Merge pull request #33507 from SneakyFish5/update-zstd

Update zstd to 1.4.4

27 files changed, 1850 insertions, 829 deletions

diff --git a/thirdparty/README.md b/thirdparty/README.md
index 8928009817..4f97d9e74a 100644
--- a/thirdparty/README.md
+++ b/thirdparty/README.md
@@ -534,7 +534,7 @@ Files extracted from upstream source:
 ## zstd
 
 - Upstream: https://github.com/facebook/zstd
-- Version: 1.4.3
+- Version: 1.4.4
 - License: BSD-3-Clause
 
 Files extracted from upstream source:
diff --git a/thirdparty/zstd/common/bitstream.h b/thirdparty/zstd/common/bitstream.h
index 7bdb060460..1c294b80d1 100644
--- a/thirdparty/zstd/common/bitstream.h
+++ b/thirdparty/zstd/common/bitstream.h
@@ -164,7 +164,7 @@ MEM_STATIC unsigned BIT_highbit32 (U32 val)
         _BitScanReverse ( &r, val );
         return (unsigned) r;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* Use GCC Intrinsic */
-        return 31 - __builtin_clz (val);
+        return __builtin_clz (val) ^ 31;
 #   elif defined(__ICCARM__)    /* IAR Intrinsic */
         return 31 - __CLZ(val);
 #   else   /* Software version */
@@ -244,9 +244,9 @@ MEM_STATIC void BIT_flushBitsFast(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
     assert(bitC->bitPos < sizeof(bitC->bitContainer) * 8);
+    assert(bitC->ptr <= bitC->endPtr);
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
-    assert(bitC->ptr <= bitC->endPtr);
     bitC->bitPos &= 7;
     bitC->bitContainer >>= nbBytes*8;
 }
@@ -260,6 +260,7 @@ MEM_STATIC void BIT_flushBits(BIT_CStream_t* bitC)
 {
     size_t const nbBytes = bitC->bitPos >> 3;
     assert(bitC->bitPos < sizeof(bitC->bitContainer) * 8);
+    assert(bitC->ptr <= bitC->endPtr);
     MEM_writeLEST(bitC->ptr, bitC->bitContainer);
     bitC->ptr += nbBytes;
     if (bitC->ptr > bitC->endPtr) bitC->ptr = bitC->endPtr;
diff --git a/thirdparty/zstd/common/compiler.h b/thirdparty/zstd/common/compiler.h
index 6686b837d6..1877a0c1d9 100644
--- a/thirdparty/zstd/common/compiler.h
+++ b/thirdparty/zstd/common/compiler.h
@@ -61,6 +61,13 @@
 #  define HINT_INLINE static INLINE_KEYWORD FORCE_INLINE_ATTR
 #endif
 
+/* UNUSED_ATTR tells the compiler it is okay if the function is unused. */
+#if defined(__GNUC__)
+#  define UNUSED_ATTR __attribute__((unused))
+#else
+#  define UNUSED_ATTR
+#endif
+
 /* force no inlining */
 #ifdef _MSC_VER
 #  define FORCE_NOINLINE static __declspec(noinline)
@@ -127,9 +134,14 @@
     }                                     \
 }
 
-/* vectorization */
+/* vectorization
+ * older GCC (pre gcc-4.3 picked as the cutoff) uses a different syntax */
 #if !defined(__clang__) && defined(__GNUC__)
-#  define DONT_VECTORIZE __attribute__((optimize("no-tree-vectorize")))
+#  if (__GNUC__ == 4 && __GNUC_MINOR__ > 3) || (__GNUC__ >= 5)
+#    define DONT_VECTORIZE __attribute__((optimize("no-tree-vectorize")))
+#  else
+#    define DONT_VECTORIZE _Pragma("GCC optimize(\"no-tree-vectorize\")")
+#  endif
 #else
 #  define DONT_VECTORIZE
 #endif
diff --git a/thirdparty/zstd/common/fse.h b/thirdparty/zstd/common/fse.h
index 811c670bdd..a7553e3721 100644
--- a/thirdparty/zstd/common/fse.h
+++ b/thirdparty/zstd/common/fse.h
@@ -308,7 +308,7 @@ If there is an error, the function will return an error code, which can be teste
 *******************************************/
 /* FSE buffer bounds */
 #define FSE_NCOUNTBOUND 512
-#define FSE_BLOCKBOUND(size) (size + (size>>7))
+#define FSE_BLOCKBOUND(size) (size + (size>>7) + 4 /* fse states */ + sizeof(size_t) /* bitContainer */)
 #define FSE_COMPRESSBOUND(size) (FSE_NCOUNTBOUND + FSE_BLOCKBOUND(size))   /* Macro version, useful for static allocation */
 
 /* It is possible to statically allocate FSE CTable/DTable as a table of FSE_CTable/FSE_DTable using below macros */
diff --git a/thirdparty/zstd/common/fse_decompress.c b/thirdparty/zstd/common/fse_decompress.c
index 72bbead5be..4f07378982 100644
--- a/thirdparty/zstd/common/fse_decompress.c
+++ b/thirdparty/zstd/common/fse_decompress.c
@@ -52,7 +52,9 @@
 #define FSE_STATIC_ASSERT(c) DEBUG_STATIC_ASSERT(c)   /* use only *after* variable declarations */
 
 /* check and forward error code */
+#ifndef CHECK_F
 #define CHECK_F(f) { size_t const e = f; if (FSE_isError(e)) return e; }
+#endif
 
 
 /* **************************************************************
diff --git a/thirdparty/zstd/common/mem.h b/thirdparty/zstd/common/mem.h
index c10d7f61e1..530d30c8f7 100644
--- a/thirdparty/zstd/common/mem.h
+++ b/thirdparty/zstd/common/mem.h
@@ -47,6 +47,79 @@ extern "C" {
 #define MEM_STATIC_ASSERT(c)   { enum { MEM_static_assert = 1/(int)(!!(c)) }; }
 MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (sizeof(size_t)==8)); }
 
+/* detects whether we are being compiled under msan */
+#if defined (__has_feature)
+#  if __has_feature(memory_sanitizer)
+#    define MEMORY_SANITIZER 1
+#  endif
+#endif
+
+#if defined (MEMORY_SANITIZER)
+/* Not all platforms that support msan provide sanitizers/msan_interface.h.
+ * We therefore declare the functions we need ourselves, rather than trying to
+ * include the header file... */
+
+#include <stdint.h> /* intptr_t */
+
+/* Make memory region fully initialized (without changing its contents). */
+void __msan_unpoison(const volatile void *a, size_t size);
+
+/* Make memory region fully uninitialized (without changing its contents).
+   This is a legacy interface that does not update origin information. Use
+   __msan_allocated_memory() instead. */
+void __msan_poison(const volatile void *a, size_t size);
+
+/* Returns the offset of the first (at least partially) poisoned byte in the
+   memory range, or -1 if the whole range is good. */
+intptr_t __msan_test_shadow(const volatile void *x, size_t size);
+#endif
+
+/* detects whether we are being compiled under asan */
+#if defined (__has_feature)
+#  if __has_feature(address_sanitizer)
+#    define ADDRESS_SANITIZER 1
+#  endif
+#elif defined(__SANITIZE_ADDRESS__)
+#  define ADDRESS_SANITIZER 1
+#endif
+
+#if defined (ADDRESS_SANITIZER)
+/* Not all platforms that support asan provide sanitizers/asan_interface.h.
+ * We therefore declare the functions we need ourselves, rather than trying to
+ * include the header file... */
+
+/**
+ * Marks a memory region (<c>[addr, addr+size)</c>) as unaddressable.
+ *
+ * This memory must be previously allocated by your program. Instrumented
+ * code is forbidden from accessing addresses in this region until it is
+ * unpoisoned. This function is not guaranteed to poison the entire region -
+ * it could poison only a subregion of <c>[addr, addr+size)</c> due to ASan
+ * alignment restrictions.
+ *
+ * \note This function is not thread-safe because no two threads can poison or
+ * unpoison memory in the same memory region simultaneously.
+ *
+ * \param addr Start of memory region.
+ * \param size Size of memory region. */
+void __asan_poison_memory_region(void const volatile *addr, size_t size);
+
+/**
+ * Marks a memory region (<c>[addr, addr+size)</c>) as addressable.
+ *
+ * This memory must be previously allocated by your program. Accessing
+ * addresses in this region is allowed until this region is poisoned again.
+ * This function could unpoison a super-region of <c>[addr, addr+size)</c> due
+ * to ASan alignment restrictions.
+ *
+ * \note This function is not thread-safe because no two threads can
+ * poison or unpoison memory in the same memory region simultaneously.
+ *
+ * \param addr Start of memory region.
+ * \param size Size of memory region. */
+void __asan_unpoison_memory_region(void const volatile *addr, size_t size);
+#endif
+
 
 /*-**************************************************************
 *  Basic Types
diff --git a/thirdparty/zstd/common/pool.c b/thirdparty/zstd/common/pool.c
index 7a82945432..f575935076 100644
--- a/thirdparty/zstd/common/pool.c
+++ b/thirdparty/zstd/common/pool.c
@@ -127,9 +127,13 @@ POOL_ctx* POOL_create_advanced(size_t numThreads, size_t queueSize,
     ctx->queueTail = 0;
     ctx->numThreadsBusy = 0;
     ctx->queueEmpty = 1;
-    (void)ZSTD_pthread_mutex_init(&ctx->queueMutex, NULL);
-    (void)ZSTD_pthread_cond_init(&ctx->queuePushCond, NULL);
-    (void)ZSTD_pthread_cond_init(&ctx->queuePopCond, NULL);
+    {
+        int error = 0;
+        error |= ZSTD_pthread_mutex_init(&ctx->queueMutex, NULL);
+        error |= ZSTD_pthread_cond_init(&ctx->queuePushCond, NULL);
+        error |= ZSTD_pthread_cond_init(&ctx->queuePopCond, NULL);
+        if (error) { POOL_free(ctx); return NULL; }
+    }
     ctx->shutdown = 0;
     /* Allocate space for the thread handles */
     ctx->threads = (ZSTD_pthread_t*)ZSTD_malloc(numThreads * sizeof(ZSTD_pthread_t), customMem);
diff --git a/thirdparty/zstd/common/threading.c b/thirdparty/zstd/common/threading.c
index f3d4fa8418..482664bd9a 100644
--- a/thirdparty/zstd/common/threading.c
+++ b/thirdparty/zstd/common/threading.c
@@ -14,6 +14,8 @@
  * This file will hold wrapper for systems, which do not support pthreads
  */
 
+#include "threading.h"
+
 /* create fake symbol to avoid empty translation unit warning */
 int g_ZSTD_threading_useless_symbol;
 
@@ -28,7 +30,6 @@ int g_ZSTD_threading_useless_symbol;
 /* ===  Dependencies  === */
 #include <process.h>
 #include <errno.h>
-#include "threading.h"
 
 
 /* ===  Implementation  === */
@@ -73,3 +74,47 @@ int ZSTD_pthread_join(ZSTD_pthread_t thread, void **value_ptr)
 }
 
 #endif   /* ZSTD_MULTITHREAD */
+
+#if defined(ZSTD_MULTITHREAD) && DEBUGLEVEL >= 1 && !defined(_WIN32)
+
+#include <stdlib.h>
+
+int ZSTD_pthread_mutex_init(ZSTD_pthread_mutex_t* mutex, pthread_mutexattr_t const* attr)
+{
+    *mutex = (pthread_mutex_t*)malloc(sizeof(pthread_mutex_t));
+    if (!*mutex)
+        return 1;
+    return pthread_mutex_init(*mutex, attr);
+}
+
+int ZSTD_pthread_mutex_destroy(ZSTD_pthread_mutex_t* mutex)
+{
+    if (!*mutex)
+        return 0;
+    {
+        int const ret = pthread_mutex_destroy(*mutex);
+        free(*mutex);
+        return ret;
+    }
+}
+
+int ZSTD_pthread_cond_init(ZSTD_pthread_cond_t* cond, pthread_condattr_t const* attr)
+{
+    *cond = (pthread_cond_t*)malloc(sizeof(pthread_cond_t));
+    if (!*cond)
+        return 1;
+    return pthread_cond_init(*cond, attr);
+}
+
+int ZSTD_pthread_cond_destroy(ZSTD_pthread_cond_t* cond)
+{
+    if (!*cond)
+        return 0;
+    {
+        int const ret = pthread_cond_destroy(*cond);
+        free(*cond);
+        return ret;
+    }
+}
+
+#endif
diff --git a/thirdparty/zstd/common/threading.h b/thirdparty/zstd/common/threading.h
index d806c89d01..3193ca7db8 100644
--- a/thirdparty/zstd/common/threading.h
+++ b/thirdparty/zstd/common/threading.h
@@ -13,6 +13,8 @@
 #ifndef THREADING_H_938743
 #define THREADING_H_938743
 
+#include "debug.h"
+
 #if defined (__cplusplus)
 extern "C" {
 #endif
@@ -75,10 +77,12 @@ int ZSTD_pthread_join(ZSTD_pthread_t thread, void** value_ptr);
  */
 
 
-#elif defined(ZSTD_MULTITHREAD)   /* posix assumed ; need a better detection method */
+#elif defined(ZSTD_MULTITHREAD)    /* posix assumed ; need a better detection method */
 /* ===   POSIX Systems   === */
 #  include <pthread.h>
 
+#if DEBUGLEVEL < 1
+
 #define ZSTD_pthread_mutex_t            pthread_mutex_t
 #define ZSTD_pthread_mutex_init(a, b)   pthread_mutex_init((a), (b))
 #define ZSTD_pthread_mutex_destroy(a)   pthread_mutex_destroy((a))
@@ -96,6 +100,33 @@ int ZSTD_pthread_join(ZSTD_pthread_t thread, void** value_ptr);
 #define ZSTD_pthread_create(a, b, c, d) pthread_create((a), (b), (c), (d))
 #define ZSTD_pthread_join(a, b)         pthread_join((a),(b))
 
+#else /* DEBUGLEVEL >= 1 */
+
+/* Debug implementation of threading.
+ * In this implementation we use pointers for mutexes and condition variables.
+ * This way, if we forget to init/destroy them the program will crash or ASAN
+ * will report leaks.
+ */
+
+#define ZSTD_pthread_mutex_t            pthread_mutex_t*
+int ZSTD_pthread_mutex_init(ZSTD_pthread_mutex_t* mutex, pthread_mutexattr_t const* attr);
+int ZSTD_pthread_mutex_destroy(ZSTD_pthread_mutex_t* mutex);
+#define ZSTD_pthread_mutex_lock(a)      pthread_mutex_lock(*(a))
+#define ZSTD_pthread_mutex_unlock(a)    pthread_mutex_unlock(*(a))
+
+#define ZSTD_pthread_cond_t             pthread_cond_t*
+int ZSTD_pthread_cond_init(ZSTD_pthread_cond_t* cond, pthread_condattr_t const* attr);
+int ZSTD_pthread_cond_destroy(ZSTD_pthread_cond_t* cond);
+#define ZSTD_pthread_cond_wait(a, b)    pthread_cond_wait(*(a), *(b))
+#define ZSTD_pthread_cond_signal(a)     pthread_cond_signal(*(a))
+#define ZSTD_pthread_cond_broadcast(a)  pthread_cond_broadcast(*(a))
+
+#define ZSTD_pthread_t                  pthread_t
+#define ZSTD_pthread_create(a, b, c, d) pthread_create((a), (b), (c), (d))
+#define ZSTD_pthread_join(a, b)         pthread_join((a),(b))
+
+#endif
+
 #else  /* ZSTD_MULTITHREAD not defined */
 /* No multithreading support */
 
diff --git a/thirdparty/zstd/common/zstd_internal.h b/thirdparty/zstd/common/zstd_internal.h
index 585fd6b19e..dcdcbdb81c 100644
--- a/thirdparty/zstd/common/zstd_internal.h
+++ b/thirdparty/zstd/common/zstd_internal.h
@@ -197,79 +197,56 @@ static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
 static void ZSTD_copy16(void* dst, const void* src) { memcpy(dst, src, 16); }
 #define COPY16(d,s) { ZSTD_copy16(d,s); d+=16; s+=16; }
 
-#define WILDCOPY_OVERLENGTH 8
-#define VECLEN 16
+#define WILDCOPY_OVERLENGTH 32
+#define WILDCOPY_VECLEN 16
 
 typedef enum {
     ZSTD_no_overlap,
-    ZSTD_overlap_src_before_dst,
+    ZSTD_overlap_src_before_dst
     /*  ZSTD_overlap_dst_before_src, */
 } ZSTD_overlap_e;
 
 /*! ZSTD_wildcopy() :
- *  custom version of memcpy(), can overwrite up to WILDCOPY_OVERLENGTH bytes (if length==0) */
+ *  Custom version of memcpy(), can over read/write up to WILDCOPY_OVERLENGTH bytes (if length==0)
+ *  @param ovtype controls the overlap detection
+ *         - ZSTD_no_overlap: The source and destination are guaranteed to be at least WILDCOPY_VECLEN bytes apart.
+ *         - ZSTD_overlap_src_before_dst: The src and dst may overlap, but they MUST be at least 8 bytes apart.
+ *           The src buffer must be before the dst buffer.
+ */
 MEM_STATIC FORCE_INLINE_ATTR DONT_VECTORIZE
-void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e ovtype)
+void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e const ovtype)
 {
     ptrdiff_t diff = (BYTE*)dst - (const BYTE*)src;
     const BYTE* ip = (const BYTE*)src;
     BYTE* op = (BYTE*)dst;
     BYTE* const oend = op + length;
 
-    assert(diff >= 8 || (ovtype == ZSTD_no_overlap && diff < -8));
-    if (length < VECLEN || (ovtype == ZSTD_overlap_src_before_dst && diff < VECLEN)) {
-      do
-          COPY8(op, ip)
-      while (op < oend);
-    }
-    else {
-      if ((length & 8) == 0)
-        COPY8(op, ip);
-      do {
+    assert(diff >= 8 || (ovtype == ZSTD_no_overlap && diff <= -WILDCOPY_VECLEN));
+
+    if (ovtype == ZSTD_overlap_src_before_dst && diff < WILDCOPY_VECLEN) {
+        /* Handle short offset copies. */
+        do {
+            COPY8(op, ip)
+        } while (op < oend);
+    } else {
+        assert(diff >= WILDCOPY_VECLEN || diff <= -WILDCOPY_VECLEN);
+        /* Separate out the first two COPY16() calls because the copy length is
+         * almost certain to be short, so the branches have different
+         * probabilities.
+         * On gcc-9 unrolling once is +1.6%, twice is +2%, thrice is +1.8%.
+         * On clang-8 unrolling once is +1.4%, twice is +3.3%, thrice is +3%.
+         */
         COPY16(op, ip);
-      }
-      while (op < oend);
-    }
-}
-
-/*! ZSTD_wildcopy_16min() :
- *  same semantics as ZSTD_wilcopy() except guaranteed to be able to copy 16 bytes at the start */
-MEM_STATIC FORCE_INLINE_ATTR DONT_VECTORIZE
-void ZSTD_wildcopy_16min(void* dst, const void* src, ptrdiff_t length, ZSTD_overlap_e ovtype)
-{
-    ptrdiff_t diff = (BYTE*)dst - (const BYTE*)src;
-    const BYTE* ip = (const BYTE*)src;
-    BYTE* op = (BYTE*)dst;
-    BYTE* const oend = op + length;
-
-    assert(length >= 8);
-    assert(diff >= 8 || (ovtype == ZSTD_no_overlap && diff < -8));
-
-    if (ovtype == ZSTD_overlap_src_before_dst && diff < VECLEN) {
-      do
-          COPY8(op, ip)
-      while (op < oend);
-    }
-    else {
-      if ((length & 8) == 0)
-        COPY8(op, ip);
-      do {
         COPY16(op, ip);
-      }
-      while (op < oend);
+        if (op >= oend) return;
+        do {
+            COPY16(op, ip);
+            COPY16(op, ip);
+        }
+        while (op < oend);
     }
 }
 
-MEM_STATIC void ZSTD_wildcopy_e(void* dst, const void* src, void* dstEnd)   /* should be faster for decoding, but strangely, not verified on all platform */
-{
-    const BYTE* ip = (const BYTE*)src;
-    BYTE* op = (BYTE*)dst;
-    BYTE* const oend = (BYTE*)dstEnd;
-    do
-        COPY8(op, ip)
-    while (op < oend);
-}
-
 
 /*-*******************************************
 *  Private declarations
@@ -323,7 +300,7 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)   /* compress, dictBuilder, decodeCorpus
         _BitScanReverse(&r, val);
         return (unsigned)r;
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* GCC Intrinsic */
-        return 31 - __builtin_clz(val);
+        return __builtin_clz (val) ^ 31;
 #   elif defined(__ICCARM__)    /* IAR Intrinsic */
         return 31 - __CLZ(val);
 #   else   /* Software version */
diff --git a/thirdparty/zstd/compress/zstd_compress.c b/thirdparty/zstd/compress/zstd_compress.c
index cd73db13be..35346b92cb 100644
--- a/thirdparty/zstd/compress/zstd_compress.c
+++ b/thirdparty/zstd/compress/zstd_compress.c
@@ -42,15 +42,15 @@ size_t ZSTD_compressBound(size_t srcSize) {
 *  Context memory management
 ***************************************/
 struct ZSTD_CDict_s {
-    void* dictBuffer;
     const void* dictContent;
     size_t dictContentSize;
-    void* workspace;
-    size_t workspaceSize;
+    U32* entropyWorkspace; /* entropy workspace of HUF_WORKSPACE_SIZE bytes */
+    ZSTD_cwksp workspace;
     ZSTD_matchState_t matchState;
     ZSTD_compressedBlockState_t cBlockState;
     ZSTD_customMem customMem;
     U32 dictID;
+    int compressionLevel; /* 0 indicates that advanced API was used to select CDict params */
 };  /* typedef'd to ZSTD_CDict within "zstd.h" */
 
