary is presumed stable : it must remain accessible and unmodified during next decompression.
+ * @return : 1 if OK, 0 if error
  */
 LZ4LIB_API int LZ4_setStreamDecode (LZ4_streamDecode_t* LZ4_streamDecode, const char* dictionary, int dictSize);
 
-/*!
-LZ4_decompress_*_continue() :
-    These decoding functions allow decompression of multiple blocks in "streaming" mode.
-    Previously decoded blocks *must* remain available at the memory position where they were decoded (up to 64 KB)
-    In the case of a ring buffers, decoding buffer must be either :
-    - Exactly same size as encoding buffer, with same update rule (block boundaries at same positions)
-      In which case, the decoding & encoding ring buffer can have any size, including very small ones ( < 64 KB).
-    - Larger than encoding buffer, by a minimum of maxBlockSize more bytes.
-      maxBlockSize is implementation dependent. It's the maximum size you intend to compress into a single block.
-      In which case, encoding and decoding buffers do not need to be synchronized,
-      and encoding ring buffer can have any size, including small ones ( < 64 KB).
-    - _At least_ 64 KB + 8 bytes + maxBlockSize.
-      In which case, encoding and decoding buffers do not need to be synchronized,
-      and encoding ring buffer can have any size, including larger than decoding buffer.
-    Whenever these conditions are not possible, save the last 64KB of decoded data into a safe buffer,
-    and indicate where it is saved using LZ4_setStreamDecode()
+/*! LZ4_decoderRingBufferSize() : v1.8.2+
+ *  Note : in a ring buffer scenario (optional),
+ *  blocks are presumed decompressed next to each other
+ *  up to the moment there is not enough remaining space for next block (remainingSize < maxBlockSize),
+ *  at which stage it resumes from beginning of ring buffer.
+ *  When setting such a ring buffer for streaming decompression,
+ *  provides the minimum size of this ring buffer
+ *  to be compatible with any source respecting maxBlockSize condition.
+ * @return : minimum ring buffer size,
+ *           or 0 if there is an error (invalid maxBlockSize).
+ */
+LZ4LIB_API int LZ4_decoderRingBufferSize(int maxBlockSize);
+#define LZ4_DECODER_RING_BUFFER_SIZE(maxBlockSize) (65536 + 14 + (maxBlockSize))  /* for static allocation; maxBlockSize presumed valid */
+
+/*! LZ4_decompress_*_continue() :
+ *  These decoding functions allow decompression of consecutive blocks in "streaming" mode.
+ *  A block is an unsplittable entity, it must be presented entirely to a decompression function.
+ *  Decompression functions only accepts one block at a time.
+ *  The last 64KB of previously decoded data *must* remain available and unmodified at the memory position where they were decoded.
+ *  If less than 64KB of data has been decoded, all the data must be present.
+ *
+ *  Special : if decompression side sets a ring buffer, it must respect one of the following conditions :
+ *  - Decompression buffer size is _at least_ LZ4_decoderRingBufferSize(maxBlockSize).
+ *    maxBlockSize is the maximum size of any single block. It can have any value > 16 bytes.
+ *    In which case, encoding and decoding buffers do not need to be synchronized.
+ *    Actually, data can be produced by any source compliant with LZ4 format specification, and respecting maxBlockSize.
+ *  - Synchronized mode :
+ *    Decompression buffer size is _exactly_ the same as compression buffer size,
+ *    and follows exactly same update rule (block boundaries at same positions),
+ *    and decoding function is provided with exact decompressed size of each block (exception for last block of the stream),
+ *    _then_ decoding & encoding ring buffer can have any size, including small ones ( < 64 KB).
+ *  - Decompression buffer is larger than encoding buffer, by a minimum of maxBlockSize more bytes.
+ *    In which case, encoding and decoding buffers do not need to be synchronized,
+ *    and encoding ring buffer can have any size, including small ones ( < 64 KB).
+ *
+ *  Whenever these conditions are not possible,
+ *  save the last 64KB of decoded data into a safe buffer where it can't be modified during decompression,
+ *  then indicate where this data is saved using LZ4_setStreamDecode(), before decompressing next block.
 */
-LZ4LIB_API int LZ4_decompress_safe_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* source, char* dest, int compressedSize, int maxDecompressedSize);
-LZ4LIB_API int LZ4_decompress_fast_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* source, char* dest, int originalSize);
+LZ4LIB_API int LZ4_decompress_safe_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* src, char* dst, int srcSize, int dstCapacity);
 
 
 /*! LZ4_decompress_*_usingDict() :
  *  These decoding functions work the same as
  *  a combination of LZ4_setStreamDecode() followed by LZ4_decompress_*_continue()
  *  They are stand-alone, and don't need an LZ4_streamDecode_t structure.
+ *  Dictionary is presumed stable : it must remain accessible and unmodified during decompression.
+ *  Performance tip : Decompression speed can be substantially increased
+ *                    when dst == dictStart + dictSize.
  */
-LZ4LIB_API int LZ4_decompress_safe_usingDict (const char* source, char* dest, int compressedSize, int maxDecompressedSize, const char* dictStart, int dictSize);
-LZ4LIB_API int LZ4_decompress_fast_usingDict (const char* source, char* dest, int originalSize, const char* dictStart, int dictSize);
+LZ4LIB_API int LZ4_decompress_safe_usingDict (const char* src, char* dst, int srcSize, int dstCapcity, const char* dictStart, int dictSize);
 
+#endif /* LZ4_H_2983827168210 */
 
-/*^**********************************************
+
+/*^*************************************
  * !!!!!!   STATIC LINKING ONLY   !!!!!!
- ***********************************************/
-/*-************************************
- *  Private definitions
- **************************************
- * Do not use these definitions.
- * They are exposed to allow static allocation of `LZ4_stream_t` and `LZ4_streamDecode_t`.
- * Using these definitions will expose code to API and/or ABI break in future versions of the library.
- **************************************/
+ ***************************************/
+
+/*-****************************************************************************
+ * Experimental section
+ *
+ * Symbols declared in this section must be considered unstable. Their
+ * signatures or semantics may change, or they may be removed altogether in the
+ * future. They are therefore only safe to depend on when the caller is
+ * statically linked against the library.
+ *
+ * To protect against unsafe usage, not only are the declarations guarded,
+ * the definitions are hidden by default
+ * when building LZ4 as a shared/dynamic library.
+ *
+ * In order to access these declarations,
+ * define LZ4_STATIC_LINKING_ONLY in your application
+ * before including LZ4's headers.
+ *
+ * In order to make their implementations accessible dynamically, you must
+ * define LZ4_PUBLISH_STATIC_FUNCTIONS when building the LZ4 library.
+ ******************************************************************************/
+
+#ifdef LZ4_STATIC_LINKING_ONLY
+
+#ifndef LZ4_STATIC_3504398509
+#define LZ4_STATIC_3504398509
+
+#ifdef LZ4_PUBLISH_STATIC_FUNCTIONS
+#define LZ4LIB_STATIC_API LZ4LIB_API
+#else
+#define LZ4LIB_STATIC_API
+#endif
+
+
+/*! LZ4_compress_fast_extState_fastReset() :
+ *  A variant of LZ4_compress_fast_extState().
+ *
+ *  Using this variant avoids an expensive initialization step.
+ *  It is only safe to call if the state buffer is known to be correctly initialized already
+ *  (see above comment on LZ4_resetStream_fast() for a definition of "correctly initialized").
+ *  From a high level, the difference is that
+ *  this function initializes the provided state with a call to something like LZ4_resetStream_fast()
+ *  while LZ4_compress_fast_extState() starts with a call to LZ4_resetStream().
+ */
+LZ4LIB_STATIC_API int LZ4_compress_fast_extState_fastReset (void* state, const char* src, char* dst, int srcSize, int dstCapacity, int acceleration);
+
+/*! LZ4_attach_dictionary() :
+ *  This is an experimental API that allows
+ *  efficient use of a static dictionary many times.
+ *
+ *  Rather than re-loading the dictionary buffer into a working context before
+ *  each compression, or copying a pre-loaded dictionary's LZ4_stream_t into a
+ *  working LZ4_stream_t, this function introduces a no-copy setup mechanism,
+ *  in which the working stream references the dictionary stream in-place.
+ *
+ *  Several assumptions are made about the state of the dictionary stream.
+ *  Currently, only streams which have been prepared by LZ4_loadDict() should
+ *  be expected to work.
+ *
+ *  Alternatively, the provided dictionaryStream may be NULL,
+ *  in which case any existing dictionary stream is unset.
+ *
+ *  If a dictionary is provided, it replaces any pre-existing stream history.
+ *  The dictionary contents are the only history that can be referenced and
+ *  logically immediately precede the data compressed in the first subsequent
+ *  compression call.
+ *
+ *  The dictionary will only remain attached to the working stream through the
+ *  first compression call, at the end of which it is cleared. The dictionary
+ *  stream (and source buffer) must remain in-place / accessible / unchanged
+ *  through the completion of the first compression call on the stream.
+ */
+LZ4LIB_STATIC_API void LZ4_attach_dictionary(LZ4_stream_t* workingStream, const LZ4_stream_t* dictionaryStream);