 ZSTD_CCtx* ZSTD_createCCtx(void)
@@ -84,23 +84,26 @@ ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem)
 
 ZSTD_CCtx* ZSTD_initStaticCCtx(void *workspace, size_t workspaceSize)
 {
-    ZSTD_CCtx* const cctx = (ZSTD_CCtx*) workspace;
+    ZSTD_cwksp ws;
+    ZSTD_CCtx* cctx;
     if (workspaceSize <= sizeof(ZSTD_CCtx)) return NULL;  /* minimum size */
     if ((size_t)workspace & 7) return NULL;  /* must be 8-aligned */
-    memset(workspace, 0, workspaceSize);   /* may be a bit generous, could memset be smaller ? */
+    ZSTD_cwksp_init(&ws, workspace, workspaceSize);
+
+    cctx = (ZSTD_CCtx*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CCtx));
+    if (cctx == NULL) {
+        return NULL;
+    }
+    memset(cctx, 0, sizeof(ZSTD_CCtx));
+    ZSTD_cwksp_move(&cctx->workspace, &ws);
     cctx->staticSize = workspaceSize;
-    cctx->workSpace = (void*)(cctx+1);
-    cctx->workSpaceSize = workspaceSize - sizeof(ZSTD_CCtx);
 
     /* statically sized space. entropyWorkspace never moves (but prev/next block swap places) */
-    if (cctx->workSpaceSize < HUF_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t)) return NULL;
-    assert(((size_t)cctx->workSpace & (sizeof(void*)-1)) == 0);   /* ensure correct alignment */
-    cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)cctx->workSpace;
-    cctx->blockState.nextCBlock = cctx->blockState.prevCBlock + 1;
-    {
-        void* const ptr = cctx->blockState.nextCBlock + 1;
-        cctx->entropyWorkspace = (U32*)ptr;
-    }
+    if (!ZSTD_cwksp_check_available(&cctx->workspace, HUF_WORKSPACE_SIZE + 2 * sizeof(ZSTD_compressedBlockState_t))) return NULL;
+    cctx->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
+    cctx->blockState.nextCBlock = (ZSTD_compressedBlockState_t*)ZSTD_cwksp_reserve_object(&cctx->workspace, sizeof(ZSTD_compressedBlockState_t));
+    cctx->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(
+        &cctx->workspace, HUF_WORKSPACE_SIZE);
     cctx->bmi2 = ZSTD_cpuid_bmi2(ZSTD_cpuid());
     return cctx;
 }
@@ -128,11 +131,11 @@ static void ZSTD_freeCCtxContent(ZSTD_CCtx* cctx)
 {
     assert(cctx != NULL);
     assert(cctx->staticSize == 0);
-    ZSTD_free(cctx->workSpace, cctx->customMem); cctx->workSpace = NULL;
     ZSTD_clearAllDicts(cctx);
 #ifdef ZSTD_MULTITHREAD
     ZSTDMT_freeCCtx(cctx->mtctx); cctx->mtctx = NULL;
 #endif
+    ZSTD_cwksp_free(&cctx->workspace, cctx->customMem);
 }
 
 size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)
@@ -140,8 +143,13 @@ size_t ZSTD_freeCCtx(ZSTD_CCtx* cctx)
     if (cctx==NULL) return 0;   /* support free on NULL */
     RETURN_ERROR_IF(cctx->staticSize, memory_allocation,
                     "not compatible with static CCtx");
-    ZSTD_freeCCtxContent(cctx);
-    ZSTD_free(cctx, cctx->customMem);
+    {
+        int cctxInWorkspace = ZSTD_cwksp_owns_buffer(&cctx->workspace, cctx);
+        ZSTD_freeCCtxContent(cctx);
+        if (!cctxInWorkspace) {
+            ZSTD_free(cctx, cctx->customMem);
+        }
+    }
     return 0;
 }
 
@@ -160,7 +168,9 @@ static size_t ZSTD_sizeof_mtctx(const ZSTD_CCtx* cctx)
 size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx)
 {
     if (cctx==NULL) return 0;   /* support sizeof on NULL */
-    return sizeof(*cctx) + cctx->workSpaceSize
+    /* cctx may be in the workspace */
+    return (cctx->workspace.workspace == cctx ? 0 : sizeof(*cctx))
+           + ZSTD_cwksp_sizeof(&cctx->workspace)
            + ZSTD_sizeof_localDict(cctx->localDict)
            + ZSTD_sizeof_mtctx(cctx);
 }
@@ -229,23 +239,23 @@ size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_paramete
     RETURN_ERROR_IF(!cctxParams, GENERIC);
     FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) );
     memset(cctxParams, 0, sizeof(*cctxParams));
+    assert(!ZSTD_checkCParams(params.cParams));
     cctxParams->cParams = params.cParams;
     cctxParams->fParams = params.fParams;
     cctxParams->compressionLevel = ZSTD_CLEVEL_DEFAULT;   /* should not matter, as all cParams are presumed properly defined */
-    assert(!ZSTD_checkCParams(params.cParams));
     return 0;
 }
 
 /* ZSTD_assignParamsToCCtxParams() :
  * params is presumed valid at this stage */
 static ZSTD_CCtx_params ZSTD_assignParamsToCCtxParams(
-        ZSTD_CCtx_params cctxParams, ZSTD_parameters params)
+        const ZSTD_CCtx_params* cctxParams, ZSTD_parameters params)
 {
-    ZSTD_CCtx_params ret = cctxParams;
+    ZSTD_CCtx_params ret = *cctxParams;
+    assert(!ZSTD_checkCParams(params.cParams));
     ret.cParams = params.cParams;
     ret.fParams = params.fParams;
     ret.compressionLevel = ZSTD_CLEVEL_DEFAULT;   /* should not matter, as all cParams are presumed properly defined */
-    assert(!ZSTD_checkCParams(params.cParams));
     return ret;
 }
 
@@ -378,7 +388,7 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
     case ZSTD_c_forceAttachDict:
         ZSTD_STATIC_ASSERT(ZSTD_dictDefaultAttach < ZSTD_dictForceCopy);
         bounds.lowerBound = ZSTD_dictDefaultAttach;
-        bounds.upperBound = ZSTD_dictForceCopy;       /* note : how to ensure at compile time that this is the highest value enum ? */
+        bounds.upperBound = ZSTD_dictForceLoad;       /* note : how to ensure at compile time that this is the highest value enum ? */
         return bounds;
 
     case ZSTD_c_literalCompressionMode:
@@ -392,6 +402,11 @@ ZSTD_bounds ZSTD_cParam_getBounds(ZSTD_cParameter param)
         bounds.upperBound = ZSTD_TARGETCBLOCKSIZE_MAX;
         return bounds;
 
+    case ZSTD_c_srcSizeHint:
+        bounds.lowerBound = ZSTD_SRCSIZEHINT_MIN;
+        bounds.upperBound = ZSTD_SRCSIZEHINT_MAX;
+        return bounds;
+
     default:
         {   ZSTD_bounds const boundError = { ERROR(parameter_unsupported), 0, 0 };
             return boundError;
@@ -448,6 +463,7 @@ static int ZSTD_isUpdateAuthorized(ZSTD_cParameter param)
     case ZSTD_c_forceAttachDict:
     case ZSTD_c_literalCompressionMode:
     case ZSTD_c_targetCBlockSize:
+    case ZSTD_c_srcSizeHint:
     default:
         return 0;
     }
@@ -494,6 +510,7 @@ size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, int value)
     case ZSTD_c_ldmMinMatch:
     case ZSTD_c_ldmBucketSizeLog:
     case ZSTD_c_targetCBlockSize:
+    case ZSTD_c_srcSizeHint:
         break;
 
     default: RETURN_ERROR(parameter_unsupported);
@@ -517,33 +534,33 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         if (value) {  /* 0 : does not change current level */
             CCtxParams->compressionLevel = value;
         }
-        if (CCtxParams->compressionLevel >= 0) return CCtxParams->compressionLevel;
+        if (CCtxParams->compressionLevel >= 0) return (size_t)CCtxParams->compressionLevel;
         return 0;  /* return type (size_t) cannot represent negative values */
     }
 
     case ZSTD_c_windowLog :
         if (value!=0)   /* 0 => use default */
             BOUNDCHECK(ZSTD_c_windowLog, value);
-        CCtxParams->cParams.windowLog = value;
+        CCtxParams->cParams.windowLog = (U32)value;
         return CCtxParams->cParams.windowLog;
 
     case ZSTD_c_hashLog :
         if (value!=0)   /* 0 => use default */
             BOUNDCHECK(ZSTD_c_hashLog, value);
-        CCtxParams->cParams.hashLog = value;
+        CCtxParams->cParams.hashLog = (U32)value;
         return CCtxParams->cParams.hashLog;
 
     case ZSTD_c_chainLog :
         if (value!=0)   /* 0 => use default */
             BOUNDCHECK(ZSTD_c_chainLog, value);
-        CCtxParams->cParams.chainLog = value;
+        CCtxParams->cParams.chainLog = (U32)value;
         return CCtxParams->cParams.chainLog;
 
     case ZSTD_c_searchLog :
         if (value!=0)   /* 0 => use default */
             BOUNDCHECK(ZSTD_c_searchLog, value);
-        CCtxParams->cParams.searchLog = value;
-        return value;
+        CCtxParams->cParams.searchLog = (U32)value;
+        return (size_t)value;
 
     case ZSTD_c_minMatch :
         if (value!=0)   /* 0 => use default */
@@ -674,6 +691,12 @@ size_t ZSTD_CCtxParams_setParameter(ZSTD_CCtx_params* CCtxParams,
         CCtxParams->targetCBlockSize = value;
         return CCtxParams->targetCBlockSize;
 
+    case ZSTD_c_srcSizeHint :
+        if (value!=0)    /* 0 ==> default */
+            BOUNDCHECK(ZSTD_c_srcSizeHint, value);
+        CCtxParams->srcSizeHint = value;
+        return CCtxParams->srcSizeHint;
+
     default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
     }
 }
@@ -779,6 +802,9 @@ size_t ZSTD_CCtxParams_getParameter(
     case ZSTD_c_targetCBlockSize :
         *value = (int)CCtxParams->targetCBlockSize;
         break;
+    case ZSTD_c_srcSizeHint :
+        *value = (int)CCtxParams->srcSizeHint;
+        break;
     default: RETURN_ERROR(parameter_unsupported, "unknown parameter");
     }
     return 0;
@@ -1029,7 +1055,11 @@ ZSTD_adjustCParams(ZSTD_compressionParameters cPar,
 ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
         const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize)
 {
-    ZSTD_compressionParameters cParams = ZSTD_getCParams(CCtxParams->compressionLevel, srcSizeHint, dictSize);
+    ZSTD_compressionParameters cParams;
+    if (srcSizeHint == ZSTD_CONTENTSIZE_UNKNOWN && CCtxParams->srcSizeHint > 0) {
+      srcSizeHint = CCtxParams->srcSizeHint;
+    }
+    cParams = ZSTD_getCParams(CCtxParams->compressionLevel, srcSizeHint, dictSize);
     if (CCtxParams->ldmParams.enableLdm) cParams.windowLog = ZSTD_LDM_DEFAULT_WINDOW_LOG;
     if (CCtxParams->cParams.windowLog) cParams.windowLog = CCtxParams->cParams.windowLog;
     if (CCtxParams->cParams.hashLog) cParams.hashLog = CCtxParams->cParams.hashLog;
@@ -1049,10 +1079,19 @@ ZSTD_sizeof_matchState(const ZSTD_compressionParameters* const cParams,
     size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);
     size_t const hSize = ((size_t)1) << cParams->hashLog;
     U32    const hashLog3 = (forCCtx && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
-    size_t const h3Size = ((size_t)1) << hashLog3;
-    size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
-    size_t const optPotentialSpace = ((MaxML+1) + (MaxLL+1) + (MaxOff+1) + (1<<Litbits)) * sizeof(U32)
-                          + (ZSTD_OPT_NUM+1) * (sizeof(ZSTD_match_t)+sizeof(ZSTD_optimal_t));
+    size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
+    /* We don't use ZSTD_cwksp_alloc_size() here because the tables aren't
+     * surrounded by redzones in ASAN. */
+    size_t const tableSpace = chainSize * sizeof(U32)
+                            + hSize * sizeof(U32)
+                            + h3Size * sizeof(U32);
+    size_t const optPotentialSpace =
+        ZSTD_cwksp_alloc_size((MaxML+1) * sizeof(U32))
+      + ZSTD_cwksp_alloc_size((MaxLL+1) * sizeof(U32))
+      + ZSTD_cwksp_alloc_size((MaxOff+1) * sizeof(U32))
+      + ZSTD_cwksp_alloc_size((1<<Litbits) * sizeof(U32))
+      + ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t))
+      + ZSTD_cwksp_alloc_size((ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
     size_t const optSpace = (forCCtx && (cParams->strategy >= ZSTD_btopt))
                                 ? optPotentialSpace
                                 : 0;
@@ -1069,20 +1108,23 @@ size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params)
         size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
         U32    const divider = (cParams.minMatch==3) ? 3 : 4;
         size_t const maxNbSeq = blockSize / divider;
-        size_t const tokenSpace = WILDCOPY_OVERLENGTH + blockSize + 11*maxNbSeq;
-        size_t const entropySpace = HUF_WORKSPACE_SIZE;
-        size_t const blockStateSpace = 2 * sizeof(ZSTD_compressedBlockState_t);
+        size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+                                + ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))
+                                + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
+        size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
+        size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
         size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 1);
 
         size_t const ldmSpace = ZSTD_ldm_getTableSize(params->ldmParams);
-        size_t const ldmSeqSpace = ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq);
+        size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(ZSTD_ldm_getMaxNbSeq(params->ldmParams, blockSize) * sizeof(rawSeq));
 
         size_t const neededSpace = entropySpace + blockStateSpace + tokenSpace +
                                    matchStateSize + ldmSpace + ldmSeqSpace;
+        size_t const cctxSpace = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx));
 
-        DEBUGLOG(5, "sizeof(ZSTD_CCtx) : %u", (U32)sizeof(ZSTD_CCtx));
-        DEBUGLOG(5, "estimate workSpace : %u", (U32)neededSpace);
-        return sizeof(ZSTD_CCtx) + neededSpace;
+        DEBUGLOG(5, "sizeof(ZSTD_CCtx) : %u", (U32)cctxSpace);
+        DEBUGLOG(5, "estimate workspace : %u", (U32)neededSpace);
+        return cctxSpace + neededSpace;
     }
 }
 
@@ -1118,7 +1160,8 @@ size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_params* params)
         size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, (size_t)1 << cParams.windowLog);
         size_t const inBuffSize = ((size_t)1 << cParams.windowLog) + blockSize;
         size_t const outBuffSize = ZSTD_compressBound(blockSize) + 1;
-        size_t const streamingSize = inBuffSize + outBuffSize;
+        size_t const streamingSize = ZSTD_cwksp_alloc_size(inBuffSize)
+                                   + ZSTD_cwksp_alloc_size(outBuffSize);
 
         return CCtxSize + streamingSize;
     }
@@ -1186,17 +1229,6 @@ size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx)
     return 0;   /* over-simplification; could also check if context is currently running in streaming mode, and in which case, report how many bytes are left to be flushed within output buffer */
 }
 
-
-
-static U32 ZSTD_equivalentCParams(ZSTD_compressionParameters cParams1,
-                                  ZSTD_compressionParameters cParams2)
-{
-    return (cParams1.hashLog  == cParams2.hashLog)
-         & (cParams1.chainLog == cParams2.chainLog)
-         & (cParams1.strategy == cParams2.strategy)   /* opt parser space */
-         & ((cParams1.minMatch==3) == (cParams2.minMatch==3));  /* hashlog3 space */
-}
-
 static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,
                                     ZSTD_compressionParameters cParams2)
 {
@@ -1211,71 +1243,6 @@ static void ZSTD_assertEqualCParams(ZSTD_compressionParameters cParams1,
     assert(cParams1.strategy     == cParams2.strategy);
 }
 
-/** The parameters are equivalent if ldm is not enabled in both sets or
- *  all the parameters are equivalent. */
-static U32 ZSTD_equivalentLdmParams(ldmParams_t ldmParams1,
-                                    ldmParams_t ldmParams2)
-{
-    return (!ldmParams1.enableLdm && !ldmParams2.enableLdm) ||
-           (ldmParams1.enableLdm == ldmParams2.enableLdm &&
-            ldmParams1.hashLog == ldmParams2.hashLog &&
-            ldmParams1.bucketSizeLog == ldmParams2.bucketSizeLog &&
-            ldmParams1.minMatchLength == ldmParams2.minMatchLength &&
-            ldmParams1.hashRateLog == ldmParams2.hashRateLog);
-}
-
-typedef enum { ZSTDb_not_buffered, ZSTDb_buffered } ZSTD_buffered_policy_e;
-
-/* ZSTD_sufficientBuff() :
- * check internal buffers exist for streaming if buffPol == ZSTDb_buffered .
- * Note : they are assumed to be correctly sized if ZSTD_equivalentCParams()==1 */
-static U32 ZSTD_sufficientBuff(size_t bufferSize1, size_t maxNbSeq1,
-                            size_t maxNbLit1,
-                            ZSTD_buffered_policy_e buffPol2,
-                            ZSTD_compressionParameters cParams2,
-                            U64 pledgedSrcSize)
-{
-    size_t const windowSize2 = MAX(1, (size_t)MIN(((U64)1 << cParams2.windowLog), pledgedSrcSize));
-    size_t const blockSize2 = MIN(ZSTD_BLOCKSIZE_MAX, windowSize2);
-    size_t const maxNbSeq2 = blockSize2 / ((cParams2.minMatch == 3) ? 3 : 4);
-    size_t const maxNbLit2 = blockSize2;
-    size_t const neededBufferSize2 = (buffPol2==ZSTDb_buffered) ? windowSize2 + blockSize2 : 0;
-    DEBUGLOG(4, "ZSTD_sufficientBuff: is neededBufferSize2=%u <= bufferSize1=%u",
-                (U32)neededBufferSize2, (U32)bufferSize1);
-    DEBUGLOG(4, "ZSTD_sufficientBuff: is maxNbSeq2=%u <= maxNbSeq1=%u",
-                (U32)maxNbSeq2, (U32)maxNbSeq1);
-    DEBUGLOG(4, "ZSTD_sufficientBuff: is maxNbLit2=%u <= maxNbLit1=%u",
-                (U32)maxNbLit2, (U32)maxNbLit1);
-    return (maxNbLit2 <= maxNbLit1)
-         & (maxNbSeq2 <= maxNbSeq1)
-         & (neededBufferSize2 <= bufferSize1);
-}
-
-/** Equivalence for resetCCtx purposes */
-static U32 ZSTD_equivalentParams(ZSTD_CCtx_params params1,
-                                 ZSTD_CCtx_params params2,
-                                 size_t buffSize1,
-                                 size_t maxNbSeq1, size_t maxNbLit1,
-                                 ZSTD_buffered_policy_e buffPol2,
-                                 U64 pledgedSrcSize)
-{
-    DEBUGLOG(4, "ZSTD_equivalentParams: pledgedSrcSize=%u", (U32)pledgedSrcSize);
-    if (!ZSTD_equivalentCParams(params1.cParams, params2.cParams)) {
-      DEBUGLOG(4, "ZSTD_equivalentCParams() == 0");
-      return 0;
-    }
-    if (!ZSTD_equivalentLdmParams(params1.ldmParams, params2.ldmParams)) {
-      DEBUGLOG(4, "ZSTD_equivalentLdmParams() == 0");
-      return 0;
-    }
-    if (!ZSTD_sufficientBuff(buffSize1, maxNbSeq1, maxNbLit1, buffPol2,
-                             params2.cParams, pledgedSrcSize)) {
-      DEBUGLOG(4, "ZSTD_sufficientBuff() == 0");
-      return 0;
-    }
-    return 1;
-}
-
 static void ZSTD_reset_compressedBlockState(ZSTD_compressedBlockState_t* bs)
 {
     int i;
@@ -1301,87 +1268,104 @@ static void ZSTD_invalidateMatchState(ZSTD_matchState_t* ms)
     ms->dictMatchState = NULL;
 }
 
-/*! ZSTD_continueCCtx() :
- *  reuse CCtx without reset (note : requires no dictionary) */
-static size_t ZSTD_continueCCtx(ZSTD_CCtx* cctx, ZSTD_CCtx_params params, U64 pledgedSrcSize)
-{
-    size_t const windowSize = MAX(1, (size_t)MIN(((U64)1 << params.cParams.windowLog), pledgedSrcSize));
-    size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
-    DEBUGLOG(4, "ZSTD_continueCCtx: re-use context in place");
+/**
+ * Indicates whether this compression proceeds directly from user-provided
+ * source buffer to user-provided destination buffer (ZSTDb_not_buffered), or
+ * whether the context needs to buffer the input/output (ZSTDb_buffered).
+ */
+typedef enum {
+    ZSTDb_not_buffered,
+    ZSTDb_buffered
+} ZSTD_buffered_policy_e;
 