+
+
+/*! In-place compression and decompression
+ *
+ * It's possible to have input and output sharing the same buffer,
+ * for highly contrained memory environments.
+ * In both cases, it requires input to lay at the end of the buffer,
+ * and decompression to start at beginning of the buffer.
+ * Buffer size must feature some margin, hence be larger than final size.
+ *
+ * |<------------------------buffer--------------------------------->|
+ *                             |<-----------compressed data--------->|
+ * |<-----------decompressed size------------------>|
+ *                                                  |<----margin---->|
+ *
+ * This technique is more useful for decompression,
+ * since decompressed size is typically larger,
+ * and margin is short.
+ *
+ * In-place decompression will work inside any buffer
+ * which size is >= LZ4_DECOMPRESS_INPLACE_BUFFER_SIZE(decompressedSize).
+ * This presumes that decompressedSize > compressedSize.
+ * Otherwise, it means compression actually expanded data,
+ * and it would be more efficient to store such data with a flag indicating it's not compressed.
+ * This can happen when data is not compressible (already compressed, or encrypted).
+ *
+ * For in-place compression, margin is larger, as it must be able to cope with both
+ * history preservation, requiring input data to remain unmodified up to LZ4_DISTANCE_MAX,
+ * and data expansion, which can happen when input is not compressible.
+ * As a consequence, buffer size requirements are much higher,
+ * and memory savings offered by in-place compression are more limited.
+ *
+ * There are ways to limit this cost for compression :
+ * - Reduce history size, by modifying LZ4_DISTANCE_MAX.
+ *   Note that it is a compile-time constant, so all compressions will apply this limit.
+ *   Lower values will reduce compression ratio, except when input_size < LZ4_DISTANCE_MAX,
+ *   so it's a reasonable trick when inputs are known to be small.
+ * - Require the compressor to deliver a "maximum compressed size".
+ *   This is the `dstCapacity` parameter in `LZ4_compress*()`.
+ *   When this size is < LZ4_COMPRESSBOUND(inputSize), then compression can fail,
+ *   in which case, the return code will be 0 (zero).
+ *   The caller must be ready for these cases to happen,
+ *   and typically design a backup scheme to send data uncompressed.
+ * The combination of both techniques can significantly reduce
+ * the amount of margin required for in-place compression.
+ *
+ * In-place compression can work in any buffer
+ * which size is >= (maxCompressedSize)
+ * with maxCompressedSize == LZ4_COMPRESSBOUND(srcSize) for guaranteed compression success.
+ * LZ4_COMPRESS_INPLACE_BUFFER_SIZE() depends on both maxCompressedSize and LZ4_DISTANCE_MAX,
+ * so it's possible to reduce memory requirements by playing with them.
+ */
+
+#define LZ4_DECOMPRESS_INPLACE_MARGIN(compressedSize)          (((compressedSize) >> 8) + 32)
+#define LZ4_DECOMPRESS_INPLACE_BUFFER_SIZE(decompressedSize)   ((decompressedSize) + LZ4_DECOMPRESS_INPLACE_MARGIN(decompressedSize))  /**< note: presumes that compressedSize < decompressedSize. note2: margin is overestimated a bit, since it could use compressedSize instead */
+
+#ifndef LZ4_DISTANCE_MAX   /* history window size; can be user-defined at compile time */
+#  define LZ4_DISTANCE_MAX 65535   /* set to maximum value by default */
+#endif
+
+#define LZ4_COMPRESS_INPLACE_MARGIN                           (LZ4_DISTANCE_MAX + 32)   /* LZ4_DISTANCE_MAX can be safely replaced by srcSize when it's smaller */
+#define LZ4_COMPRESS_INPLACE_BUFFER_SIZE(maxCompressedSize)   ((maxCompressedSize) + LZ4_COMPRESS_INPLACE_MARGIN)  /**< maxCompressedSize is generally LZ4_COMPRESSBOUND(inputSize), but can be set to any lower value, with the risk that compression can fail (return code 0(zero)) */
+
+#endif   /* LZ4_STATIC_3504398509 */
+#endif   /* LZ4_STATIC_LINKING_ONLY */
+
+
+
+#ifndef LZ4_H_98237428734687
+#define LZ4_H_98237428734687
+
+/*-************************************************************
+ *  PRIVATE DEFINITIONS
+ **************************************************************
+ * Do not use these definitions directly.
+ * They are only exposed to allow static allocation of `LZ4_stream_t` and `LZ4_streamDecode_t`.
+ * Accessing members will expose code to API and/or ABI break in future versions of the library.
+ **************************************************************/
 #define LZ4_HASHLOG   (LZ4_MEMORY_USAGE-2)
 #define LZ4_HASHTABLESIZE (1 << LZ4_MEMORY_USAGE)
 #define LZ4_HASH_SIZE_U32 (1 << LZ4_HASHLOG)       /* required as macro for static allocation */
@@ -332,14 +560,16 @@  LZ4LIB_API int LZ4_decompress_fast_usingDict (const char* source, char* dest, in
 #if defined(__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
 #include <stdint.h>
 