-    cctx->blockSize = blockSize;   /* previous block size could be different even for same windowLog, due to pledgedSrcSize */
-    cctx->appliedParams = params;
-    cctx->blockState.matchState.cParams = params.cParams;
-    cctx->pledgedSrcSizePlusOne = pledgedSrcSize+1;
-    cctx->consumedSrcSize = 0;
-    cctx->producedCSize = 0;
-    if (pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN)
-        cctx->appliedParams.fParams.contentSizeFlag = 0;
-    DEBUGLOG(4, "pledged content size : %u ; flag : %u",
-        (U32)pledgedSrcSize, cctx->appliedParams.fParams.contentSizeFlag);
-    cctx->stage = ZSTDcs_init;
-    cctx->dictID = 0;
-    if (params.ldmParams.enableLdm)
-        ZSTD_window_clear(&cctx->ldmState.window);
-    ZSTD_referenceExternalSequences(cctx, NULL, 0);
-    ZSTD_invalidateMatchState(&cctx->blockState.matchState);
-    ZSTD_reset_compressedBlockState(cctx->blockState.prevCBlock);
-    XXH64_reset(&cctx->xxhState, 0);
-    return 0;
-}
+/**
+ * Controls, for this matchState reset, whether the tables need to be cleared /
+ * prepared for the coming compression (ZSTDcrp_makeClean), or whether the
+ * tables can be left unclean (ZSTDcrp_leaveDirty), because we know that a
+ * subsequent operation will overwrite the table space anyways (e.g., copying
+ * the matchState contents in from a CDict).
+ */
+typedef enum {
+    ZSTDcrp_makeClean,
+    ZSTDcrp_leaveDirty
+} ZSTD_compResetPolicy_e;
 
-typedef enum { ZSTDcrp_continue, ZSTDcrp_noMemset } ZSTD_compResetPolicy_e;
+/**
+ * Controls, for this matchState reset, whether indexing can continue where it
+ * left off (ZSTDirp_continue), or whether it needs to be restarted from zero
+ * (ZSTDirp_reset).
+ */
+typedef enum {
+    ZSTDirp_continue,
+    ZSTDirp_reset
+} ZSTD_indexResetPolicy_e;
 
-typedef enum { ZSTD_resetTarget_CDict, ZSTD_resetTarget_CCtx } ZSTD_resetTarget_e;
+typedef enum {
+    ZSTD_resetTarget_CDict,
+    ZSTD_resetTarget_CCtx
+} ZSTD_resetTarget_e;
 
-static void*
+static size_t
 ZSTD_reset_matchState(ZSTD_matchState_t* ms,
-                      void* ptr,
+                      ZSTD_cwksp* ws,
                 const ZSTD_compressionParameters* cParams,
-                      ZSTD_compResetPolicy_e const crp, ZSTD_resetTarget_e const forWho)
+                const ZSTD_compResetPolicy_e crp,
+                const ZSTD_indexResetPolicy_e forceResetIndex,
+                const ZSTD_resetTarget_e forWho)
 {
     size_t const chainSize = (cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cParams->chainLog);
     size_t const hSize = ((size_t)1) << cParams->hashLog;
     U32    const hashLog3 = ((forWho == ZSTD_resetTarget_CCtx) && cParams->minMatch==3) ? MIN(ZSTD_HASHLOG3_MAX, cParams->windowLog) : 0;
-    size_t const h3Size = ((size_t)1) << hashLog3;
-    size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
-
-    assert(((size_t)ptr & 3) == 0);
+    size_t const h3Size = hashLog3 ? ((size_t)1) << hashLog3 : 0;
+
+    DEBUGLOG(4, "reset indices : %u", forceResetIndex == ZSTDirp_reset);
+    if (forceResetIndex == ZSTDirp_reset) {
+        memset(&ms->window, 0, sizeof(ms->window));
+        ms->window.dictLimit = 1;    /* start from 1, so that 1st position is valid */
+        ms->window.lowLimit = 1;     /* it ensures first and later CCtx usages compress the same */
+        ms->window.nextSrc = ms->window.base + 1;   /* see issue #1241 */
+        ZSTD_cwksp_mark_tables_dirty(ws);
+    }
 
     ms->hashLog3 = hashLog3;
-    memset(&ms->window, 0, sizeof(ms->window));
-    ms->window.dictLimit = 1;    /* start from 1, so that 1st position is valid */
-    ms->window.lowLimit = 1;     /* it ensures first and later CCtx usages compress the same */
-    ms->window.nextSrc = ms->window.base + 1;   /* see issue #1241 */
+
     ZSTD_invalidateMatchState(ms);
 
+    assert(!ZSTD_cwksp_reserve_failed(ws)); /* check that allocation hasn't already failed */
+
+    ZSTD_cwksp_clear_tables(ws);
+
+    DEBUGLOG(5, "reserving table space");
+    /* table Space */
+    ms->hashTable = (U32*)ZSTD_cwksp_reserve_table(ws, hSize * sizeof(U32));
+    ms->chainTable = (U32*)ZSTD_cwksp_reserve_table(ws, chainSize * sizeof(U32));
+    ms->hashTable3 = (U32*)ZSTD_cwksp_reserve_table(ws, h3Size * sizeof(U32));
+    RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
+                    "failed a workspace allocation in ZSTD_reset_matchState");
+
+    DEBUGLOG(4, "reset table : %u", crp!=ZSTDcrp_leaveDirty);
+    if (crp!=ZSTDcrp_leaveDirty) {
+        /* reset tables only */
+        ZSTD_cwksp_clean_tables(ws);
+    }
+
     /* opt parser space */
     if ((forWho == ZSTD_resetTarget_CCtx) && (cParams->strategy >= ZSTD_btopt)) {
         DEBUGLOG(4, "reserving optimal parser space");
-        ms->opt.litFreq = (unsigned*)ptr;
-        ms->opt.litLengthFreq = ms->opt.litFreq + (1<<Litbits);
-        ms->opt.matchLengthFreq = ms->opt.litLengthFreq + (MaxLL+1);
-        ms->opt.offCodeFreq = ms->opt.matchLengthFreq + (MaxML+1);
-        ptr = ms->opt.offCodeFreq + (MaxOff+1);
-        ms->opt.matchTable = (ZSTD_match_t*)ptr;
-        ptr = ms->opt.matchTable + ZSTD_OPT_NUM+1;
-        ms->opt.priceTable = (ZSTD_optimal_t*)ptr;
-        ptr = ms->opt.priceTable + ZSTD_OPT_NUM+1;
+        ms->opt.litFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (1<<Litbits) * sizeof(unsigned));
+        ms->opt.litLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxLL+1) * sizeof(unsigned));
+        ms->opt.matchLengthFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxML+1) * sizeof(unsigned));
+        ms->opt.offCodeFreq = (unsigned*)ZSTD_cwksp_reserve_aligned(ws, (MaxOff+1) * sizeof(unsigned));
+        ms->opt.matchTable = (ZSTD_match_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_match_t));
+        ms->opt.priceTable = (ZSTD_optimal_t*)ZSTD_cwksp_reserve_aligned(ws, (ZSTD_OPT_NUM+1) * sizeof(ZSTD_optimal_t));
     }
 
-    /* table Space */
-    DEBUGLOG(4, "reset table : %u", crp!=ZSTDcrp_noMemset);
-    assert(((size_t)ptr & 3) == 0);  /* ensure ptr is properly aligned */
-    if (crp!=ZSTDcrp_noMemset) memset(ptr, 0, tableSpace);   /* reset tables only */
-    ms->hashTable = (U32*)(ptr);
-    ms->chainTable = ms->hashTable + hSize;
-    ms->hashTable3 = ms->chainTable + chainSize;
-    ptr = ms->hashTable3 + h3Size;
-
     ms->cParams = *cParams;
 
-    assert(((size_t)ptr & 3) == 0);
-    return ptr;
+    RETURN_ERROR_IF(ZSTD_cwksp_reserve_failed(ws), memory_allocation,
+                    "failed a workspace allocation in ZSTD_reset_matchState");
+
+    return 0;
 }
 
 /* ZSTD_indexTooCloseToMax() :
@@ -1397,13 +1381,6 @@ static int ZSTD_indexTooCloseToMax(ZSTD_window_t w)
     return (size_t)(w.nextSrc - w.base) > (ZSTD_CURRENT_MAX - ZSTD_INDEXOVERFLOW_MARGIN);
 }
 
-#define ZSTD_WORKSPACETOOLARGE_FACTOR 3 /* define "workspace is too large" as this number of times larger than needed */
-#define ZSTD_WORKSPACETOOLARGE_MAXDURATION 128  /* when workspace is continuously too large
-                                         * during at least this number of times,
-                                         * context's memory usage is considered wasteful,
-                                         * because it's sized to handle a worst case scenario which rarely happens.
-                                         * In which case, resize it down to free some memory */
-
 /*! ZSTD_resetCCtx_internal() :
     note : `params` are assumed fully validated at this stage */
 static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
@@ -1412,30 +1389,12 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
                                       ZSTD_compResetPolicy_e const crp,
                                       ZSTD_buffered_policy_e const zbuff)
 {
+    ZSTD_cwksp* const ws = &zc->workspace;
     DEBUGLOG(4, "ZSTD_resetCCtx_internal: pledgedSrcSize=%u, wlog=%u",
                 (U32)pledgedSrcSize, params.cParams.windowLog);
     assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
 
-    if (crp == ZSTDcrp_continue) {
-        if (ZSTD_equivalentParams(zc->appliedParams, params,
-                                  zc->inBuffSize,
-                                  zc->seqStore.maxNbSeq, zc->seqStore.maxNbLit,
-                                  zbuff, pledgedSrcSize) ) {
-            DEBUGLOG(4, "ZSTD_equivalentParams()==1 -> consider continue mode");
-            zc->workSpaceOversizedDuration += (zc->workSpaceOversizedDuration > 0);   /* if it was too large, it still is */
-            if (zc->workSpaceOversizedDuration <= ZSTD_WORKSPACETOOLARGE_MAXDURATION) {
-                DEBUGLOG(4, "continue mode confirmed (wLog1=%u, blockSize1=%zu)",
-                            zc->appliedParams.cParams.windowLog, zc->blockSize);
-                if (ZSTD_indexTooCloseToMax(zc->blockState.matchState.window)) {
-                    /* prefer a reset, faster than a rescale */
-                    ZSTD_reset_matchState(&zc->blockState.matchState,
-                                           zc->entropyWorkspace + HUF_WORKSPACE_SIZE_U32,
-                                          &params.cParams,
-                                           crp, ZSTD_resetTarget_CCtx);
-                }
-                return ZSTD_continueCCtx(zc, params, pledgedSrcSize);
-    }   }   }
-    DEBUGLOG(4, "ZSTD_equivalentParams()==0 -> reset CCtx");
+    zc->isFirstBlock = 1;
 
     if (params.ldmParams.enableLdm) {
         /* Adjust long distance matching parameters */
@@ -1449,58 +1408,74 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
         size_t const blockSize = MIN(ZSTD_BLOCKSIZE_MAX, windowSize);
         U32    const divider = (params.cParams.minMatch==3) ? 3 : 4;
         size_t const maxNbSeq = blockSize / divider;
-        size_t const tokenSpace = WILDCOPY_OVERLENGTH + blockSize + 11*maxNbSeq;
+        size_t const tokenSpace = ZSTD_cwksp_alloc_size(WILDCOPY_OVERLENGTH + blockSize)
+                                + ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(seqDef))
+                                + 3 * ZSTD_cwksp_alloc_size(maxNbSeq * sizeof(BYTE));
         size_t const buffOutSize = (zbuff==ZSTDb_buffered) ? ZSTD_compressBound(blockSize)+1 : 0;
         size_t const buffInSize = (zbuff==ZSTDb_buffered) ? windowSize + blockSize : 0;
         size_t const matchStateSize = ZSTD_sizeof_matchState(&params.cParams, /* forCCtx */ 1);
         size_t const maxNbLdmSeq = ZSTD_ldm_getMaxNbSeq(params.ldmParams, blockSize);
-        void* ptr;   /* used to partition workSpace */
 
-        /* Check if workSpace is large enough, alloc a new one if needed */
-        {   size_t const entropySpace = HUF_WORKSPACE_SIZE;
-            size_t const blockStateSpace = 2 * sizeof(ZSTD_compressedBlockState_t);
-            size_t const bufferSpace = buffInSize + buffOutSize;
+        ZSTD_indexResetPolicy_e needsIndexReset = ZSTDirp_continue;
+
+        if (ZSTD_indexTooCloseToMax(zc->blockState.matchState.window)) {
+            needsIndexReset = ZSTDirp_reset;
+        }
+
+        ZSTD_cwksp_bump_oversized_duration(ws, 0);
+
+        /* Check if workspace is large enough, alloc a new one if needed */
+        {   size_t const cctxSpace = zc->staticSize ? ZSTD_cwksp_alloc_size(sizeof(ZSTD_CCtx)) : 0;
+            size_t const entropySpace = ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE);
+            size_t const blockStateSpace = 2 * ZSTD_cwksp_alloc_size(sizeof(ZSTD_compressedBlockState_t));
+            size_t const bufferSpace = ZSTD_cwksp_alloc_size(buffInSize) + ZSTD_cwksp_alloc_size(buffOutSize);
             size_t const ldmSpace = ZSTD_ldm_getTableSize(params.ldmParams);
-            size_t const ldmSeqSpace = maxNbLdmSeq * sizeof(rawSeq);
+            size_t const ldmSeqSpace = ZSTD_cwksp_alloc_size(maxNbLdmSeq * sizeof(rawSeq));
 
-            size_t const neededSpace = entropySpace + blockStateSpace + ldmSpace +
-                                       ldmSeqSpace + matchStateSize + tokenSpace +
-                                       bufferSpace;
+            size_t const neededSpace =
+                cctxSpace +
+                entropySpace +
+                blockStateSpace +
+                ldmSpace +
+                ldmSeqSpace +
+                matchStateSize +
+                tokenSpace +
+                bufferSpace;
 
-            int const workSpaceTooSmall = zc->workSpaceSize < neededSpace;
-            int const workSpaceTooLarge = zc->workSpaceSize > ZSTD_WORKSPACETOOLARGE_FACTOR * neededSpace;
-            int const workSpaceWasteful = workSpaceTooLarge && (zc->workSpaceOversizedDuration > ZSTD_WORKSPACETOOLARGE_MAXDURATION);
-            zc->workSpaceOversizedDuration = workSpaceTooLarge ? zc->workSpaceOversizedDuration+1 : 0;
+            int const workspaceTooSmall = ZSTD_cwksp_sizeof(ws) < neededSpace;
+            int const workspaceWasteful = ZSTD_cwksp_check_wasteful(ws, neededSpace);
 
             DEBUGLOG(4, "Need %zuKB workspace, including %zuKB for match state, and %zuKB for buffers",
                         neededSpace>>10, matchStateSize>>10, bufferSpace>>10);
             DEBUGLOG(4, "windowSize: %zu - blockSize: %zu", windowSize, blockSize);
 
-            if (workSpaceTooSmall || workSpaceWasteful) {
-                DEBUGLOG(4, "Resize workSpaceSize from %zuKB to %zuKB",
-                            zc->workSpaceSize >> 10,
+            if (workspaceTooSmall || workspaceWasteful) {
+                DEBUGLOG(4, "Resize workspaceSize from %zuKB to %zuKB",
+                            ZSTD_cwksp_sizeof(ws) >> 10,
                             neededSpace >> 10);
 
                 RETURN_ERROR_IF(zc->staticSize, memory_allocation, "static cctx : no resize");
 
-                zc->workSpaceSize = 0;
-                ZSTD_free(zc->workSpace, zc->customMem);
-                zc->workSpace = ZSTD_malloc(neededSpace, zc->customMem);
-                RETURN_ERROR_IF(zc->workSpace == NULL, memory_allocation);
-                zc->workSpaceSize = neededSpace;
-                zc->workSpaceOversizedDuration = 0;
+                needsIndexReset = ZSTDirp_reset;
 
+                ZSTD_cwksp_free(ws, zc->customMem);
+                FORWARD_IF_ERROR(ZSTD_cwksp_create(ws, neededSpace, zc->customMem));
+
+                DEBUGLOG(5, "reserving object space");
                 /* Statically sized space.
                  * entropyWorkspace never moves,
                  * though prev/next block swap places */
-                assert(((size_t)zc->workSpace & 3) == 0);   /* ensure correct alignment */
-                assert(zc->workSpaceSize >= 2 * sizeof(ZSTD_compressedBlockState_t));
-                zc->blockState.prevCBlock = (ZSTD_compressedBlockState_t*)zc->workSpace;
-                zc->blockState.nextCBlock = zc->blockState.prevCBlock + 1;
-                ptr = zc->blockState.nextCBlock + 1;
-                zc->entropyWorkspace = (U32*)ptr;
+                assert(ZSTD_cwksp_check_available(ws, 2 * sizeof(ZSTD_compressedBlockState_t)));
+                zc->blockState.prevCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));
+                RETURN_ERROR_IF(zc->blockState.prevCBlock == NULL, memory_allocation, "couldn't allocate prevCBlock");
+                zc->blockState.nextCBlock = (ZSTD_compressedBlockState_t*) ZSTD_cwksp_reserve_object(ws, sizeof(ZSTD_compressedBlockState_t));
+                RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate nextCBlock");
+                zc->entropyWorkspace = (U32*) ZSTD_cwksp_reserve_object(ws, HUF_WORKSPACE_SIZE);
+                RETURN_ERROR_IF(zc->blockState.nextCBlock == NULL, memory_allocation, "couldn't allocate entropyWorkspace");
         }   }
 
+        ZSTD_cwksp_clear(ws);
+
         /* init params */
         zc->appliedParams = params;
         zc->blockState.matchState.cParams = params.cParams;
@@ -1519,58 +1494,58 @@ static size_t ZSTD_resetCCtx_internal(ZSTD_CCtx* zc,
 
         ZSTD_reset_compressedBlockState(zc->blockState.prevCBlock);
 
-        ptr = ZSTD_reset_matchState(&zc->blockState.matchState,
-                                     zc->entropyWorkspace + HUF_WORKSPACE_SIZE_U32,
-                                    &params.cParams,
-                                     crp, ZSTD_resetTarget_CCtx);
-
-        /* ldm hash table */
-        /* initialize bucketOffsets table later for pointer alignment */
-        if (params.ldmParams.enableLdm) {
-            size_t const ldmHSize = ((size_t)1) << params.ldmParams.hashLog;
-            memset(ptr, 0, ldmHSize * sizeof(ldmEntry_t));
-            assert(((size_t)ptr & 3) == 0); /* ensure ptr is properly aligned */
-            zc->ldmState.hashTable = (ldmEntry_t*)ptr;
-            ptr = zc->ldmState.hashTable + ldmHSize;
-            zc->ldmSequences = (rawSeq*)ptr;
-            ptr = zc->ldmSequences + maxNbLdmSeq;
-            zc->maxNbLdmSequences = maxNbLdmSeq;
-
-            memset(&zc->ldmState.window, 0, sizeof(zc->ldmState.window));
-        }
-        assert(((size_t)ptr & 3) == 0); /* ensure ptr is properly aligned */
-
-        /* sequences storage */
-        zc->seqStore.maxNbSeq = maxNbSeq;
-        zc->seqStore.sequencesStart = (seqDef*)ptr;
-        ptr = zc->seqStore.sequencesStart + maxNbSeq;
-        zc->seqStore.llCode = (BYTE*) ptr;
-        zc->seqStore.mlCode = zc->seqStore.llCode + maxNbSeq;
-        zc->seqStore.ofCode = zc->seqStore.mlCode + maxNbSeq;
-        zc->seqStore.litStart = zc->seqStore.ofCode + maxNbSeq;
         /* ZSTD_wildcopy() is used to copy into the literals buffer,
          * so we have to oversize the buffer by WILDCOPY_OVERLENGTH bytes.
          */
+        zc->seqStore.litStart = ZSTD_cwksp_reserve_buffer(ws, blockSize + WILDCOPY_OVERLENGTH);
         zc->seqStore.maxNbLit = blockSize;
-        ptr = zc->seqStore.litStart + blockSize + WILDCOPY_OVERLENGTH;
+
+        /* buffers */
+        zc->inBuffSize = buffInSize;
+        zc->inBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffInSize);
+        zc->outBuffSize = buffOutSize;
+        zc->outBuff = (char*)ZSTD_cwksp_reserve_buffer(ws, buffOutSize);
 
         /* ldm bucketOffsets table */
         if (params.ldmParams.enableLdm) {
+            /* TODO: avoid memset? */
             size_t const ldmBucketSize =
                   ((size_t)1) << (params.ldmParams.hashLog -
                                   params.ldmParams.bucketSizeLog);
-            memset(ptr, 0, ldmBucketSize);
-            zc->ldmState.bucketOffsets = (BYTE*)ptr;
-            ptr = zc->ldmState.bucketOffsets + ldmBucketSize;
-            ZSTD_window_clear(&zc->ldmState.window);
+            zc->ldmState.bucketOffsets = ZSTD_cwksp_reserve_buffer(ws, ldmBucketSize);
+            memset(zc->ldmState.bucketOffsets, 0, ldmBucketSize);
         }
+
+        /* sequences storage */
         ZSTD_referenceExternalSequences(zc, NULL, 0);
+        zc->seqStore.maxNbSeq = maxNbSeq;
+        zc->seqStore.llCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
+        zc->seqStore.mlCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
+        zc->seqStore.ofCode = ZSTD_cwksp_reserve_buffer(ws, maxNbSeq * sizeof(BYTE));
+        zc->seqStore.sequencesStart = (seqDef*)ZSTD_cwksp_reserve_aligned(ws, maxNbSeq * sizeof(seqDef));
+
+        FORWARD_IF_ERROR(ZSTD_reset_matchState(
+            &zc->blockState.matchState,
+            ws,
+            &params.cParams,
+            crp,
+            needsIndexReset,
+            ZSTD_resetTarget_CCtx));
 