-typedef struct {
+typedef struct LZ4_stream_t_internal LZ4_stream_t_internal;
+struct LZ4_stream_t_internal {
     uint32_t hashTable[LZ4_HASH_SIZE_U32];
     uint32_t currentOffset;
-    uint32_t initCheck;
+    uint16_t dirty;
+    uint16_t tableType;
     const uint8_t* dictionary;
-    uint8_t* bufferStart;   /* obsolete, used for slideInputBuffer */
+    const LZ4_stream_t_internal* dictCtx;
     uint32_t dictSize;
-} LZ4_stream_t_internal;
+};
 
 typedef struct {
     const uint8_t* externalDict;
@@ -350,49 +580,67 @@  typedef struct {
 
 #else
 
-typedef struct {
+typedef struct LZ4_stream_t_internal LZ4_stream_t_internal;
+struct LZ4_stream_t_internal {
     unsigned int hashTable[LZ4_HASH_SIZE_U32];
     unsigned int currentOffset;
-    unsigned int initCheck;
+    unsigned short dirty;
+    unsigned short tableType;
     const unsigned char* dictionary;
-    unsigned char* bufferStart;   /* obsolete, used for slideInputBuffer */
+    const LZ4_stream_t_internal* dictCtx;
     unsigned int dictSize;
-} LZ4_stream_t_internal;
+};
 
 typedef struct {
     const unsigned char* externalDict;
-    size_t extDictSize;
     const unsigned char* prefixEnd;
+    size_t extDictSize;
     size_t prefixSize;
 } LZ4_streamDecode_t_internal;
 
 #endif
 
-/*!
- * LZ4_stream_t :
- * information structure to track an LZ4 stream.
- * init this structure before first use.
- * note : only use in association with static linking !
- *        this definition is not API/ABI safe,
- *        and may change in a future version !
+/*! LZ4_stream_t :
+ *  information structure to track an LZ4 stream.
+ *  LZ4_stream_t can also be created using LZ4_createStream(), which is recommended.
+ *  The structure definition can be convenient for static allocation
+ *  (on stack, or as part of larger structure).
+ *  Init this structure with LZ4_initStream() before first use.
+ *  note : only use this definition in association with static linking !
+ *    this definition is not API/ABI safe, and may change in a future version.
  */
-#define LZ4_STREAMSIZE_U64 ((1 << (LZ4_MEMORY_USAGE-3)) + 4)
+#define LZ4_STREAMSIZE_U64 ((1 << (LZ4_MEMORY_USAGE-3)) + 4 + ((sizeof(void*)==16) ? 4 : 0) /*AS-400*/ )
 #define LZ4_STREAMSIZE     (LZ4_STREAMSIZE_U64 * sizeof(unsigned long long))
 union LZ4_stream_u {
     unsigned long long table[LZ4_STREAMSIZE_U64];
     LZ4_stream_t_internal internal_donotuse;
 } ;  /* previously typedef'd to LZ4_stream_t */
 