-        /* buffers */
-        zc->inBuffSize = buffInSize;
-        zc->inBuff = (char*)ptr;
-        zc->outBuffSize = buffOutSize;
-        zc->outBuff = zc->inBuff + buffInSize;
+        /* ldm hash table */
+        if (params.ldmParams.enableLdm) {
+            /* TODO: avoid memset? */
+            size_t const ldmHSize = ((size_t)1) << params.ldmParams.hashLog;
+            zc->ldmState.hashTable = (ldmEntry_t*)ZSTD_cwksp_reserve_aligned(ws, ldmHSize * sizeof(ldmEntry_t));
+            memset(zc->ldmState.hashTable, 0, ldmHSize * sizeof(ldmEntry_t));
+            zc->ldmSequences = (rawSeq*)ZSTD_cwksp_reserve_aligned(ws, maxNbLdmSeq * sizeof(rawSeq));
+            zc->maxNbLdmSequences = maxNbLdmSeq;
+
+            memset(&zc->ldmState.window, 0, sizeof(zc->ldmState.window));
+            ZSTD_window_clear(&zc->ldmState.window);
+        }
+
+        DEBUGLOG(3, "wksp: finished allocating, %zd bytes remain available", ZSTD_cwksp_available_space(ws));
 
         return 0;
     }
@@ -1604,15 +1579,15 @@ static const size_t attachDictSizeCutoffs[ZSTD_STRATEGY_MAX+1] = {
 };
 
 static int ZSTD_shouldAttachDict(const ZSTD_CDict* cdict,
-                                 ZSTD_CCtx_params params,
+                                 const ZSTD_CCtx_params* params,
                                  U64 pledgedSrcSize)
 {
     size_t cutoff = attachDictSizeCutoffs[cdict->matchState.cParams.strategy];
     return ( pledgedSrcSize <= cutoff
           || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
-          || params.attachDictPref == ZSTD_dictForceAttach )
-        && params.attachDictPref != ZSTD_dictForceCopy
-        && !params.forceWindow; /* dictMatchState isn't correctly
+          || params->attachDictPref == ZSTD_dictForceAttach )
+        && params->attachDictPref != ZSTD_dictForceCopy
+        && !params->forceWindow; /* dictMatchState isn't correctly
                                  * handled in _enforceMaxDist */
 }
 
@@ -1630,8 +1605,8 @@ ZSTD_resetCCtx_byAttachingCDict(ZSTD_CCtx* cctx,
          * has its own tables. */
         params.cParams = ZSTD_adjustCParams_internal(*cdict_cParams, pledgedSrcSize, 0);
         params.cParams.windowLog = windowLog;
-        ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
-                                ZSTDcrp_continue, zbuff);
+        FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
+                                                 ZSTDcrp_makeClean, zbuff));
         assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);
     }
 
@@ -1679,30 +1654,36 @@ static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx,
         /* Copy only compression parameters related to tables. */
         params.cParams = *cdict_cParams;
         params.cParams.windowLog = windowLog;
-        ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
-                                ZSTDcrp_noMemset, zbuff);
+        FORWARD_IF_ERROR(ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
+                                                 ZSTDcrp_leaveDirty, zbuff));
         assert(cctx->appliedParams.cParams.strategy == cdict_cParams->strategy);
         assert(cctx->appliedParams.cParams.hashLog == cdict_cParams->hashLog);
         assert(cctx->appliedParams.cParams.chainLog == cdict_cParams->chainLog);
     }
 
+    ZSTD_cwksp_mark_tables_dirty(&cctx->workspace);
+
     /* copy tables */
     {   size_t const chainSize = (cdict_cParams->strategy == ZSTD_fast) ? 0 : ((size_t)1 << cdict_cParams->chainLog);
         size_t const hSize =  (size_t)1 << cdict_cParams->hashLog;
-        size_t const tableSpace = (chainSize + hSize) * sizeof(U32);
-        assert((U32*)cctx->blockState.matchState.chainTable == (U32*)cctx->blockState.matchState.hashTable + hSize);  /* chainTable must follow hashTable */
-        assert((U32*)cctx->blockState.matchState.hashTable3 == (U32*)cctx->blockState.matchState.chainTable + chainSize);
-        assert((U32*)cdict->matchState.chainTable == (U32*)cdict->matchState.hashTable + hSize);  /* chainTable must follow hashTable */
-        assert((U32*)cdict->matchState.hashTable3 == (U32*)cdict->matchState.chainTable + chainSize);
-        memcpy(cctx->blockState.matchState.hashTable, cdict->matchState.hashTable, tableSpace);   /* presumes all tables follow each other */
+
+        memcpy(cctx->blockState.matchState.hashTable,
+               cdict->matchState.hashTable,
+               hSize * sizeof(U32));
+        memcpy(cctx->blockState.matchState.chainTable,
+               cdict->matchState.chainTable,
+               chainSize * sizeof(U32));
     }
 
     /* Zero the hashTable3, since the cdict never fills it */
-    {   size_t const h3Size = (size_t)1 << cctx->blockState.matchState.hashLog3;
+    {   int const h3log = cctx->blockState.matchState.hashLog3;
+        size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;
         assert(cdict->matchState.hashLog3 == 0);
         memset(cctx->blockState.matchState.hashTable3, 0, h3Size * sizeof(U32));
     }
 
+    ZSTD_cwksp_mark_tables_clean(&cctx->workspace);
+
     /* copy dictionary offsets */
     {   ZSTD_matchState_t const* srcMatchState = &cdict->matchState;
         ZSTD_matchState_t* dstMatchState = &cctx->blockState.matchState;
@@ -1724,7 +1705,7 @@ static size_t ZSTD_resetCCtx_byCopyingCDict(ZSTD_CCtx* cctx,
  * in-place. We decide here which strategy to use. */
 static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
                             const ZSTD_CDict* cdict,
-                            ZSTD_CCtx_params params,
+                            const ZSTD_CCtx_params* params,
                             U64 pledgedSrcSize,
                             ZSTD_buffered_policy_e zbuff)
 {
@@ -1734,10 +1715,10 @@ static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
 
     if (ZSTD_shouldAttachDict(cdict, params, pledgedSrcSize)) {
         return ZSTD_resetCCtx_byAttachingCDict(
-            cctx, cdict, params, pledgedSrcSize, zbuff);
+            cctx, cdict, *params, pledgedSrcSize, zbuff);
     } else {
         return ZSTD_resetCCtx_byCopyingCDict(
-            cctx, cdict, params, pledgedSrcSize, zbuff);
+            cctx, cdict, *params, pledgedSrcSize, zbuff);
     }
 }
 
@@ -1763,7 +1744,7 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
         params.cParams = srcCCtx->appliedParams.cParams;
         params.fParams = fParams;
         ZSTD_resetCCtx_internal(dstCCtx, params, pledgedSrcSize,
-                                ZSTDcrp_noMemset, zbuff);
+                                ZSTDcrp_leaveDirty, zbuff);
         assert(dstCCtx->appliedParams.cParams.windowLog == srcCCtx->appliedParams.cParams.windowLog);
         assert(dstCCtx->appliedParams.cParams.strategy == srcCCtx->appliedParams.cParams.strategy);
         assert(dstCCtx->appliedParams.cParams.hashLog == srcCCtx->appliedParams.cParams.hashLog);
@@ -1771,16 +1752,27 @@ static size_t ZSTD_copyCCtx_internal(ZSTD_CCtx* dstCCtx,
         assert(dstCCtx->blockState.matchState.hashLog3 == srcCCtx->blockState.matchState.hashLog3);
     }
 
+    ZSTD_cwksp_mark_tables_dirty(&dstCCtx->workspace);
+
     /* copy tables */
     {   size_t const chainSize = (srcCCtx->appliedParams.cParams.strategy == ZSTD_fast) ? 0 : ((size_t)1 << srcCCtx->appliedParams.cParams.chainLog);
         size_t const hSize =  (size_t)1 << srcCCtx->appliedParams.cParams.hashLog;
-        size_t const h3Size = (size_t)1 << srcCCtx->blockState.matchState.hashLog3;
-        size_t const tableSpace = (chainSize + hSize + h3Size) * sizeof(U32);
-        assert((U32*)dstCCtx->blockState.matchState.chainTable == (U32*)dstCCtx->blockState.matchState.hashTable + hSize);  /* chainTable must follow hashTable */
-        assert((U32*)dstCCtx->blockState.matchState.hashTable3 == (U32*)dstCCtx->blockState.matchState.chainTable + chainSize);
-        memcpy(dstCCtx->blockState.matchState.hashTable, srcCCtx->blockState.matchState.hashTable, tableSpace);   /* presumes all tables follow each other */
+        int const h3log = srcCCtx->blockState.matchState.hashLog3;
+        size_t const h3Size = h3log ? ((size_t)1 << h3log) : 0;
+
+        memcpy(dstCCtx->blockState.matchState.hashTable,
+               srcCCtx->blockState.matchState.hashTable,
+               hSize * sizeof(U32));
+        memcpy(dstCCtx->blockState.matchState.chainTable,
+               srcCCtx->blockState.matchState.chainTable,
+               chainSize * sizeof(U32));
+        memcpy(dstCCtx->blockState.matchState.hashTable3,
+               srcCCtx->blockState.matchState.hashTable3,
+               h3Size * sizeof(U32));
     }
 
+    ZSTD_cwksp_mark_tables_clean(&dstCCtx->workspace);
+
     /* copy dictionary offsets */
     {
         const ZSTD_matchState_t* srcMatchState = &srcCCtx->blockState.matchState;
@@ -1831,6 +1823,20 @@ ZSTD_reduceTable_internal (U32* const table, U32 const size, U32 const reducerVa
     int rowNb;
     assert((size & (ZSTD_ROWSIZE-1)) == 0);  /* multiple of ZSTD_ROWSIZE */
     assert(size < (1U<<31));   /* can be casted to int */
+
+#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
+    /* To validate that the table re-use logic is sound, and that we don't
+     * access table space that we haven't cleaned, we re-"poison" the table
+     * space every time we mark it dirty.
+     *
+     * This function however is intended to operate on those dirty tables and
+     * re-clean them. So when this function is used correctly, we can unpoison
+     * the memory it operated on. This introduces a blind spot though, since
+     * if we now try to operate on __actually__ poisoned memory, we will not
+     * detect that. */
+    __msan_unpoison(table, size * sizeof(U32));
+#endif
+
     for (rowNb=0 ; rowNb < nbRows ; rowNb++) {
         int column;
         for (column=0; column<ZSTD_ROWSIZE; column++) {
@@ -1938,7 +1944,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                 ZSTD_entropyCTables_t* nextEntropy,
                           const ZSTD_CCtx_params* cctxParams,
                                 void* dst, size_t dstCapacity,
-                                void* workspace, size_t wkspSize,
+                                void* entropyWorkspace, size_t entropyWkspSize,
                           const int bmi2)
 {
     const int longOffsets = cctxParams->cParams.windowLog > STREAM_ACCUMULATOR_MIN;
@@ -1971,7 +1977,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                     ZSTD_disableLiteralsCompression(cctxParams),
                                     op, dstCapacity,
                                     literals, litSize,
-                                    workspace, wkspSize,
+                                    entropyWorkspace, entropyWkspSize,
                                     bmi2);
         FORWARD_IF_ERROR(cSize);
         assert(cSize <= dstCapacity);
@@ -1981,12 +1987,17 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     /* Sequences Header */
     RETURN_ERROR_IF((oend-op) < 3 /*max nbSeq Size*/ + 1 /*seqHead*/,
                     dstSize_tooSmall);
-    if (nbSeq < 0x7F)
+    if (nbSeq < 128) {
         *op++ = (BYTE)nbSeq;
-    else if (nbSeq < LONGNBSEQ)
-        op[0] = (BYTE)((nbSeq>>8) + 0x80), op[1] = (BYTE)nbSeq, op+=2;
-    else
-        op[0]=0xFF, MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ)), op+=3;
+    } else if (nbSeq < LONGNBSEQ) {
+        op[0] = (BYTE)((nbSeq>>8) + 0x80);
+        op[1] = (BYTE)nbSeq;
+        op+=2;
+    } else {
+        op[0]=0xFF;
+        MEM_writeLE16(op+1, (U16)(nbSeq - LONGNBSEQ));
+        op+=3;
+    }
     assert(op <= oend);
     if (nbSeq==0) {
         /* Copy the old tables over as if we repeated them */
@@ -2002,7 +2013,7 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     ZSTD_seqToCodes(seqStorePtr);
     /* build CTable for Literal Lengths */
     {   unsigned max = MaxLL;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, workspace, wkspSize);   /* can't fail */
+        size_t const mostFrequent = HIST_countFast_wksp(count, &max, llCodeTable, nbSeq, entropyWorkspace, entropyWkspSize);   /* can't fail */
         DEBUGLOG(5, "Building LL table");
         nextEntropy->fse.litlength_repeatMode = prevEntropy->fse.litlength_repeatMode;
         LLtype = ZSTD_selectEncodingType(&nextEntropy->fse.litlength_repeatMode,
@@ -2012,10 +2023,14 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         ZSTD_defaultAllowed, strategy);
         assert(set_basic < set_compressed && set_rle < set_compressed);
         assert(!(LLtype < set_compressed && nextEntropy->fse.litlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
-        {   size_t const countSize = ZSTD_buildCTable(op, (size_t)(oend - op), CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
-                                                    count, max, llCodeTable, nbSeq, LL_defaultNorm, LL_defaultNormLog, MaxLL,
-                                                    prevEntropy->fse.litlengthCTable, sizeof(prevEntropy->fse.litlengthCTable),
-                                                    workspace, wkspSize);
+        {   size_t const countSize = ZSTD_buildCTable(
+                op, (size_t)(oend - op),
+                CTable_LitLength, LLFSELog, (symbolEncodingType_e)LLtype,
+                count, max, llCodeTable, nbSeq,
+                LL_defaultNorm, LL_defaultNormLog, MaxLL,
+                prevEntropy->fse.litlengthCTable,
+                sizeof(prevEntropy->fse.litlengthCTable),
+                entropyWorkspace, entropyWkspSize);
             FORWARD_IF_ERROR(countSize);
             if (LLtype == set_compressed)
                 lastNCount = op;
@@ -2024,7 +2039,8 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     }   }
     /* build CTable for Offsets */
     {   unsigned max = MaxOff;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, ofCodeTable, nbSeq, workspace, wkspSize);  /* can't fail */
+        size_t const mostFrequent = HIST_countFast_wksp(
+            count, &max, ofCodeTable, nbSeq, entropyWorkspace, entropyWkspSize);  /* can't fail */
         /* We can only use the basic table if max <= DefaultMaxOff, otherwise the offsets are too large */
         ZSTD_defaultPolicy_e const defaultPolicy = (max <= DefaultMaxOff) ? ZSTD_defaultAllowed : ZSTD_defaultDisallowed;
         DEBUGLOG(5, "Building OF table");
@@ -2035,10 +2051,14 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         OF_defaultNorm, OF_defaultNormLog,
                                         defaultPolicy, strategy);
         assert(!(Offtype < set_compressed && nextEntropy->fse.offcode_repeatMode != FSE_repeat_none)); /* We don't copy tables */
-        {   size_t const countSize = ZSTD_buildCTable(op, (size_t)(oend - op), CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
-                                                    count, max, ofCodeTable, nbSeq, OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
-                                                    prevEntropy->fse.offcodeCTable, sizeof(prevEntropy->fse.offcodeCTable),
-                                                    workspace, wkspSize);
+        {   size_t const countSize = ZSTD_buildCTable(
+                op, (size_t)(oend - op),
+                CTable_OffsetBits, OffFSELog, (symbolEncodingType_e)Offtype,
+                count, max, ofCodeTable, nbSeq,
+                OF_defaultNorm, OF_defaultNormLog, DefaultMaxOff,
+                prevEntropy->fse.offcodeCTable,
+                sizeof(prevEntropy->fse.offcodeCTable),
+                entropyWorkspace, entropyWkspSize);
             FORWARD_IF_ERROR(countSize);
             if (Offtype == set_compressed)
                 lastNCount = op;
@@ -2047,7 +2067,8 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
     }   }
     /* build CTable for MatchLengths */
     {   unsigned max = MaxML;
-        size_t const mostFrequent = HIST_countFast_wksp(count, &max, mlCodeTable, nbSeq, workspace, wkspSize);   /* can't fail */
+        size_t const mostFrequent = HIST_countFast_wksp(
+            count, &max, mlCodeTable, nbSeq, entropyWorkspace, entropyWkspSize);   /* can't fail */
         DEBUGLOG(5, "Building ML table (remaining space : %i)", (int)(oend-op));
         nextEntropy->fse.matchlength_repeatMode = prevEntropy->fse.matchlength_repeatMode;
         MLtype = ZSTD_selectEncodingType(&nextEntropy->fse.matchlength_repeatMode,
@@ -2056,10 +2077,14 @@ ZSTD_compressSequences_internal(seqStore_t* seqStorePtr,
                                         ML_defaultNorm, ML_defaultNormLog,
                                         ZSTD_defaultAllowed, strategy);
         assert(!(MLtype < set_compressed && nextEntropy->fse.matchlength_repeatMode != FSE_repeat_none)); /* We don't copy tables */
-        {   size_t const countSize = ZSTD_buildCTable(op, (size_t)(oend - op), CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
-                                                    count, max, mlCodeTable, nbSeq, ML_defaultNorm, ML_defaultNormLog, MaxML,
-                                                    prevEntropy->fse.matchlengthCTable, sizeof(prevEntropy->fse.matchlengthCTable),
-                                                    workspace, wkspSize);
+        {   size_t const countSize = ZSTD_buildCTable(
+                op, (size_t)(oend - op),
+                CTable_MatchLength, MLFSELog, (symbolEncodingType_e)MLtype,
+                count, max, mlCodeTable, nbSeq,
+                ML_defaultNorm, ML_defaultNormLog, MaxML,
+                prevEntropy->fse.matchlengthCTable,
+                sizeof(prevEntropy->fse.matchlengthCTable),
+                entropyWorkspace, entropyWkspSize);
             FORWARD_IF_ERROR(countSize);
             if (MLtype == set_compressed)
                 lastNCount = op;
@@ -2107,13 +2132,13 @@ ZSTD_compressSequences(seqStore_t* seqStorePtr,
                        const ZSTD_CCtx_params* cctxParams,
                              void* dst, size_t dstCapacity,
                              size_t srcSize,
-                             void* workspace, size_t wkspSize,
+                             void* entropyWorkspace, size_t entropyWkspSize,
                              int bmi2)
 {
     size_t const cSize = ZSTD_compressSequences_internal(
                             seqStorePtr, prevEntropy, nextEntropy, cctxParams,
                             dst, dstCapacity,
-                            workspace, wkspSize, bmi2);
+                            entropyWorkspace, entropyWkspSize, bmi2);
     if (cSize == 0) return 0;
     /* When srcSize <= dstCapacity, there is enough space to write a raw uncompressed block.
      * Since we ran out of space, block must be not compressible, so fall back to raw uncompressed block.
@@ -2264,11 +2289,99 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize)
     return ZSTDbss_compress;
 }
 
+static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc)
+{
+    const seqStore_t* seqStore = ZSTD_getSeqStore(zc);
+    const seqDef* seqs = seqStore->sequencesStart;
+    size_t seqsSize = seqStore->sequences - seqs;
+
+    ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex];
+    size_t i; size_t position; int repIdx;
+
+    assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences);
+    for (i = 0, position = 0; i < seqsSize; ++i) {
+        outSeqs[i].offset = seqs[i].offset;
+        outSeqs[i].litLength = seqs[i].litLength;
+        outSeqs[i].matchLength = seqs[i].matchLength + MINMATCH;
+
+        if (i == seqStore->longLengthPos) {
+            if (seqStore->longLengthID == 1) {
+                outSeqs[i].litLength += 0x10000;
+            } else if (seqStore->longLengthID == 2) {
+                outSeqs[i].matchLength += 0x10000;
+            }
+        }
+
+        if (outSeqs[i].offset <= ZSTD_REP_NUM) {
+            outSeqs[i].rep = outSeqs[i].offset;
+            repIdx = (unsigned int)i - outSeqs[i].offset;
+
+            if (outSeqs[i].litLength == 0) {
+                if (outSeqs[i].offset < 3) {
+                    --repIdx;
+                } else {
+                    repIdx = (unsigned int)i - 1;
+                }
+                ++outSeqs[i].rep;
+            }
+            assert(repIdx >= -3);
+            outSeqs[i].offset = repIdx >= 0 ? outSeqs[repIdx].offset : repStartValue[-repIdx - 1];
+            if (outSeqs[i].rep == 4) {
+                --outSeqs[i].offset;
+            }
+        } else {
+            outSeqs[i].offset -= ZSTD_REP_NUM;
+        }
+
+        position += outSeqs[i].litLength;
+        outSeqs[i].matchPos = (unsigned int)position;
+        position += outSeqs[i].matchLength;
+    }
+    zc->seqCollector.seqIndex += seqsSize;
+}
+
+size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
+    size_t outSeqsSize, const void* src, size_t srcSize)
+{
+    const size_t dstCapacity = ZSTD_compressBound(srcSize);
+    void* dst = ZSTD_malloc(dstCapacity, ZSTD_defaultCMem);
+    SeqCollector seqCollector;
+
+    RETURN_ERROR_IF(dst == NULL, memory_allocation);
+
+    seqCollector.collectSequences = 1;
+    seqCollector.seqStart = outSeqs;
+    seqCollector.seqIndex = 0;
+    seqCollector.maxSequences = outSeqsSize;
+    zc->seqCollector = seqCollector;
+
+    ZSTD_compress2(zc, dst, dstCapacity, src, srcSize);
+    ZSTD_free(dst, ZSTD_defaultCMem);
+    return zc->seqCollector.seqIndex;
+}
+
+/* Returns true if the given block is a RLE block */
+static int ZSTD_isRLE(const BYTE *ip, size_t length) {
+    size_t i;
+    if (length < 2) return 1;
+    for (i = 1; i < length; ++i) {
+        if (ip[0] != ip[i]) return 0;
+    }
+    return 1;
+}
+
 static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
                                         void* dst, size_t dstCapacity,
-                                        const void* src, size_t srcSize)
+                                        const void* src, size_t srcSize, U32 frame)
 {
+    /* This the upper bound for the length of an rle block.
+     * This isn't the actual upper bound. Finding the real threshold
+     * needs further investigation.
+     */
+    const U32 rleMaxLength = 25;
     size_t cSize;
+    const BYTE* ip = (const BYTE*)src;
+    BYTE* op = (BYTE*)dst;
     DEBUGLOG(5, "ZSTD_compressBlock_internal (dstCapacity=%u, dictLimit=%u, nextToUpdate=%u)",
                 (unsigned)dstCapacity, (unsigned)zc->blockState.matchState.window.dictLimit,
                 (unsigned)zc->blockState.matchState.nextToUpdate);
@@ -2278,6 +2391,11 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
         if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; }
     }
 
+    if (zc->seqCollector.collectSequences) {
+        ZSTD_copyBlockSequences(zc);
+        return 0;
+    }
+
     /* encode sequences and literals */
     cSize = ZSTD_compressSequences(&zc->seqStore,
             &zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
@@ -2287,8 +2405,21 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
             zc->entropyWorkspace, HUF_WORKSPACE_SIZE /* statically allocated in resetCCtx */,
             zc->bmi2);
 
+    if (frame &&
+        /* We don't want to emit our first block as a RLE even if it qualifies because
+         * doing so will cause the decoder (cli only) to throw a "should consume all input error."
+         * This is only an issue for zstd <= v1.4.3
+         */
+        !zc->isFirstBlock &&
+        cSize < rleMaxLength &&
+        ZSTD_isRLE(ip, srcSize))
+    {
+        cSize = 1;
+        op[0] = ip[0];
+    }
+
 out:
-    if (!ZSTD_isError(cSize) && cSize != 0) {
+    if (!ZSTD_isError(cSize) && cSize > 1) {
         /* confirm repcodes and entropy tables when emitting a compressed block */
         ZSTD_compressedBlockState_t* const tmp = zc->blockState.prevCBlock;
         zc->blockState.prevCBlock = zc->blockState.nextCBlock;
@@ -2305,7 +2436,11 @@ out:
 }
 
 
-static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms, ZSTD_CCtx_params const* params, void const* ip, void const* iend)
+static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms,
+                                         ZSTD_cwksp* ws,
+                                         ZSTD_CCtx_params const* params,
+                                         void const* ip,
+                                         void const* iend)
 {
     if (ZSTD_window_needOverflowCorrection(ms->window, iend)) {
         U32 const maxDist = (U32)1 << params->cParams.windowLog;
@@ -2314,7 +2449,9 @@ static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms, ZSTD_CCtx_params
         ZSTD_STATIC_ASSERT(ZSTD_CHAINLOG_MAX <= 30);
         ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX_32 <= 30);
         ZSTD_STATIC_ASSERT(ZSTD_WINDOWLOG_MAX <= 31);
+        ZSTD_cwksp_mark_tables_dirty(ws);
         ZSTD_reduceIndex(ms, params, correction);
+        ZSTD_cwksp_mark_tables_clean(ws);
         if (ms->nextToUpdate < correction) ms->nextToUpdate = 0;
         else ms->nextToUpdate -= correction;
         /* invalidate dictionaries on overflow correction */
@@ -2323,7 +2460,6 @@ static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms, ZSTD_CCtx_params
     }
 }
 
-
 /*! ZSTD_compress_frameChunk() :
 *   Compress a chunk of data into one or multiple blocks.
 *   All blocks will be terminated, all input will be consumed.
@@ -2357,7 +2493,8 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
                         "not enough space to store compressed block");
         if (remaining < blockSize) blockSize = remaining;
 
-        ZSTD_overflowCorrectIfNeeded(ms, &cctx->appliedParams, ip, ip + blockSize);
+        ZSTD_overflowCorrectIfNeeded(
+            ms, &cctx->workspace, &cctx->appliedParams, ip, ip + blockSize);
         ZSTD_checkDictValidity(&ms->window, ip + blockSize, maxDist, &ms->loadedDictEnd, &ms->dictMatchState);
 
         /* Ensure hash/chain table insertion resumes no sooner than lowlimit */
@@ -2365,15 +2502,16 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
 
         {   size_t cSize = ZSTD_compressBlock_internal(cctx,
                                 op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize,
-                                ip, blockSize);
+                                ip, blockSize, 1 /* frame */);
             FORWARD_IF_ERROR(cSize);
-
             if (cSize == 0) {  /* block is not compressible */
                 cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
                 FORWARD_IF_ERROR(cSize);
             } else {
-                U32 const cBlockHeader24 = lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
-                MEM_writeLE24(op, cBlockHeader24);
+                const U32 cBlockHeader = cSize == 1 ?
+                    lastBlock + (((U32)bt_rle)<<1) + (U32)(blockSize << 3) :
+                    lastBlock + (((U32)bt_compressed)<<1) + (U32)(cSize << 3);
+                MEM_writeLE24(op, cBlockHeader);
                 cSize += ZSTD_blockHeaderSize;
             }
 
@@ -2383,6 +2521,7 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
             op += cSize;
             assert(dstCapacity >= cSize);
             dstCapacity -= cSize;
+            cctx->isFirstBlock = 0;
             DEBUGLOG(5, "ZSTD_compress_frameChunk: adding a block of size %u",
                         (unsigned)cSize);
     }   }
@@ -2393,25 +2532,25 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
 
 
 static size_t ZSTD_writeFrameHeader(void* dst, size_t dstCapacity,
-                                    ZSTD_CCtx_params params, U64 pledgedSrcSize, U32 dictID)
+                                    const ZSTD_CCtx_params* params, U64 pledgedSrcSize, U32 dictID)
 {   BYTE* const op = (BYTE*)dst;
     U32   const dictIDSizeCodeLength = (dictID>0) + (dictID>=256) + (dictID>=65536);   /* 0-3 */
-    U32   const dictIDSizeCode = params.fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength;   /* 0-3 */
-    U32   const checksumFlag = params.fParams.checksumFlag>0;
-    U32   const windowSize = (U32)1 << params.cParams.windowLog;
-    U32   const singleSegment = params.fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);
-    BYTE  const windowLogByte = (BYTE)((params.cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
-    U32   const fcsCode = params.fParams.contentSizeFlag ?
+    U32   const dictIDSizeCode = params->fParams.noDictIDFlag ? 0 : dictIDSizeCodeLength;   /* 0-3 */
+    U32   const checksumFlag = params->fParams.checksumFlag>0;
+    U32   const windowSize = (U32)1 << params->cParams.windowLog;
+    U32   const singleSegment = params->fParams.contentSizeFlag && (windowSize >= pledgedSrcSize);
+    BYTE  const windowLogByte = (BYTE)((params->cParams.windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN) << 3);
+    U32   const fcsCode = params->fParams.contentSizeFlag ?
                      (pledgedSrcSize>=256) + (pledgedSrcSize>=65536+256) + (pledgedSrcSize>=0xFFFFFFFFU) : 0;  /* 0-3 */
     BYTE  const frameHeaderDescriptionByte = (BYTE)(dictIDSizeCode + (checksumFlag<<2) + (singleSegment<<5) + (fcsCode<<6) );
     size_t pos=0;
 
-    assert(!(params.fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));
+    assert(!(params->fParams.contentSizeFlag && pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN));
     RETURN_ERROR_IF(dstCapacity < ZSTD_FRAMEHEADERSIZE_MAX, dstSize_tooSmall);
     DEBUGLOG(4, "ZSTD_writeFrameHeader : dictIDFlag : %u ; dictID : %u ; dictIDSizeCode : %u",
-                !params.fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);
+                !params->fParams.noDictIDFlag, (unsigned)dictID, (unsigned)dictIDSizeCode);
 
-    if (params.format == ZSTD_f_zstd1) {
+    if (params->format == ZSTD_f_zstd1) {
         MEM_writeLE32(dst, ZSTD_MAGICNUMBER);
         pos = 4;
     }
@@ -2477,7 +2616,7 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
                     "missing init (ZSTD_compressBegin)");
 
     if (frame && (cctx->stage==ZSTDcs_init)) {
-        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->appliedParams,
+        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams,
                                        cctx->pledgedSrcSizePlusOne-1, cctx->dictID);
         FORWARD_IF_ERROR(fhSize);
         assert(fhSize <= dstCapacity);
@@ -2497,13 +2636,15 @@ static size_t ZSTD_compressContinue_internal (ZSTD_CCtx* cctx,
 
     if (!frame) {
         /* overflow check and correction for block mode */
-        ZSTD_overflowCorrectIfNeeded(ms, &cctx->appliedParams, src, (BYTE const*)src + srcSize);
+        ZSTD_overflowCorrectIfNeeded(
+            ms, &cctx->workspace, &cctx->appliedParams,
+            src, (BYTE const*)src + srcSize);
     }
 
     DEBUGLOG(5, "ZSTD_compressContinue_internal (blockSize=%u)", (unsigned)cctx->blockSize);
     {   size_t const cSize = frame ?
                              ZSTD_compress_frameChunk (cctx, dst, dstCapacity, src, srcSize, lastFrameChunk) :
-                             ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize);
+                             ZSTD_compressBlock_internal (cctx, dst, dstCapacity, src, srcSize, 0 /* frame */);
         FORWARD_IF_ERROR(cSize);
         cctx->consumedSrcSize += srcSize;
         cctx->producedCSize += (cSize + fhSize);
@@ -2550,6 +2691,7 @@ size_t ZSTD_compressBlock(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const
  *  @return : 0, or an error code
  */
 static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
+                                         ZSTD_cwksp* ws,
                                          ZSTD_CCtx_params const* params,
                                          const void* src, size_t srcSize,
                                          ZSTD_dictTableLoadMethod_e dtlm)
@@ -2570,7 +2712,7 @@ static size_t ZSTD_loadDictionaryContent(ZSTD_matchState_t* ms,
         size_t const chunk = MIN(remaining, ZSTD_CHUNKSIZE_MAX);
         const BYTE* const ichunk = ip + chunk;
 
-        ZSTD_overflowCorrectIfNeeded(ms, params, ip, ichunk);
+        ZSTD_overflowCorrectIfNeeded(ms, ws, params, ip, ichunk);
 
         switch(params->cParams.strategy)
         {
@@ -2629,10 +2771,11 @@ static size_t ZSTD_checkDictNCount(short* normalizedCounter, unsigned dictMaxSym
 /*! ZSTD_loadZstdDictionary() :
  * @return : dictID, or an error code
  *  assumptions : magic number supposed already checked
- *                dictSize supposed > 8
+ *                dictSize supposed >= 8
  */
 static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
                                       ZSTD_matchState_t* ms,
+                                      ZSTD_cwksp* ws,
                                       ZSTD_CCtx_params const* params,
                                       const void* dict, size_t dictSize,
                                       ZSTD_dictTableLoadMethod_e dtlm,
@@ -2645,7 +2788,7 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
     size_t dictID;
 
     ZSTD_STATIC_ASSERT(HUF_WORKSPACE_SIZE >= (1<<MAX(MLFSELog,LLFSELog)));
-    assert(dictSize > 8);
+    assert(dictSize >= 8);
     assert(MEM_readLE32(dictPtr) == ZSTD_MAGIC_DICTIONARY);
 
     dictPtr += 4;   /* skip magic number */
@@ -2728,7 +2871,8 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
         bs->entropy.fse.offcode_repeatMode = FSE_repeat_valid;
         bs->entropy.fse.matchlength_repeatMode = FSE_repeat_valid;
         bs->entropy.fse.litlength_repeatMode = FSE_repeat_valid;
-        FORWARD_IF_ERROR(ZSTD_loadDictionaryContent(ms, params, dictPtr, dictContentSize, dtlm));
+        FORWARD_IF_ERROR(ZSTD_loadDictionaryContent(
+            ms, ws, params, dictPtr, dictContentSize, dtlm));
         return dictID;
     }
 }
@@ -2738,6 +2882,7 @@ static size_t ZSTD_loadZstdDictionary(ZSTD_compressedBlockState_t* bs,
 static size_t
 ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
                                ZSTD_matchState_t* ms,
+                               ZSTD_cwksp* ws,
                          const ZSTD_CCtx_params* params,
                          const void* dict, size_t dictSize,
                                ZSTD_dictContentType_e dictContentType,
@@ -2745,27 +2890,35 @@ ZSTD_compress_insertDictionary(ZSTD_compressedBlockState_t* bs,
                                void* workspace)
 {
     DEBUGLOG(4, "ZSTD_compress_insertDictionary (dictSize=%u)", (U32)dictSize);
-    if ((dict==NULL) || (dictSize<=8)) return 0;
+    if ((dict==NULL) || (dictSize<8)) {
+        RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong);
+        return 0;
+    }
 
     ZSTD_reset_compressedBlockState(bs);
 
     /* dict restricted modes */
     if (dictContentType == ZSTD_dct_rawContent)
-        return ZSTD_loadDictionaryContent(ms, params, dict, dictSize, dtlm);
+        return ZSTD_loadDictionaryContent(ms, ws, params, dict, dictSize, dtlm);
 
     if (MEM_readLE32(dict) != ZSTD_MAGIC_DICTIONARY) {
         if (dictContentType == ZSTD_dct_auto) {
             DEBUGLOG(4, "raw content dictionary detected");
-            return ZSTD_loadDictionaryContent(ms, params, dict, dictSize, dtlm);
+            return ZSTD_loadDictionaryContent(
+                ms, ws, params, dict, dictSize, dtlm);
         }
         RETURN_ERROR_IF(dictContentType == ZSTD_dct_fullDict, dictionary_wrong);
         assert(0);   /* impossible */
     }
 
     /* dict as full zstd dictionary */
-    return ZSTD_loadZstdDictionary(bs, ms, params, dict, dictSize, dtlm, workspace);
+    return ZSTD_loadZstdDictionary(
+        bs, ms, ws, params, dict, dictSize, dtlm, workspace);
 }
 
+#define ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF (128 KB)
+#define ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER (6)
+
 /*! ZSTD_compressBegin_internal() :
  * @return : 0, or an error code */
 static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
@@ -2773,23 +2926,34 @@ static size_t ZSTD_compressBegin_internal(ZSTD_CCtx* cctx,
                                     ZSTD_dictContentType_e dictContentType,
                                     ZSTD_dictTableLoadMethod_e dtlm,
                                     const ZSTD_CDict* cdict,
-                                    ZSTD_CCtx_params params, U64 pledgedSrcSize,
+                                    const ZSTD_CCtx_params* params, U64 pledgedSrcSize,
                                     ZSTD_buffered_policy_e zbuff)
 {
-    DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u", params.cParams.windowLog);
+    DEBUGLOG(4, "ZSTD_compressBegin_internal: wlog=%u", params->cParams.windowLog);
     /* params are supposed to be fully validated at this point */
-    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
+    assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
     assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
-
-    if (cdict && cdict->dictContentSize>0) {
+    if ( (cdict)
+      && (cdict->dictContentSize > 0)
+      && ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
+        || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
+        || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
+        || cdict->compressionLevel == 0)
+      && (params->attachDictPref != ZSTD_dictForceLoad) ) {
         return ZSTD_resetCCtx_usingCDict(cctx, cdict, params, pledgedSrcSize, zbuff);
     }
 
-    FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, params, pledgedSrcSize,
-                                     ZSTDcrp_continue, zbuff) );
-    {   size_t const dictID = ZSTD_compress_insertDictionary(
-                cctx->blockState.prevCBlock, &cctx->blockState.matchState,
-                &params, dict, dictSize, dictContentType, dtlm, cctx->entropyWorkspace);
+    FORWARD_IF_ERROR( ZSTD_resetCCtx_internal(cctx, *params, pledgedSrcSize,
+                                     ZSTDcrp_makeClean, zbuff) );
+    {   size_t const dictID = cdict ?
+                ZSTD_compress_insertDictionary(
+                        cctx->blockState.prevCBlock, &cctx->blockState.matchState,
+                        &cctx->workspace, params, cdict->dictContent, cdict->dictContentSize,
+                        dictContentType, dtlm, cctx->entropyWorkspace)
+              : ZSTD_compress_insertDictionary(
+                        cctx->blockState.prevCBlock, &cctx->blockState.matchState,
+                        &cctx->workspace, params, dict, dictSize,
+                        dictContentType, dtlm, cctx->entropyWorkspace);
         FORWARD_IF_ERROR(dictID);
         assert(dictID <= UINT_MAX);
         cctx->dictID = (U32)dictID;
@@ -2802,12 +2966,12 @@ size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
                                     ZSTD_dictContentType_e dictContentType,
                                     ZSTD_dictTableLoadMethod_e dtlm,
                                     const ZSTD_CDict* cdict,
-                                    ZSTD_CCtx_params params,
+                                    const ZSTD_CCtx_params* params,
                                     unsigned long long pledgedSrcSize)
 {
-    DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u", params.cParams.windowLog);
+    DEBUGLOG(4, "ZSTD_compressBegin_advanced_internal: wlog=%u", params->cParams.windowLog);
     /* compression parameters verification and optimization */
-    FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) );
+    FORWARD_IF_ERROR( ZSTD_checkCParams(params->cParams) );
     return ZSTD_compressBegin_internal(cctx,
                                        dict, dictSize, dictContentType, dtlm,
                                        cdict,
@@ -2822,21 +2986,21 @@ size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx,
                                    ZSTD_parameters params, unsigned long long pledgedSrcSize)
 {
     ZSTD_CCtx_params const cctxParams =
-            ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+            ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params);
     return ZSTD_compressBegin_advanced_internal(cctx,
                                             dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast,
                                             NULL /*cdict*/,
-                                            cctxParams, pledgedSrcSize);
+                                            &cctxParams, pledgedSrcSize);
 }
 
 size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel)
 {
     ZSTD_parameters const params = ZSTD_getParams(compressionLevel, ZSTD_CONTENTSIZE_UNKNOWN, dictSize);
     ZSTD_CCtx_params const cctxParams =
-            ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+            ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params);
     DEBUGLOG(4, "ZSTD_compressBegin_usingDict (dictSize=%u)", (unsigned)dictSize);
     return ZSTD_compressBegin_internal(cctx, dict, dictSize, ZSTD_dct_auto, ZSTD_dtlm_fast, NULL,
-                                       cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered);
+                                       &cctxParams, ZSTD_CONTENTSIZE_UNKNOWN, ZSTDb_not_buffered);
 }
 
 size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel)
@@ -2859,7 +3023,7 @@ static size_t ZSTD_writeEpilogue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity)
 