+/*! LZ4_initStream() : v1.9.0+
+ *  An LZ4_stream_t structure must be initialized at least once.
+ *  This is automatically done when invoking LZ4_createStream(),
+ *  but it's not when the structure is simply declared on stack (for example).
+ *
+ *  Use LZ4_initStream() to properly initialize a newly declared LZ4_stream_t.
+ *  It can also initialize any arbitrary buffer of sufficient size,
+ *  and will @return a pointer of proper type upon initialization.
+ *
+ *  Note : initialization fails if size and alignment conditions are not respected.
+ *         In which case, the function will @return NULL.
+ *  Note2: An LZ4_stream_t structure guarantees correct alignment and size.
+ *  Note3: Before v1.9.0, use LZ4_resetStream() instead
+ */
+LZ4LIB_API LZ4_stream_t* LZ4_initStream (void* buffer, size_t size);
+
 
-/*!
- * LZ4_streamDecode_t :
- * information structure to track an LZ4 stream during decompression.
- * init this structure  using LZ4_setStreamDecode (or memset()) before first use
- * note : only use in association with static linking !
- *        this definition is not API/ABI safe,
- *        and may change in a future version !
+/*! LZ4_streamDecode_t :
+ *  information structure to track an LZ4 stream during decompression.
+ *  init this structure  using LZ4_setStreamDecode() before first use.
+ *  note : only use in association with static linking !
+ *         this definition is not API/ABI safe,
+ *         and may change in a future version !
  */
-#define LZ4_STREAMDECODESIZE_U64  4
+#define LZ4_STREAMDECODESIZE_U64 (4 + ((sizeof(void*)==16) ? 2 : 0) /*AS-400*/ )
 #define LZ4_STREAMDECODESIZE     (LZ4_STREAMDECODESIZE_U64 * sizeof(unsigned long long))
 union LZ4_streamDecode_u {
     unsigned long long table[LZ4_STREAMDECODESIZE_U64];
@@ -400,15 +648,22 @@  union LZ4_streamDecode_u {
 } ;   /* previously typedef'd to LZ4_streamDecode_t */
 
 
-/*=************************************
+
+/*-************************************
 *  Obsolete Functions
 **************************************/
-/* Deprecation warnings */
-/* Should these warnings be a problem,
-   it is generally possible to disable them,
-   typically with -Wno-deprecated-declarations for gcc
-   or _CRT_SECURE_NO_WARNINGS in Visual.
-   Otherwise, it's also possible to define LZ4_DISABLE_DEPRECATE_WARNINGS */
+
+/*! Deprecation warnings
+ *
+ *  Deprecated functions make the compiler generate a warning when invoked.
+ *  This is meant to invite users to update their source code.
+ *  Should deprecation warnings be a problem, it is generally possible to disable them,
+ *  typically with -Wno-deprecated-declarations for gcc
+ *  or _CRT_SECURE_NO_WARNINGS in Visual.
+ *
+ *  Another method is to define LZ4_DISABLE_DEPRECATE_WARNINGS
+ *  before including the header file.
+ */
 #ifdef LZ4_DISABLE_DEPRECATE_WARNINGS
 #  define LZ4_DEPRECATED(message)   /* disable deprecation warnings */
 #else
@@ -428,36 +683,82 @@  union LZ4_streamDecode_u {
 #endif /* LZ4_DISABLE_DEPRECATE_WARNINGS */
 