     /* special case : empty frame */
     if (cctx->stage == ZSTDcs_init) {
-        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, cctx->appliedParams, 0, 0);
+        fhSize = ZSTD_writeFrameHeader(dst, dstCapacity, &cctx->appliedParams, 0, 0);
         FORWARD_IF_ERROR(fhSize);
         dstCapacity -= fhSize;
         op += fhSize;
@@ -2920,13 +3084,13 @@ static size_t ZSTD_compress_internal (ZSTD_CCtx* cctx,
                                       ZSTD_parameters params)
 {
     ZSTD_CCtx_params const cctxParams =
-            ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+            ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params);
     DEBUGLOG(4, "ZSTD_compress_internal");
     return ZSTD_compress_advanced_internal(cctx,
                                            dst, dstCapacity,
                                            src, srcSize,
                                            dict, dictSize,
-                                           cctxParams);
+                                           &cctxParams);
 }
 
 size_t ZSTD_compress_advanced (ZSTD_CCtx* cctx,
@@ -2950,7 +3114,7 @@ size_t ZSTD_compress_advanced_internal(
         void* dst, size_t dstCapacity,
         const void* src, size_t srcSize,
         const void* dict,size_t dictSize,
-        ZSTD_CCtx_params params)
+        const ZSTD_CCtx_params* params)
 {
     DEBUGLOG(4, "ZSTD_compress_advanced_internal (srcSize:%u)", (unsigned)srcSize);
     FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
@@ -2966,9 +3130,9 @@ size_t ZSTD_compress_usingDict(ZSTD_CCtx* cctx,
                                int compressionLevel)
 {
     ZSTD_parameters const params = ZSTD_getParams(compressionLevel, srcSize + (!srcSize), dict ? dictSize : 0);
-    ZSTD_CCtx_params cctxParams = ZSTD_assignParamsToCCtxParams(cctx->requestedParams, params);
+    ZSTD_CCtx_params cctxParams = ZSTD_assignParamsToCCtxParams(&cctx->requestedParams, params);
     assert(params.fParams.contentSizeFlag == 1);
-    return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, cctxParams);
+    return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, dict, dictSize, &cctxParams);
 }
 
 size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
@@ -3003,8 +3167,11 @@ size_t ZSTD_estimateCDictSize_advanced(
         ZSTD_dictLoadMethod_e dictLoadMethod)
 {
     DEBUGLOG(5, "sizeof(ZSTD_CDict) : %u", (unsigned)sizeof(ZSTD_CDict));
-    return sizeof(ZSTD_CDict) + HUF_WORKSPACE_SIZE + ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0)
-           + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize);
+    return ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
+         + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
+         + ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0)
+         + (dictLoadMethod == ZSTD_dlm_byRef ? 0
+            : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void *))));
 }
 
 size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel)
@@ -3017,7 +3184,9 @@ size_t ZSTD_sizeof_CDict(const ZSTD_CDict* cdict)
 {
     if (cdict==NULL) return 0;   /* support sizeof on NULL */
     DEBUGLOG(5, "sizeof(*cdict) : %u", (unsigned)sizeof(*cdict));
-    return cdict->workspaceSize + (cdict->dictBuffer ? cdict->dictContentSize : 0) + sizeof(*cdict);
+    /* cdict may be in the workspace */
+    return (cdict->workspace.workspace == cdict ? 0 : sizeof(*cdict))
+        + ZSTD_cwksp_sizeof(&cdict->workspace);
 }
 
 static size_t ZSTD_initCDict_internal(
@@ -3031,28 +3200,29 @@ static size_t ZSTD_initCDict_internal(
     assert(!ZSTD_checkCParams(cParams));
     cdict->matchState.cParams = cParams;
     if ((dictLoadMethod == ZSTD_dlm_byRef) || (!dictBuffer) || (!dictSize)) {
-        cdict->dictBuffer = NULL;
         cdict->dictContent = dictBuffer;
     } else {
-        void* const internalBuffer = ZSTD_malloc(dictSize, cdict->customMem);
-        cdict->dictBuffer = internalBuffer;
-        cdict->dictContent = internalBuffer;
+         void *internalBuffer = ZSTD_cwksp_reserve_object(&cdict->workspace, ZSTD_cwksp_align(dictSize, sizeof(void*)));
         RETURN_ERROR_IF(!internalBuffer, memory_allocation);
+        cdict->dictContent = internalBuffer;
         memcpy(internalBuffer, dictBuffer, dictSize);
     }
     cdict->dictContentSize = dictSize;
 
+    cdict->entropyWorkspace = (U32*)ZSTD_cwksp_reserve_object(&cdict->workspace, HUF_WORKSPACE_SIZE);
+
+
     /* Reset the state to no dictionary */
     ZSTD_reset_compressedBlockState(&cdict->cBlockState);
-    {   void* const end = ZSTD_reset_matchState(&cdict->matchState,
-                            (U32*)cdict->workspace + HUF_WORKSPACE_SIZE_U32,
-                            &cParams,
-                             ZSTDcrp_continue, ZSTD_resetTarget_CDict);
-        assert(end == (char*)cdict->workspace + cdict->workspaceSize);
-        (void)end;
-    }
+    FORWARD_IF_ERROR(ZSTD_reset_matchState(
+        &cdict->matchState,
+        &cdict->workspace,
+        &cParams,
+        ZSTDcrp_makeClean,
+        ZSTDirp_reset,
+        ZSTD_resetTarget_CDict));
     /* (Maybe) load the dictionary
-     * Skips loading the dictionary if it is <= 8 bytes.
+     * Skips loading the dictionary if it is < 8 bytes.
      */
     {   ZSTD_CCtx_params params;
         memset(&params, 0, sizeof(params));
@@ -3060,9 +3230,9 @@ static size_t ZSTD_initCDict_internal(
         params.fParams.contentSizeFlag = 1;
         params.cParams = cParams;
         {   size_t const dictID = ZSTD_compress_insertDictionary(
-                    &cdict->cBlockState, &cdict->matchState, &params,
-                    cdict->dictContent, cdict->dictContentSize,
-                    dictContentType, ZSTD_dtlm_full, cdict->workspace);
+                    &cdict->cBlockState, &cdict->matchState, &cdict->workspace,
+                    &params, cdict->dictContent, cdict->dictContentSize,
+                    dictContentType, ZSTD_dtlm_full, cdict->entropyWorkspace);
             FORWARD_IF_ERROR(dictID);
             assert(dictID <= (size_t)(U32)-1);
             cdict->dictID = (U32)dictID;
@@ -3080,18 +3250,29 @@ ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,
     DEBUGLOG(3, "ZSTD_createCDict_advanced, mode %u", (unsigned)dictContentType);
     if (!customMem.customAlloc ^ !customMem.customFree) return NULL;
 
-    {   ZSTD_CDict* const cdict = (ZSTD_CDict*)ZSTD_malloc(sizeof(ZSTD_CDict), customMem);
-        size_t const workspaceSize = HUF_WORKSPACE_SIZE + ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0);
+    {   size_t const workspaceSize =
+            ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict)) +
+            ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE) +
+            ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0) +
+            (dictLoadMethod == ZSTD_dlm_byRef ? 0
+             : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))));
         void* const workspace = ZSTD_malloc(workspaceSize, customMem);
+        ZSTD_cwksp ws;
+        ZSTD_CDict* cdict;
 
-        if (!cdict || !workspace) {
-            ZSTD_free(cdict, customMem);
+        if (!workspace) {
             ZSTD_free(workspace, customMem);
             return NULL;
         }
+
+        ZSTD_cwksp_init(&ws, workspace, workspaceSize);
+
+        cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));
+        assert(cdict != NULL);
+        ZSTD_cwksp_move(&cdict->workspace, &ws);
         cdict->customMem = customMem;
-        cdict->workspace = workspace;
-        cdict->workspaceSize = workspaceSize;
+        cdict->compressionLevel = 0; /* signals advanced API usage */
+
         if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
                                         dictBuffer, dictSize,
                                         dictLoadMethod, dictContentType,
@@ -3107,9 +3288,12 @@ ZSTD_CDict* ZSTD_createCDict_advanced(const void* dictBuffer, size_t dictSize,
 ZSTD_CDict* ZSTD_createCDict(const void* dict, size_t dictSize, int compressionLevel)
 {
     ZSTD_compressionParameters cParams = ZSTD_getCParams(compressionLevel, 0, dictSize);
-    return ZSTD_createCDict_advanced(dict, dictSize,
-                                     ZSTD_dlm_byCopy, ZSTD_dct_auto,
-                                     cParams, ZSTD_defaultCMem);
+    ZSTD_CDict* cdict = ZSTD_createCDict_advanced(dict, dictSize,
+                                                  ZSTD_dlm_byCopy, ZSTD_dct_auto,
+                                                  cParams, ZSTD_defaultCMem);
+    if (cdict)
+        cdict->compressionLevel = compressionLevel == 0 ? ZSTD_CLEVEL_DEFAULT : compressionLevel;
+    return cdict;
 }
 
 ZSTD_CDict* ZSTD_createCDict_byReference(const void* dict, size_t dictSize, int compressionLevel)
@@ -3124,9 +3308,11 @@ size_t ZSTD_freeCDict(ZSTD_CDict* cdict)
 {
     if (cdict==NULL) return 0;   /* support free on NULL */
     {   ZSTD_customMem const cMem = cdict->customMem;
-        ZSTD_free(cdict->workspace, cMem);
-        ZSTD_free(cdict->dictBuffer, cMem);
-        ZSTD_free(cdict, cMem);
+        int cdictInWorkspace = ZSTD_cwksp_owns_buffer(&cdict->workspace, cdict);
+        ZSTD_cwksp_free(&cdict->workspace, cMem);
+        if (!cdictInWorkspace) {
+            ZSTD_free(cdict, cMem);
+        }
         return 0;
     }
 }
@@ -3152,28 +3338,30 @@ const ZSTD_CDict* ZSTD_initStaticCDict(
                                  ZSTD_compressionParameters cParams)
 {
     size_t const matchStateSize = ZSTD_sizeof_matchState(&cParams, /* forCCtx */ 0);
-    size_t const neededSize = sizeof(ZSTD_CDict) + (dictLoadMethod == ZSTD_dlm_byRef ? 0 : dictSize)
-                            + HUF_WORKSPACE_SIZE + matchStateSize;
-    ZSTD_CDict* const cdict = (ZSTD_CDict*) workspace;
-    void* ptr;
+    size_t const neededSize = ZSTD_cwksp_alloc_size(sizeof(ZSTD_CDict))
+                            + (dictLoadMethod == ZSTD_dlm_byRef ? 0
+                               : ZSTD_cwksp_alloc_size(ZSTD_cwksp_align(dictSize, sizeof(void*))))
+                            + ZSTD_cwksp_alloc_size(HUF_WORKSPACE_SIZE)
+                            + matchStateSize;
+    ZSTD_CDict* cdict;
+
     if ((size_t)workspace & 7) return NULL;  /* 8-aligned */
+
+    {
+        ZSTD_cwksp ws;
+        ZSTD_cwksp_init(&ws, workspace, workspaceSize);
+        cdict = (ZSTD_CDict*)ZSTD_cwksp_reserve_object(&ws, sizeof(ZSTD_CDict));
+        if (cdict == NULL) return NULL;
+        ZSTD_cwksp_move(&cdict->workspace, &ws);
+    }
+
     DEBUGLOG(4, "(workspaceSize < neededSize) : (%u < %u) => %u",
         (unsigned)workspaceSize, (unsigned)neededSize, (unsigned)(workspaceSize < neededSize));
     if (workspaceSize < neededSize) return NULL;
 
-    if (dictLoadMethod == ZSTD_dlm_byCopy) {
-        memcpy(cdict+1, dict, dictSize);
-        dict = cdict+1;
-        ptr = (char*)workspace + sizeof(ZSTD_CDict) + dictSize;
-    } else {
-        ptr = cdict+1;
-    }
-    cdict->workspace = ptr;
-    cdict->workspaceSize = HUF_WORKSPACE_SIZE + matchStateSize;
-
     if (ZSTD_isError( ZSTD_initCDict_internal(cdict,
                                               dict, dictSize,
-                                              ZSTD_dlm_byRef, dictContentType,
+                                              dictLoadMethod, dictContentType,
                                               cParams) ))
         return NULL;
 
@@ -3195,7 +3383,15 @@ size_t ZSTD_compressBegin_usingCDict_advanced(
     DEBUGLOG(4, "ZSTD_compressBegin_usingCDict_advanced");
     RETURN_ERROR_IF(cdict==NULL, dictionary_wrong);
     {   ZSTD_CCtx_params params = cctx->requestedParams;
-        params.cParams = ZSTD_getCParamsFromCDict(cdict);
+        params.cParams = ( pledgedSrcSize < ZSTD_USE_CDICT_PARAMS_SRCSIZE_CUTOFF
+                        || pledgedSrcSize < cdict->dictContentSize * ZSTD_USE_CDICT_PARAMS_DICTSIZE_MULTIPLIER
+                        || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN
+                        || cdict->compressionLevel == 0 )
+                      && (params.attachDictPref != ZSTD_dictForceLoad) ?
+                ZSTD_getCParamsFromCDict(cdict)
+              : ZSTD_getCParams(cdict->compressionLevel,
+                                pledgedSrcSize,
+                                cdict->dictContentSize);
         /* Increase window log to fit the entire dictionary and source if the
          * source size is known. Limit the increase to 19, which is the
          * window log for compression level 1 with the largest source size.
@@ -3209,7 +3405,7 @@ size_t ZSTD_compressBegin_usingCDict_advanced(
         return ZSTD_compressBegin_internal(cctx,
                                            NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast,
                                            cdict,
-                                           params, pledgedSrcSize,
+                                           &params, pledgedSrcSize,
                                            ZSTDb_not_buffered);
     }
 }
@@ -3300,7 +3496,7 @@ static size_t ZSTD_resetCStream_internal(ZSTD_CStream* cctx,
     FORWARD_IF_ERROR( ZSTD_compressBegin_internal(cctx,
                                          dict, dictSize, dictContentType, ZSTD_dtlm_fast,
                                          cdict,
-                                         params, pledgedSrcSize,
+                                         &params, pledgedSrcSize,
                                          ZSTDb_buffered) );
 
     cctx->inToCompress = 0;
@@ -3334,13 +3530,14 @@ size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pss)
  *  Assumption 2 : either dict, or cdict, is defined, not both */
 size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
                     const void* dict, size_t dictSize, const ZSTD_CDict* cdict,
-                    ZSTD_CCtx_params params, unsigned long long pledgedSrcSize)
+                    const ZSTD_CCtx_params* params,
+                    unsigned long long pledgedSrcSize)
 {
     DEBUGLOG(4, "ZSTD_initCStream_internal");
     FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
     FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
-    assert(!ZSTD_isError(ZSTD_checkCParams(params.cParams)));
-    zcs->requestedParams = params;
+    assert(!ZSTD_isError(ZSTD_checkCParams(params->cParams)));
+    zcs->requestedParams = *params;
     assert(!((dict) && (cdict)));  /* either dict or cdict, not both */
     if (dict) {
         FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) );
@@ -3379,7 +3576,7 @@ size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict)
 /* ZSTD_initCStream_advanced() :
  * pledgedSrcSize must be exact.
  * if srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN.
- * dict is loaded with default parameters ZSTD_dm_auto and ZSTD_dlm_byCopy. */
+ * dict is loaded with default parameters ZSTD_dct_auto and ZSTD_dlm_byCopy. */
 size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
                                  const void* dict, size_t dictSize,
                                  ZSTD_parameters params, unsigned long long pss)
@@ -3393,7 +3590,7 @@ size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
     FORWARD_IF_ERROR( ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only) );
     FORWARD_IF_ERROR( ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize) );
     FORWARD_IF_ERROR( ZSTD_checkCParams(params.cParams) );
-    zcs->requestedParams = ZSTD_assignParamsToCCtxParams(zcs->requestedParams, params);
+    zcs->requestedParams = ZSTD_assignParamsToCCtxParams(&zcs->requestedParams, params);
     FORWARD_IF_ERROR( ZSTD_CCtx_loadDictionary(zcs, dict, dictSize) );
     return 0;
 }
@@ -3643,7 +3840,7 @@ size_t ZSTD_compressStream2( ZSTD_CCtx* cctx,
             if (cctx->mtctx == NULL) {
                 DEBUGLOG(4, "ZSTD_compressStream2: creating new mtctx for nbWorkers=%u",
                             params.nbWorkers);
-                cctx->mtctx = ZSTDMT_createCCtx_advanced(params.nbWorkers, cctx->customMem);
+                cctx->mtctx = ZSTDMT_createCCtx_advanced((U32)params.nbWorkers, cctx->customMem);
                 RETURN_ERROR_IF(cctx->mtctx == NULL, memory_allocation);
             }
             /* mt compression */
@@ -3771,8 +3968,8 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     { 19, 12, 13,  1,  6,  1, ZSTD_fast    },  /* base for negative levels */
     { 19, 13, 14,  1,  7,  0, ZSTD_fast    },  /* level  1 */
     { 20, 15, 16,  1,  6,  0, ZSTD_fast    },  /* level  2 */
-    { 21, 16, 17,  1,  5,  1, ZSTD_dfast   },  /* level  3 */
-    { 21, 18, 18,  1,  5,  1, ZSTD_dfast   },  /* level  4 */
+    { 21, 16, 17,  1,  5,  0, ZSTD_dfast   },  /* level  3 */
+    { 21, 18, 18,  1,  5,  0, ZSTD_dfast   },  /* level  4 */
     { 21, 18, 19,  2,  5,  2, ZSTD_greedy  },  /* level  5 */
     { 21, 19, 19,  3,  5,  4, ZSTD_greedy  },  /* level  6 */
     { 21, 19, 19,  3,  5,  8, ZSTD_lazy    },  /* level  7 */
@@ -3796,8 +3993,8 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     /* W,  C,  H,  S,  L,  T, strat */
     { 18, 12, 13,  1,  5,  1, ZSTD_fast    },  /* base for negative levels */
     { 18, 13, 14,  1,  6,  0, ZSTD_fast    },  /* level  1 */
-    { 18, 14, 14,  1,  5,  1, ZSTD_dfast   },  /* level  2 */
-    { 18, 16, 16,  1,  4,  1, ZSTD_dfast   },  /* level  3 */
+    { 18, 14, 14,  1,  5,  0, ZSTD_dfast   },  /* level  2 */
+    { 18, 16, 16,  1,  4,  0, ZSTD_dfast   },  /* level  3 */
     { 18, 16, 17,  2,  5,  2, ZSTD_greedy  },  /* level  4.*/
     { 18, 18, 18,  3,  5,  2, ZSTD_greedy  },  /* level  5.*/
     { 18, 18, 19,  3,  5,  4, ZSTD_lazy    },  /* level  6.*/
@@ -3823,8 +4020,8 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     { 17, 12, 12,  1,  5,  1, ZSTD_fast    },  /* base for negative levels */
     { 17, 12, 13,  1,  6,  0, ZSTD_fast    },  /* level  1 */
     { 17, 13, 15,  1,  5,  0, ZSTD_fast    },  /* level  2 */
-    { 17, 15, 16,  2,  5,  1, ZSTD_dfast   },  /* level  3 */
-    { 17, 17, 17,  2,  4,  1, ZSTD_dfast   },  /* level  4 */
+    { 17, 15, 16,  2,  5,  0, ZSTD_dfast   },  /* level  3 */
+    { 17, 17, 17,  2,  4,  0, ZSTD_dfast   },  /* level  4 */
     { 17, 16, 17,  3,  4,  2, ZSTD_greedy  },  /* level  5 */
     { 17, 17, 17,  3,  4,  4, ZSTD_lazy    },  /* level  6 */
     { 17, 17, 17,  3,  4,  8, ZSTD_lazy2   },  /* level  7 */
@@ -3849,7 +4046,7 @@ static const ZSTD_compressionParameters ZSTD_defaultCParameters[4][ZSTD_MAX_CLEV
     { 14, 12, 13,  1,  5,  1, ZSTD_fast    },  /* base for negative levels */
     { 14, 14, 15,  1,  5,  0, ZSTD_fast    },  /* level  1 */
     { 14, 14, 15,  1,  4,  0, ZSTD_fast    },  /* level  2 */
-    { 14, 14, 15,  2,  4,  1, ZSTD_dfast   },  /* level  3 */
+    { 14, 14, 15,  2,  4,  0, ZSTD_dfast   },  /* level  3 */
     { 14, 14, 14,  4,  4,  2, ZSTD_greedy  },  /* level  4 */
     { 14, 14, 14,  3,  4,  4, ZSTD_lazy    },  /* level  5.*/
     { 14, 14, 14,  4,  4,  8, ZSTD_lazy2   },  /* level  6 */
diff --git a/thirdparty/zstd/compress/zstd_compress_internal.h b/thirdparty/zstd/compress/zstd_compress_internal.h
index 6d623cc6be..14036f873f 100644
--- a/thirdparty/zstd/compress/zstd_compress_internal.h
+++ b/thirdparty/zstd/compress/zstd_compress_internal.h
@@ -19,6 +19,7 @@
 *  Dependencies
 ***************************************/
 #include "zstd_internal.h"
+#include "zstd_cwksp.h"
 #ifdef ZSTD_MULTITHREAD
 #  include "zstdmt_compress.h"
 #endif
@@ -192,6 +193,13 @@ typedef struct {
   size_t capacity; /* The capacity starting from `seq` pointer */
 } rawSeqStore_t;
 
+typedef struct {
+    int collectSequences;
+    ZSTD_Sequence* seqStart;
+    size_t seqIndex;
+    size_t maxSequences;
+} SeqCollector;
+
 struct ZSTD_CCtx_params_s {
     ZSTD_format_e format;
     ZSTD_compressionParameters cParams;
@@ -203,6 +211,9 @@ struct ZSTD_CCtx_params_s {
     size_t targetCBlockSize;   /* Tries to fit compressed block size to be around targetCBlockSize.
                                 * No target when targetCBlockSize == 0.
                                 * There is no guarantee on compressed block size */
+    int srcSizeHint;           /* User's best guess of source size.
+                                * Hint is not valid when srcSizeHint == 0.
+                                * There is no guarantee that hint is close to actual source size */
 
     ZSTD_dictAttachPref_e attachDictPref;
     ZSTD_literalCompressionMode_e literalCompressionMode;
@@ -228,9 +239,7 @@ struct ZSTD_CCtx_s {
     ZSTD_CCtx_params appliedParams;
     U32   dictID;
 
-    int workSpaceOversizedDuration;
-    void* workSpace;
-    size_t workSpaceSize;
+    ZSTD_cwksp workspace; /* manages buffer for dynamic allocations */
     size_t blockSize;
     unsigned long long pledgedSrcSizePlusOne;  /* this way, 0 (default) == unknown */
     unsigned long long consumedSrcSize;
@@ -238,6 +247,8 @@ struct ZSTD_CCtx_s {
     XXH64_state_t xxhState;
     ZSTD_customMem customMem;
     size_t staticSize;
+    SeqCollector seqCollector;
+    int isFirstBlock;
 
     seqStore_t seqStore;      /* sequences storage ptrs */
     ldmState_t ldmState;      /* long distance matching state */
@@ -337,26 +348,57 @@ MEM_STATIC size_t ZSTD_minGain(size_t srcSize, ZSTD_strategy strat)
     return (srcSize >> minlog) + 2;
 }
 