 /* Obsolete compression functions */
-LZ4_DEPRECATED("use LZ4_compress_default() instead") int LZ4_compress               (const char* source, char* dest, int sourceSize);
-LZ4_DEPRECATED("use LZ4_compress_default() instead") int LZ4_compress_limitedOutput (const char* source, char* dest, int sourceSize, int maxOutputSize);
-LZ4_DEPRECATED("use LZ4_compress_fast_extState() instead") int LZ4_compress_withState               (void* state, const char* source, char* dest, int inputSize);
-LZ4_DEPRECATED("use LZ4_compress_fast_extState() instead") int LZ4_compress_limitedOutput_withState (void* state, const char* source, char* dest, int inputSize, int maxOutputSize);
-LZ4_DEPRECATED("use LZ4_compress_fast_continue() instead") int LZ4_compress_continue                (LZ4_stream_t* LZ4_streamPtr, const char* source, char* dest, int inputSize);
-LZ4_DEPRECATED("use LZ4_compress_fast_continue() instead") int LZ4_compress_limitedOutput_continue  (LZ4_stream_t* LZ4_streamPtr, const char* source, char* dest, int inputSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_default() instead")       LZ4LIB_API int LZ4_compress               (const char* src, char* dest, int srcSize);
+LZ4_DEPRECATED("use LZ4_compress_default() instead")       LZ4LIB_API int LZ4_compress_limitedOutput (const char* src, char* dest, int srcSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_fast_extState() instead") LZ4LIB_API int LZ4_compress_withState               (void* state, const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_fast_extState() instead") LZ4LIB_API int LZ4_compress_limitedOutput_withState (void* state, const char* source, char* dest, int inputSize, int maxOutputSize);
+LZ4_DEPRECATED("use LZ4_compress_fast_continue() instead") LZ4LIB_API int LZ4_compress_continue                (LZ4_stream_t* LZ4_streamPtr, const char* source, char* dest, int inputSize);
+LZ4_DEPRECATED("use LZ4_compress_fast_continue() instead") LZ4LIB_API int LZ4_compress_limitedOutput_continue  (LZ4_stream_t* LZ4_streamPtr, const char* source, char* dest, int inputSize, int maxOutputSize);
 
 /* Obsolete decompression functions */
-/* These function names are completely deprecated and must no longer be used.
-   They are only provided in lz4.c for compatibility with older programs.
-    - LZ4_uncompress is the same as LZ4_decompress_fast
-    - LZ4_uncompress_unknownOutputSize is the same as LZ4_decompress_safe
-   These function prototypes are now disabled; uncomment them only if you really need them.
-   It is highly recommended to stop using these prototypes and migrate to maintained ones */
-/* int LZ4_uncompress (const char* source, char* dest, int outputSize); */
-/* int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize); */
+LZ4_DEPRECATED("use LZ4_decompress_fast() instead") LZ4LIB_API int LZ4_uncompress (const char* source, char* dest, int outputSize);
+LZ4_DEPRECATED("use LZ4_decompress_safe() instead") LZ4LIB_API int LZ4_uncompress_unknownOutputSize (const char* source, char* dest, int isize, int maxOutputSize);
 
-/* Obsolete streaming functions; use new streaming interface whenever possible */
-LZ4_DEPRECATED("use LZ4_createStream() instead") void* LZ4_create (char* inputBuffer);
-LZ4_DEPRECATED("use LZ4_createStream() instead") int   LZ4_sizeofStreamState(void);
-LZ4_DEPRECATED("use LZ4_resetStream() instead")  int   LZ4_resetStreamState(void* state, char* inputBuffer);
-LZ4_DEPRECATED("use LZ4_saveDict() instead")     char* LZ4_slideInputBuffer (void* state);
+/* Obsolete streaming functions; degraded functionality; do not use!
+ *
+ * In order to perform streaming compression, these functions depended on data
+ * that is no longer tracked in the state. They have been preserved as well as
+ * possible: using them will still produce a correct output. However, they don't
+ * actually retain any history between compression calls. The compression ratio
+ * achieved will therefore be no better than compressing each chunk
+ * independently.
+ */
+LZ4_DEPRECATED("Use LZ4_createStream() instead") LZ4LIB_API void* LZ4_create (char* inputBuffer);
+LZ4_DEPRECATED("Use LZ4_createStream() instead") LZ4LIB_API int   LZ4_sizeofStreamState(void);
+LZ4_DEPRECATED("Use LZ4_resetStream() instead")  LZ4LIB_API int   LZ4_resetStreamState(void* state, char* inputBuffer);
+LZ4_DEPRECATED("Use LZ4_saveDict() instead")     LZ4LIB_API char* LZ4_slideInputBuffer (void* state);
 