+/*! ZSTD_safecopyLiterals() :
+ *  memcpy() function that won't read beyond more than WILDCOPY_OVERLENGTH bytes past ilimit_w.
+ *  Only called when the sequence ends past ilimit_w, so it only needs to be optimized for single
+ *  large copies.
+ */
+static void ZSTD_safecopyLiterals(BYTE* op, BYTE const* ip, BYTE const* const iend, BYTE const* ilimit_w) {
+    assert(iend > ilimit_w);
+    if (ip <= ilimit_w) {
+        ZSTD_wildcopy(op, ip, ilimit_w - ip, ZSTD_no_overlap);
+        op += ilimit_w - ip;
+        ip = ilimit_w;
+    }
+    while (ip < iend) *op++ = *ip++;
+}
+
 /*! ZSTD_storeSeq() :
- *  Store a sequence (literal length, literals, offset code and match length code) into seqStore_t.
- *  `offsetCode` : distance to match + 3 (values 1-3 are repCodes).
+ *  Store a sequence (litlen, litPtr, offCode and mlBase) into seqStore_t.
+ *  `offCode` : distance to match + ZSTD_REP_MOVE (values <= ZSTD_REP_MOVE are repCodes).
  *  `mlBase` : matchLength - MINMATCH
+ *  Allowed to overread literals up to litLimit.
 */
-MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const void* literals, U32 offsetCode, size_t mlBase)
+HINT_INLINE UNUSED_ATTR
+void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const BYTE* literals, const BYTE* litLimit, U32 offCode, size_t mlBase)
 {
+    BYTE const* const litLimit_w = litLimit - WILDCOPY_OVERLENGTH;
+    BYTE const* const litEnd = literals + litLength;
 #if defined(DEBUGLEVEL) && (DEBUGLEVEL >= 6)
     static const BYTE* g_start = NULL;
     if (g_start==NULL) g_start = (const BYTE*)literals;  /* note : index only works for compression within a single segment */
     {   U32 const pos = (U32)((const BYTE*)literals - g_start);
         DEBUGLOG(6, "Cpos%7u :%3u literals, match%4u bytes at offCode%7u",
-               pos, (U32)litLength, (U32)mlBase+MINMATCH, (U32)offsetCode);
+               pos, (U32)litLength, (U32)mlBase+MINMATCH, (U32)offCode);
     }
 #endif
     assert((size_t)(seqStorePtr->sequences - seqStorePtr->sequencesStart) < seqStorePtr->maxNbSeq);
     /* copy Literals */
     assert(seqStorePtr->maxNbLit <= 128 KB);
     assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + seqStorePtr->maxNbLit);
-    ZSTD_wildcopy(seqStorePtr->lit, literals, (ptrdiff_t)litLength, ZSTD_no_overlap);
+    assert(literals + litLength <= litLimit);
+    if (litEnd <= litLimit_w) {
+        /* Common case we can use wildcopy.
+	 * First copy 16 bytes, because literals are likely short.
+	 */
+        assert(WILDCOPY_OVERLENGTH >= 16);
+        ZSTD_copy16(seqStorePtr->lit, literals);
+        if (litLength > 16) {
+            ZSTD_wildcopy(seqStorePtr->lit+16, literals+16, (ptrdiff_t)litLength-16, ZSTD_no_overlap);
+        }
+    } else {
+        ZSTD_safecopyLiterals(seqStorePtr->lit, literals, litEnd, litLimit_w);
+    }
     seqStorePtr->lit += litLength;
 
     /* literal Length */
@@ -368,7 +410,7 @@ MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const v
     seqStorePtr->sequences[0].litLength = (U16)litLength;
 
     /* match offset */
-    seqStorePtr->sequences[0].offset = offsetCode + 1;
+    seqStorePtr->sequences[0].offset = offCode + 1;
 
     /* match Length */
     if (mlBase>0xFFFF) {
@@ -910,7 +952,7 @@ ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
 size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
                      const void* dict, size_t dictSize,
                      const ZSTD_CDict* cdict,
-                     ZSTD_CCtx_params  params, unsigned long long pledgedSrcSize);
+                     const ZSTD_CCtx_params* params, unsigned long long pledgedSrcSize);
 
 void ZSTD_resetSeqStore(seqStore_t* ssPtr);
 
@@ -925,7 +967,7 @@ size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
                                     ZSTD_dictContentType_e dictContentType,
                                     ZSTD_dictTableLoadMethod_e dtlm,
                                     const ZSTD_CDict* cdict,
-                                    ZSTD_CCtx_params params,
+                                    const ZSTD_CCtx_params* params,
                                     unsigned long long pledgedSrcSize);
 
 /* ZSTD_compress_advanced_internal() :
@@ -934,7 +976,7 @@ size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
                                        void* dst, size_t dstCapacity,
                                  const void* src, size_t srcSize,
                                  const void* dict,size_t dictSize,
-                                 ZSTD_CCtx_params params);
+                                 const ZSTD_CCtx_params* params);
 
 
 /* ZSTD_writeLastEmptyBlock() :
diff --git a/thirdparty/zstd/compress/zstd_compress_literals.c b/thirdparty/zstd/compress/zstd_compress_literals.c
index eb3e5a44bc..6c13331182 100644
--- a/thirdparty/zstd/compress/zstd_compress_literals.c
+++ b/thirdparty/zstd/compress/zstd_compress_literals.c
@@ -70,7 +70,7 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
                               ZSTD_strategy strategy, int disableLiteralCompression,
                               void* dst, size_t dstCapacity,
                         const void* src, size_t srcSize,
-                              void* workspace, size_t wkspSize,
+                              void* entropyWorkspace, size_t entropyWorkspaceSize,
                         const int bmi2)
 {
     size_t const minGain = ZSTD_minGain(srcSize, strategy);
@@ -99,10 +99,15 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
     {   HUF_repeat repeat = prevHuf->repeatMode;
         int const preferRepeat = strategy < ZSTD_lazy ? srcSize <= 1024 : 0;
         if (repeat == HUF_repeat_valid && lhSize == 3) singleStream = 1;
-        cLitSize = singleStream ? HUF_compress1X_repeat(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11,
-                                      workspace, wkspSize, (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2)
-                                : HUF_compress4X_repeat(ostart+lhSize, dstCapacity-lhSize, src, srcSize, 255, 11,
-                                      workspace, wkspSize, (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
+        cLitSize = singleStream ?
+            HUF_compress1X_repeat(
+                ostart+lhSize, dstCapacity-lhSize, src, srcSize,
+                255, 11, entropyWorkspace, entropyWorkspaceSize,
+                (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2) :
+            HUF_compress4X_repeat(
+                ostart+lhSize, dstCapacity-lhSize, src, srcSize,
+                255, 11, entropyWorkspace, entropyWorkspaceSize,
+                (HUF_CElt*)nextHuf->CTable, &repeat, preferRepeat, bmi2);
         if (repeat != HUF_repeat_none) {
             /* reused the existing table */
             hType = set_repeat;
diff --git a/thirdparty/zstd/compress/zstd_compress_literals.h b/thirdparty/zstd/compress/zstd_compress_literals.h
index 7adbecc0be..97273d7cfd 100644
--- a/thirdparty/zstd/compress/zstd_compress_literals.h
+++ b/thirdparty/zstd/compress/zstd_compress_literals.h
@@ -23,7 +23,7 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
                               ZSTD_strategy strategy, int disableLiteralCompression,
                               void* dst, size_t dstCapacity,
                         const void* src, size_t srcSize,
-                              void* workspace, size_t wkspSize,
+                              void* entropyWorkspace, size_t entropyWorkspaceSize,
                         const int bmi2);
 
 #endif /* ZSTD_COMPRESS_LITERALS_H */
diff --git a/thirdparty/zstd/compress/zstd_compress_sequences.c b/thirdparty/zstd/compress/zstd_compress_sequences.c
index 3c3deae08c..0ff7a26823 100644
--- a/thirdparty/zstd/compress/zstd_compress_sequences.c
+++ b/thirdparty/zstd/compress/zstd_compress_sequences.c
@@ -222,7 +222,7 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
                 const BYTE* codeTable, size_t nbSeq,
                 const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
                 const FSE_CTable* prevCTable, size_t prevCTableSize,
-                void* workspace, size_t workspaceSize)
+                void* entropyWorkspace, size_t entropyWorkspaceSize)
 {
     BYTE* op = (BYTE*)dst;
     const BYTE* const oend = op + dstCapacity;
@@ -238,7 +238,7 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
         memcpy(nextCTable, prevCTable, prevCTableSize);
         return 0;
     case set_basic:
-        FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, workspace, workspaceSize));  /* note : could be pre-calculated */
+        FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, defaultNorm, defaultMax, defaultNormLog, entropyWorkspace, entropyWorkspaceSize));  /* note : could be pre-calculated */
         return 0;
     case set_compressed: {
         S16 norm[MaxSeq + 1];
@@ -252,7 +252,7 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
         FORWARD_IF_ERROR(FSE_normalizeCount(norm, tableLog, count, nbSeq_1, max));
         {   size_t const NCountSize = FSE_writeNCount(op, oend - op, norm, max, tableLog);   /* overflow protected */
             FORWARD_IF_ERROR(NCountSize);
-            FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, norm, max, tableLog, workspace, workspaceSize));
+            FORWARD_IF_ERROR(FSE_buildCTable_wksp(nextCTable, norm, max, tableLog, entropyWorkspace, entropyWorkspaceSize));
             return NCountSize;
         }
     }
diff --git a/thirdparty/zstd/compress/zstd_compress_sequences.h b/thirdparty/zstd/compress/zstd_compress_sequences.h
index f5234d94c8..57e8e367b0 100644
--- a/thirdparty/zstd/compress/zstd_compress_sequences.h
+++ b/thirdparty/zstd/compress/zstd_compress_sequences.h
@@ -35,7 +35,7 @@ ZSTD_buildCTable(void* dst, size_t dstCapacity,
                 const BYTE* codeTable, size_t nbSeq,
                 const S16* defaultNorm, U32 defaultNormLog, U32 defaultMax,
                 const FSE_CTable* prevCTable, size_t prevCTableSize,
-                void* workspace, size_t workspaceSize);
+                void* entropyWorkspace, size_t entropyWorkspaceSize);
 
 size_t ZSTD_encodeSequences(
             void* dst, size_t dstCapacity,
diff --git a/thirdparty/zstd/compress/zstd_cwksp.h b/thirdparty/zstd/compress/zstd_cwksp.h
new file mode 100644
index 0000000000..fc9765bd3f
--- /dev/null
+++ b/thirdparty/zstd/compress/zstd_cwksp.h
@@ -0,0 +1,535 @@
+/*
+ * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
+ * All rights reserved.
+ *
+ * This source code is licensed under both the BSD-style license (found in the
+ * LICENSE file in the root directory of this source tree) and the GPLv2 (found
+ * in the COPYING file in the root directory of this source tree).
+ * You may select, at your option, one of the above-listed licenses.
+ */
+
+#ifndef ZSTD_CWKSP_H
+#define ZSTD_CWKSP_H
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include "zstd_internal.h"
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-*************************************
+*  Constants
+***************************************/
+
+/* define "workspace is too large" as this number of times larger than needed */
+#define ZSTD_WORKSPACETOOLARGE_FACTOR 3
+
+/* when workspace is continuously too large
+ * during at least this number of times,
+ * context's memory usage is considered wasteful,
+ * because it's sized to handle a worst case scenario which rarely happens.
+ * In which case, resize it down to free some memory */
+#define ZSTD_WORKSPACETOOLARGE_MAXDURATION 128
+
+/* Since the workspace is effectively its own little malloc implementation /
+ * arena, when we run under ASAN, we should similarly insert redzones between
+ * each internal element of the workspace, so ASAN will catch overruns that
+ * reach outside an object but that stay inside the workspace.
+ *
+ * This defines the size of that redzone.
+ */
+#ifndef ZSTD_CWKSP_ASAN_REDZONE_SIZE
+#define ZSTD_CWKSP_ASAN_REDZONE_SIZE 128
+#endif
+
+/*-*************************************
+*  Structures
+***************************************/
+typedef enum {
+    ZSTD_cwksp_alloc_objects,
+    ZSTD_cwksp_alloc_buffers,
+    ZSTD_cwksp_alloc_aligned
+} ZSTD_cwksp_alloc_phase_e;
+
+/**
+ * Zstd fits all its internal datastructures into a single continuous buffer,
+ * so that it only needs to perform a single OS allocation (or so that a buffer
+ * can be provided to it and it can perform no allocations at all). This buffer
+ * is called the workspace.
+ *
+ * Several optimizations complicate that process of allocating memory ranges
+ * from this workspace for each internal datastructure:
+ *
+ * - These different internal datastructures have different setup requirements:
+ *
+ *   - The static objects need to be cleared once and can then be trivially
+ *     reused for each compression.
+ *
+ *   - Various buffers don't need to be initialized at all--they are always
+ *     written into before they're read.
+ *
+ *   - The matchstate tables have a unique requirement that they don't need
+ *     their memory to be totally cleared, but they do need the memory to have
+ *     some bound, i.e., a guarantee that all values in the memory they've been
+ *     allocated is less than some maximum value (which is the starting value
+ *     for the indices that they will then use for compression). When this
+ *     guarantee is provided to them, they can use the memory without any setup
+ *     work. When it can't, they have to clear the area.
+ *
+ * - These buffers also have different alignment requirements.
+ *
+ * - We would like to reuse the objects in the workspace for multiple
+ *   compressions without having to perform any expensive reallocation or
+ *   reinitialization work.
+ *
+ * - We would like to be able to efficiently reuse the workspace across
+ *   multiple compressions **even when the compression parameters change** and
+ *   we need to resize some of the objects (where possible).
+ *
+ * To attempt to manage this buffer, given these constraints, the ZSTD_cwksp
+ * abstraction was created. It works as follows:
+ *
+ * Workspace Layout:
+ *
+ * [                        ... workspace ...                         ]
+ * [objects][tables ... ->] free space [<- ... aligned][<- ... buffers]
+ *
+ * The various objects that live in the workspace are divided into the
+ * following categories, and are allocated separately:
+ *
+ * - Static objects: this is optionally the enclosing ZSTD_CCtx or ZSTD_CDict,
+ *   so that literally everything fits in a single buffer. Note: if present,
+ *   this must be the first object in the workspace, since ZSTD_free{CCtx,
+ *   CDict}() rely on a pointer comparison to see whether one or two frees are
+ *   required.
+ *
+ * - Fixed size objects: these are fixed-size, fixed-count objects that are
+ *   nonetheless "dynamically" allocated in the workspace so that we can
+ *   control how they're initialized separately from the broader ZSTD_CCtx.
+ *   Examples:
+ *   - Entropy Workspace
+ *   - 2 x ZSTD_compressedBlockState_t
+ *   - CDict dictionary contents
+ *
+ * - Tables: these are any of several different datastructures (hash tables,
+ *   chain tables, binary trees) that all respect a common format: they are
+ *   uint32_t arrays, all of whose values are between 0 and (nextSrc - base).
+ *   Their sizes depend on the cparams.
+ *
+ * - Aligned: these buffers are used for various purposes that require 4 byte
+ *   alignment, but don't require any initialization before they're used.
+ *
+ * - Buffers: these buffers are used for various purposes that don't require
+ *   any alignment or initialization before they're used. This means they can
+ *   be moved around at no cost for a new compression.
+ *
+ * Allocating Memory:
+ *
+ * The various types of objects must be allocated in order, so they can be
+ * correctly packed into the workspace buffer. That order is:
+ *
+ * 1. Objects
+ * 2. Buffers
+ * 3. Aligned
+ * 4. Tables
+ *
+ * Attempts to reserve objects of different types out of order will fail.
+ */
+typedef struct {
+    void* workspace;
+    void* workspaceEnd;
+
+    void* objectEnd;
+    void* tableEnd;
+    void* tableValidEnd;
+    void* allocStart;
+
+    int allocFailed;
+    int workspaceOversizedDuration;
+    ZSTD_cwksp_alloc_phase_e phase;
+} ZSTD_cwksp;
+
+/*-*************************************
+*  Functions
+***************************************/
+
+MEM_STATIC size_t ZSTD_cwksp_available_space(ZSTD_cwksp* ws);
+
+MEM_STATIC void ZSTD_cwksp_assert_internal_consistency(ZSTD_cwksp* ws) {
+    (void)ws;
+    assert(ws->workspace <= ws->objectEnd);
+    assert(ws->objectEnd <= ws->tableEnd);
+    assert(ws->objectEnd <= ws->tableValidEnd);
+    assert(ws->tableEnd <= ws->allocStart);
+    assert(ws->tableValidEnd <= ws->allocStart);
+    assert(ws->allocStart <= ws->workspaceEnd);
+}
+
+/**
+ * Align must be a power of 2.
+ */
+MEM_STATIC size_t ZSTD_cwksp_align(size_t size, size_t const align) {
+    size_t const mask = align - 1;
+    assert((align & mask) == 0);
+    return (size + mask) & ~mask;
+}
+
+/**
+ * Use this to determine how much space in the workspace we will consume to
+ * allocate this object. (Normally it should be exactly the size of the object,
+ * but under special conditions, like ASAN, where we pad each object, it might
+ * be larger.)
+ *
+ * Since tables aren't currently redzoned, you don't need to call through this
+ * to figure out how much space you need for the matchState tables. Everything
+ * else is though.
+ */
+MEM_STATIC size_t ZSTD_cwksp_alloc_size(size_t size) {
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    return size + 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+#else
+    return size;
+#endif
+}
+
+MEM_STATIC void ZSTD_cwksp_internal_advance_phase(
+        ZSTD_cwksp* ws, ZSTD_cwksp_alloc_phase_e phase) {
+    assert(phase >= ws->phase);
+    if (phase > ws->phase) {
+        if (ws->phase < ZSTD_cwksp_alloc_buffers &&
+                phase >= ZSTD_cwksp_alloc_buffers) {
+            ws->tableValidEnd = ws->objectEnd;
+        }
+        if (ws->phase < ZSTD_cwksp_alloc_aligned &&
+                phase >= ZSTD_cwksp_alloc_aligned) {
+            /* If unaligned allocations down from a too-large top have left us
+             * unaligned, we need to realign our alloc ptr. Technically, this
+             * can consume space that is unaccounted for in the neededSpace
+             * calculation. However, I believe this can only happen when the
+             * workspace is too large, and specifically when it is too large
+             * by a larger margin than the space that will be consumed. */
+            /* TODO: cleaner, compiler warning friendly way to do this??? */
+            ws->allocStart = (BYTE*)ws->allocStart - ((size_t)ws->allocStart & (sizeof(U32)-1));
+            if (ws->allocStart < ws->tableValidEnd) {
+                ws->tableValidEnd = ws->allocStart;
+            }
+        }
+        ws->phase = phase;
+    }
+}
+
+/**
+ * Returns whether this object/buffer/etc was allocated in this workspace.
+ */
+MEM_STATIC int ZSTD_cwksp_owns_buffer(const ZSTD_cwksp* ws, const void* ptr) {
+    return (ptr != NULL) && (ws->workspace <= ptr) && (ptr <= ws->workspaceEnd);
+}
+
+/**
+ * Internal function. Do not use directly.
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_internal(
+        ZSTD_cwksp* ws, size_t bytes, ZSTD_cwksp_alloc_phase_e phase) {
+    void* alloc;
+    void* bottom = ws->tableEnd;
+    ZSTD_cwksp_internal_advance_phase(ws, phase);
+    alloc = (BYTE *)ws->allocStart - bytes;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    /* over-reserve space */
+    alloc = (BYTE *)alloc - 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+#endif
+
+    DEBUGLOG(5, "cwksp: reserving %p %zd bytes, %zd bytes remaining",
+        alloc, bytes, ZSTD_cwksp_available_space(ws) - bytes);
+    ZSTD_cwksp_assert_internal_consistency(ws);
+    assert(alloc >= bottom);
+    if (alloc < bottom) {
+        DEBUGLOG(4, "cwksp: alloc failed!");
+        ws->allocFailed = 1;
+        return NULL;
+    }
+    if (alloc < ws->tableValidEnd) {
+        ws->tableValidEnd = alloc;
+    }
+    ws->allocStart = alloc;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    /* Move alloc so there's ZSTD_CWKSP_ASAN_REDZONE_SIZE unused space on
+     * either size. */
+    alloc = (BYTE *)alloc + ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+    __asan_unpoison_memory_region(alloc, bytes);
+#endif
+
+    return alloc;
+}
+
+/**
+ * Reserves and returns unaligned memory.
+ */
+MEM_STATIC BYTE* ZSTD_cwksp_reserve_buffer(ZSTD_cwksp* ws, size_t bytes) {
+    return (BYTE*)ZSTD_cwksp_reserve_internal(ws, bytes, ZSTD_cwksp_alloc_buffers);
+}
+
+/**
+ * Reserves and returns memory sized on and aligned on sizeof(unsigned).
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_aligned(ZSTD_cwksp* ws, size_t bytes) {
+    assert((bytes & (sizeof(U32)-1)) == 0);
+    return ZSTD_cwksp_reserve_internal(ws, ZSTD_cwksp_align(bytes, sizeof(U32)), ZSTD_cwksp_alloc_aligned);
+}
+
+/**
+ * Aligned on sizeof(unsigned). These buffers have the special property that
+ * their values remain constrained, allowing us to re-use them without
+ * memset()-ing them.
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_table(ZSTD_cwksp* ws, size_t bytes) {
+    const ZSTD_cwksp_alloc_phase_e phase = ZSTD_cwksp_alloc_aligned;
+    void* alloc = ws->tableEnd;
+    void* end = (BYTE *)alloc + bytes;
+    void* top = ws->allocStart;
+
+    DEBUGLOG(5, "cwksp: reserving %p table %zd bytes, %zd bytes remaining",
+        alloc, bytes, ZSTD_cwksp_available_space(ws) - bytes);
+    assert((bytes & (sizeof(U32)-1)) == 0);
+    ZSTD_cwksp_internal_advance_phase(ws, phase);
+    ZSTD_cwksp_assert_internal_consistency(ws);
+    assert(end <= top);
+    if (end > top) {
+        DEBUGLOG(4, "cwksp: table alloc failed!");
+        ws->allocFailed = 1;
+        return NULL;
+    }
+    ws->tableEnd = end;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    __asan_unpoison_memory_region(alloc, bytes);
+#endif
+
+    return alloc;
+}
+
+/**
+ * Aligned on sizeof(void*).
+ */
+MEM_STATIC void* ZSTD_cwksp_reserve_object(ZSTD_cwksp* ws, size_t bytes) {
+    size_t roundedBytes = ZSTD_cwksp_align(bytes, sizeof(void*));
+    void* alloc = ws->objectEnd;
+    void* end = (BYTE*)alloc + roundedBytes;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    /* over-reserve space */
+    end = (BYTE *)end + 2 * ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+#endif
+
+    DEBUGLOG(5,
+        "cwksp: reserving %p object %zd bytes (rounded to %zd), %zd bytes remaining",
+        alloc, bytes, roundedBytes, ZSTD_cwksp_available_space(ws) - roundedBytes);
+    assert(((size_t)alloc & (sizeof(void*)-1)) == 0);
+    assert((bytes & (sizeof(void*)-1)) == 0);
+    ZSTD_cwksp_assert_internal_consistency(ws);
+    /* we must be in the first phase, no advance is possible */
+    if (ws->phase != ZSTD_cwksp_alloc_objects || end > ws->workspaceEnd) {
+        DEBUGLOG(4, "cwksp: object alloc failed!");
+        ws->allocFailed = 1;
+        return NULL;
+    }
+    ws->objectEnd = end;
+    ws->tableEnd = end;
+    ws->tableValidEnd = end;
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    /* Move alloc so there's ZSTD_CWKSP_ASAN_REDZONE_SIZE unused space on
+     * either size. */
+    alloc = (BYTE *)alloc + ZSTD_CWKSP_ASAN_REDZONE_SIZE;
+    __asan_unpoison_memory_region(alloc, bytes);
+#endif
+
+    return alloc;
+}
+
+MEM_STATIC void ZSTD_cwksp_mark_tables_dirty(ZSTD_cwksp* ws) {
+    DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_dirty");
+
+#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
+    /* To validate that the table re-use logic is sound, and that we don't
+     * access table space that we haven't cleaned, we re-"poison" the table
+     * space every time we mark it dirty. */
+    {
+        size_t size = (BYTE*)ws->tableValidEnd - (BYTE*)ws->objectEnd;
+        assert(__msan_test_shadow(ws->objectEnd, size) == -1);
+        __msan_poison(ws->objectEnd, size);
+    }
+#endif
+
+    assert(ws->tableValidEnd >= ws->objectEnd);
+    assert(ws->tableValidEnd <= ws->allocStart);
+    ws->tableValidEnd = ws->objectEnd;
+    ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+MEM_STATIC void ZSTD_cwksp_mark_tables_clean(ZSTD_cwksp* ws) {
+    DEBUGLOG(4, "cwksp: ZSTD_cwksp_mark_tables_clean");
+    assert(ws->tableValidEnd >= ws->objectEnd);
+    assert(ws->tableValidEnd <= ws->allocStart);
+    if (ws->tableValidEnd < ws->tableEnd) {
+        ws->tableValidEnd = ws->tableEnd;
+    }
+    ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+/**
+ * Zero the part of the allocated tables not already marked clean.
+ */
+MEM_STATIC void ZSTD_cwksp_clean_tables(ZSTD_cwksp* ws) {
+    DEBUGLOG(4, "cwksp: ZSTD_cwksp_clean_tables");
+    assert(ws->tableValidEnd >= ws->objectEnd);
+    assert(ws->tableValidEnd <= ws->allocStart);
+    if (ws->tableValidEnd < ws->tableEnd) {
+        memset(ws->tableValidEnd, 0, (BYTE*)ws->tableEnd - (BYTE*)ws->tableValidEnd);
+    }
+    ZSTD_cwksp_mark_tables_clean(ws);
+}
+
+/**
+ * Invalidates table allocations.
+ * All other allocations remain valid.
+ */
+MEM_STATIC void ZSTD_cwksp_clear_tables(ZSTD_cwksp* ws) {
+    DEBUGLOG(4, "cwksp: clearing tables!");
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    {
+        size_t size = (BYTE*)ws->tableValidEnd - (BYTE*)ws->objectEnd;
+        __asan_poison_memory_region(ws->objectEnd, size);
+    }
+#endif
+
+    ws->tableEnd = ws->objectEnd;
+    ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+/**
+ * Invalidates all buffer, aligned, and table allocations.
+ * Object allocations remain valid.
+ */
+MEM_STATIC void ZSTD_cwksp_clear(ZSTD_cwksp* ws) {
+    DEBUGLOG(4, "cwksp: clearing!");
+
+#if defined (MEMORY_SANITIZER) && !defined (ZSTD_MSAN_DONT_POISON_WORKSPACE)
+    /* To validate that the context re-use logic is sound, and that we don't
+     * access stuff that this compression hasn't initialized, we re-"poison"
+     * the workspace (or at least the non-static, non-table parts of it)
+     * every time we start a new compression. */
+    {
+        size_t size = (BYTE*)ws->workspaceEnd - (BYTE*)ws->tableValidEnd;
+        __msan_poison(ws->tableValidEnd, size);
+    }
+#endif
+
+#if defined (ADDRESS_SANITIZER) && !defined (ZSTD_ASAN_DONT_POISON_WORKSPACE)
+    {
+        size_t size = (BYTE*)ws->workspaceEnd - (BYTE*)ws->objectEnd;
+        __asan_poison_memory_region(ws->objectEnd, size);
+    }
+#endif
+
+    ws->tableEnd = ws->objectEnd;
+    ws->allocStart = ws->workspaceEnd;
+    ws->allocFailed = 0;
+    if (ws->phase > ZSTD_cwksp_alloc_buffers) {
+        ws->phase = ZSTD_cwksp_alloc_buffers;
+    }
+    ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+/**
+ * The provided workspace takes ownership of the buffer [start, start+size).
+ * Any existing values in the workspace are ignored (the previously managed
+ * buffer, if present, must be separately freed).
+ */
+MEM_STATIC void ZSTD_cwksp_init(ZSTD_cwksp* ws, void* start, size_t size) {
+    DEBUGLOG(4, "cwksp: init'ing workspace with %zd bytes", size);
+    assert(((size_t)start & (sizeof(void*)-1)) == 0); /* ensure correct alignment */
+    ws->workspace = start;
+    ws->workspaceEnd = (BYTE*)start + size;
+    ws->objectEnd = ws->workspace;
+    ws->tableValidEnd = ws->objectEnd;
+    ws->phase = ZSTD_cwksp_alloc_objects;
+    ZSTD_cwksp_clear(ws);
+    ws->workspaceOversizedDuration = 0;
+    ZSTD_cwksp_assert_internal_consistency(ws);
+}
+
+MEM_STATIC size_t ZSTD_cwksp_create(ZSTD_cwksp* ws, size_t size, ZSTD_customMem customMem) {
+    void* workspace = ZSTD_malloc(size, customMem);
+    DEBUGLOG(4, "cwksp: creating new workspace with %zd bytes", size);
+    RETURN_ERROR_IF(workspace == NULL, memory_allocation);
+    ZSTD_cwksp_init(ws, workspace, size);
+    return 0;
+}
+
+MEM_STATIC void ZSTD_cwksp_free(ZSTD_cwksp* ws, ZSTD_customMem customMem) {
+    void *ptr = ws->workspace;
+    DEBUGLOG(4, "cwksp: freeing workspace");
+    memset(ws, 0, sizeof(ZSTD_cwksp));
+    ZSTD_free(ptr, customMem);
+}
+
+/**
+ * Moves the management of a workspace from one cwksp to another. The src cwksp
+ * is left in an invalid state (src must be re-init()'ed before its used again).
+ */
+MEM_STATIC void ZSTD_cwksp_move(ZSTD_cwksp* dst, ZSTD_cwksp* src) {
+    *dst = *src;
+    memset(src, 0, sizeof(ZSTD_cwksp));
+}
+
+MEM_STATIC size_t ZSTD_cwksp_sizeof(const ZSTD_cwksp* ws) {
+    return (size_t)((BYTE*)ws->workspaceEnd - (BYTE*)ws->workspace);
+}
+
+MEM_STATIC int ZSTD_cwksp_reserve_failed(const ZSTD_cwksp* ws) {
+    return ws->allocFailed;
+}
+
+/*-*************************************
+*  Functions Checking Free Space
+***************************************/
+
+MEM_STATIC size_t ZSTD_cwksp_available_space(ZSTD_cwksp* ws) {
+    return (size_t)((BYTE*)ws->allocStart - (BYTE*)ws->tableEnd);
+}
+
+MEM_STATIC int ZSTD_cwksp_check_available(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+    return ZSTD_cwksp_available_space(ws) >= additionalNeededSpace;
+}
+
+MEM_STATIC int ZSTD_cwksp_check_too_large(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+    return ZSTD_cwksp_check_available(
+        ws, additionalNeededSpace * ZSTD_WORKSPACETOOLARGE_FACTOR);
+}
+
+MEM_STATIC int ZSTD_cwksp_check_wasteful(ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+    return ZSTD_cwksp_check_too_large(ws, additionalNeededSpace)
+        && ws->workspaceOversizedDuration > ZSTD_WORKSPACETOOLARGE_MAXDURATION;
+}
+
+MEM_STATIC void ZSTD_cwksp_bump_oversized_duration(
+        ZSTD_cwksp* ws, size_t additionalNeededSpace) {
+    if (ZSTD_cwksp_check_too_large(ws, additionalNeededSpace)) {
+        ws->workspaceOversizedDuration++;
+    } else {
+        ws->workspaceOversizedDuration = 0;
+    }
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+#endif /* ZSTD_CWKSP_H */
diff --git a/thirdparty/zstd/compress/zstd_double_fast.c b/thirdparty/zstd/compress/zstd_double_fast.c
index 54467cc31b..a661a48534 100644
--- a/thirdparty/zstd/compress/zstd_double_fast.c
+++ b/thirdparty/zstd/compress/zstd_double_fast.c
@@ -148,7 +148,7 @@ size_t ZSTD_compressBlock_doubleFast_generic(
             const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
             mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
             ip++;
-            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
             goto _match_stored;
         }
 