 /* Obsolete streaming decoding functions */
-LZ4_DEPRECATED("use LZ4_decompress_safe_usingDict() instead") int LZ4_decompress_safe_withPrefix64k (const char* src, char* dst, int compressedSize, int maxDstSize);
-LZ4_DEPRECATED("use LZ4_decompress_fast_usingDict() instead") int LZ4_decompress_fast_withPrefix64k (const char* src, char* dst, int originalSize);
+LZ4_DEPRECATED("use LZ4_decompress_safe_usingDict() instead") LZ4LIB_API int LZ4_decompress_safe_withPrefix64k (const char* src, char* dst, int compressedSize, int maxDstSize);
+LZ4_DEPRECATED("use LZ4_decompress_fast_usingDict() instead") LZ4LIB_API int LZ4_decompress_fast_withPrefix64k (const char* src, char* dst, int originalSize);
+
+/*! LZ4_decompress_fast() : **unsafe!**
+ *  These functions used to be faster than LZ4_decompress_safe(),
+ *  but it has changed, and they are now slower than LZ4_decompress_safe().
+ *  This is because LZ4_decompress_fast() doesn't know the input size,
+ *  and therefore must progress more cautiously in the input buffer to not read beyond the end of block.
+ *  On top of that `LZ4_decompress_fast()` is not protected vs malformed or malicious inputs, making it a security liability.
+ *  As a consequence, LZ4_decompress_fast() is strongly discouraged, and deprecated.
+ *
+ *  The last remaining LZ4_decompress_fast() specificity is that
+ *  it can decompress a block without knowing its compressed size.
+ *  Such functionality could be achieved in a more secure manner,
+ *  by also providing the maximum size of input buffer,
+ *  but it would require new prototypes, and adaptation of the implementation to this new use case.
+ *
+ *  Parameters:
+ *  originalSize : is the uncompressed size to regenerate.
+ *                 `dst` must be already allocated, its size must be >= 'originalSize' bytes.
+ * @return : number of bytes read from source buffer (== compressed size).
+ *           The function expects to finish at block's end exactly.
+ *           If the source stream is detected malformed, the function stops decoding and returns a negative result.
+ *  note : LZ4_decompress_fast*() requires originalSize. Thanks to this information, it never writes past the output buffer.
+ *         However, since it doesn't know its 'src' size, it may read an unknown amount of input, past input buffer bounds.
+ *         Also, since match offsets are not validated, match reads from 'src' may underflow too.
+ *         These issues never happen if input (compressed) data is correct.
+ *         But they may happen if input data is invalid (error or intentional tampering).
+ *         As a consequence, use these functions in trusted environments with trusted data **only**.
+ */
+
+LZ4_DEPRECATED("This function is deprecated and unsafe. Consider using LZ4_decompress_safe() instead")
+LZ4LIB_API int LZ4_decompress_fast (const char* src, char* dst, int originalSize);
+LZ4_DEPRECATED("This function is deprecated and unsafe. Consider using LZ4_decompress_safe_continue() instead")
+LZ4LIB_API int LZ4_decompress_fast_continue (LZ4_streamDecode_t* LZ4_streamDecode, const char* src, char* dst, int originalSize);
+LZ4_DEPRECATED("This function is deprecated and unsafe. Consider using LZ4_decompress_safe_usingDict() instead")
+LZ4LIB_API int LZ4_decompress_fast_usingDict (const char* src, char* dst, int originalSize, const char* dictStart, int dictSize);
+
+/*! LZ4_resetStream() :
+ *  An LZ4_stream_t structure must be initialized at least once.
+ *  This is done with LZ4_initStream(), or LZ4_resetStream().
+ *  Consider switching to LZ4_initStream(),
+ *  invoking LZ4_resetStream() will trigger deprecation warnings in the future.
+ */
+LZ4LIB_API void LZ4_resetStream (LZ4_stream_t* streamPtr);
+
+
+#endif /* LZ4_H_98237428734687 */
 
 
 #if defined (__cplusplus)
 }
 #endif
-
-#endif /* LZ4_H_2983827168210 */