@@ -157,7 +157,7 @@ size_t ZSTD_compressBlock_doubleFast_generic(
           && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
             mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
             ip++;
-            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
             goto _match_stored;
         }
 
@@ -247,7 +247,7 @@ _match_found:
         offset_2 = offset_1;
         offset_1 = offset;
 
-        ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
 
 _match_stored:
         /* match found */
@@ -278,7 +278,7 @@ _match_stored:
                         const BYTE* const repEnd2 = repIndex2 < prefixLowestIndex ? dictEnd : iend;
                         size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixLowest) + 4;
                         U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                        ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
+                        ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
                         hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
                         hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
                         ip += repLength2;
@@ -297,7 +297,7 @@ _match_stored:
                     U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff;  /* swap offset_2 <=> offset_1 */
                     hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
                     hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
-                    ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH);
+                    ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, rLength-MINMATCH);
                     ip += rLength;
                     anchor = ip;
                     continue;   /* faster when present ... (?) */
@@ -411,7 +411,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
             const BYTE* repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
             mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
             ip++;
-            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
         } else {
             if ((matchLongIndex > dictStartIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
                 const BYTE* const matchEnd = matchLongIndex < prefixStartIndex ? dictEnd : iend;
@@ -422,7 +422,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
                 while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; }   /* catch up */
                 offset_2 = offset_1;
                 offset_1 = offset;
-                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
 
             } else if ((matchIndex > dictStartIndex) && (MEM_read32(match) == MEM_read32(ip))) {
                 size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
@@ -447,7 +447,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
                 }
                 offset_2 = offset_1;
                 offset_1 = offset;
-                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
 
             } else {
                 ip += ((ip-anchor) >> kSearchStrength) + 1;
@@ -479,7 +479,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
                     const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
                     size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
                     U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
+                    ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
                     hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
                     hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
                     ip += repLength2;
diff --git a/thirdparty/zstd/compress/zstd_fast.c b/thirdparty/zstd/compress/zstd_fast.c
index 59267ffbbc..6dbefee6b7 100644
--- a/thirdparty/zstd/compress/zstd_fast.c
+++ b/thirdparty/zstd/compress/zstd_fast.c
@@ -8,7 +8,7 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
-#include "zstd_compress_internal.h"
+#include "zstd_compress_internal.h"  /* ZSTD_hashPtr, ZSTD_count, ZSTD_storeSeq */
 #include "zstd_fast.h"
 
 
@@ -43,8 +43,8 @@ void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
 }
 
 
-FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_fast_generic(
+FORCE_INLINE_TEMPLATE size_t
+ZSTD_compressBlock_fast_generic(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize,
         U32 const mls)
@@ -74,8 +74,7 @@ size_t ZSTD_compressBlock_fast_generic(
     DEBUGLOG(5, "ZSTD_compressBlock_fast_generic");
     ip0 += (ip0 == prefixStart);
     ip1 = ip0 + 1;
-    {
-        U32 const maxRep = (U32)(ip0 - prefixStart);
+    {   U32 const maxRep = (U32)(ip0 - prefixStart);
         if (offset_2 > maxRep) offsetSaved = offset_2, offset_2 = 0;
         if (offset_1 > maxRep) offsetSaved = offset_1, offset_1 = 0;
     }
@@ -118,8 +117,7 @@ size_t ZSTD_compressBlock_fast_generic(
             match0 = match1;
             goto _offset;
         }
-        {
-            size_t const step = ((ip0-anchor) >> (kSearchStrength - 1)) + stepSize;
+        {   size_t const step = ((size_t)(ip0-anchor) >> (kSearchStrength - 1)) + stepSize;
             assert(step >= 2);
             ip0 += step;
             ip1 += step;
@@ -138,7 +136,7 @@ _offset: /* Requires: ip0, match0 */
 _match: /* Requires: ip0, match0, offcode */
         /* Count the forward length */
         mLength += ZSTD_count(ip0+mLength+4, match0+mLength+4, iend) + 4;
-        ZSTD_storeSeq(seqStore, ip0-anchor, anchor, offcode, mLength-MINMATCH);
+        ZSTD_storeSeq(seqStore, (size_t)(ip0-anchor), anchor, iend, offcode, mLength-MINMATCH);
         /* match found */
         ip0 += mLength;
         anchor = ip0;
@@ -150,16 +148,15 @@ _match: /* Requires: ip0, match0, offcode */
             hashTable[ZSTD_hashPtr(base+current0+2, hlog, mls)] = current0+2;  /* here because current+2 could be > iend-8 */
             hashTable[ZSTD_hashPtr(ip0-2, hlog, mls)] = (U32)(ip0-2-base);
 
-            while ( (ip0 <= ilimit)
-                 && ( (offset_2>0)
-                    & (MEM_read32(ip0) == MEM_read32(ip0 - offset_2)) )) {
+            while ( ((ip0 <= ilimit) & (offset_2>0))  /* offset_2==0 means offset_2 is invalidated */
+                 && (MEM_read32(ip0) == MEM_read32(ip0 - offset_2)) ) {
                 /* store sequence */
                 size_t const rLength = ZSTD_count(ip0+4, ip0+4-offset_2, iend) + 4;
-                U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff;  /* swap offset_2 <=> offset_1 */
+                { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
                 hashTable[ZSTD_hashPtr(ip0, hlog, mls)] = (U32)(ip0-base);
                 ip0 += rLength;
                 ip1 = ip0 + 1;
-                ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, 0 /*litLen*/, anchor, iend, 0 /*offCode*/, rLength-MINMATCH);
                 anchor = ip0;
                 continue;   /* faster when present (confirmed on gcc-8) ... (?) */
             }
@@ -179,8 +176,7 @@ size_t ZSTD_compressBlock_fast(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize)
 {
-    ZSTD_compressionParameters const* cParams = &ms->cParams;
-    U32 const mls = cParams->minMatch;
+    U32 const mls = ms->cParams.minMatch;
     assert(ms->dictMatchState == NULL);
     switch(mls)
     {
@@ -265,7 +261,7 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
             const BYTE* const repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
             mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixStart) + 4;
             ip++;
-            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, mLength-MINMATCH);
         } else if ( (matchIndex <= prefixStartIndex) ) {
             size_t const dictHash = ZSTD_hashPtr(ip, dictHLog, mls);
             U32 const dictMatchIndex = dictHashTable[dictHash];
@@ -285,7 +281,7 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
                 } /* catch up */
                 offset_2 = offset_1;
                 offset_1 = offset;
-                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
             }
         } else if (MEM_read32(match) != MEM_read32(ip)) {
             /* it's not a match, and we're not going to check the dictionary */
@@ -300,7 +296,7 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
                  && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
             offset_2 = offset_1;
             offset_1 = offset;
-            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
         }
 
         /* match found */
@@ -325,7 +321,7 @@ size_t ZSTD_compressBlock_fast_dictMatchState_generic(
                     const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
                     size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
                     U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
+                    ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, repLength2-MINMATCH);
                     hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
                     ip += repLength2;
                     anchor = ip;
@@ -348,8 +344,7 @@ size_t ZSTD_compressBlock_fast_dictMatchState(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize)
 {
-    ZSTD_compressionParameters const* cParams = &ms->cParams;
-    U32 const mls = cParams->minMatch;
+    U32 const mls = ms->cParams.minMatch;
     assert(ms->dictMatchState != NULL);
     switch(mls)
     {
@@ -408,16 +403,17 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
         const U32    repIndex = current + 1 - offset_1;
         const BYTE* const repBase = repIndex < prefixStartIndex ? dictBase : base;
         const BYTE* const repMatch = repBase + repIndex;
-        size_t mLength;
         hashTable[h] = current;   /* update hash table */
         assert(offset_1 <= current +1);   /* check repIndex */
 
         if ( (((U32)((prefixStartIndex-1) - repIndex) >= 3) /* intentional underflow */ & (repIndex > dictStartIndex))
            && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
             const BYTE* const repMatchEnd = repIndex < prefixStartIndex ? dictEnd : iend;
-            mLength = ZSTD_count_2segments(ip+1 +4, repMatch +4, iend, repMatchEnd, prefixStart) + 4;
+            size_t const rLength = ZSTD_count_2segments(ip+1 +4, repMatch +4, iend, repMatchEnd, prefixStart) + 4;
             ip++;
-            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, 0, rLength-MINMATCH);
+            ip += rLength;
+            anchor = ip;
         } else {
             if ( (matchIndex < dictStartIndex) ||
                  (MEM_read32(match) != MEM_read32(ip)) ) {
@@ -427,19 +423,15 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
             }
             {   const BYTE* const matchEnd = matchIndex < prefixStartIndex ? dictEnd : iend;
                 const BYTE* const lowMatchPtr = matchIndex < prefixStartIndex ? dictStart : prefixStart;
-                U32 offset;
-                mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
+                U32 const offset = current - matchIndex;
+                size_t mLength = ZSTD_count_2segments(ip+4, match+4, iend, matchEnd, prefixStart) + 4;
                 while (((ip>anchor) & (match>lowMatchPtr)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; }   /* catch up */
-                offset = current - matchIndex;
-                offset_2 = offset_1;
-                offset_1 = offset;
-                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                offset_2 = offset_1; offset_1 = offset;  /* update offset history */
+                ZSTD_storeSeq(seqStore, (size_t)(ip-anchor), anchor, iend, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ip += mLength;
+                anchor = ip;
         }   }
 
-        /* found a match : store it */
-        ip += mLength;
-        anchor = ip;
-
         if (ip <= ilimit) {
             /* Fill Table */
             hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2;
@@ -448,13 +440,13 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(
             while (ip <= ilimit) {
                 U32 const current2 = (U32)(ip-base);
                 U32 const repIndex2 = current2 - offset_2;
-                const BYTE* repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
+                const BYTE* const repMatch2 = repIndex2 < prefixStartIndex ? dictBase + repIndex2 : base + repIndex2;
                 if ( (((U32)((prefixStartIndex-1) - repIndex2) >= 3) & (repIndex2 > dictStartIndex))  /* intentional overflow */
                    && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
                     const BYTE* const repEnd2 = repIndex2 < prefixStartIndex ? dictEnd : iend;
                     size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, prefixStart) + 4;
-                    U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
+                    { U32 const tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset; }  /* swap offset_2 <=> offset_1 */
+                    ZSTD_storeSeq(seqStore, 0 /*litlen*/, anchor, iend, 0 /*offcode*/, repLength2-MINMATCH);
                     hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
                     ip += repLength2;
                     anchor = ip;
@@ -476,8 +468,7 @@ size_t ZSTD_compressBlock_fast_extDict(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize)
 {
-    ZSTD_compressionParameters const* cParams = &ms->cParams;
-    U32 const mls = cParams->minMatch;
+    U32 const mls = ms->cParams.minMatch;
     switch(mls)
     {
     default: /* includes case 3 */
diff --git a/thirdparty/zstd/compress/zstd_lazy.c b/thirdparty/zstd/compress/zstd_lazy.c
index 0af41724c7..9ad7e03b54 100644
--- a/thirdparty/zstd/compress/zstd_lazy.c
+++ b/thirdparty/zstd/compress/zstd_lazy.c
@@ -810,7 +810,7 @@ ZSTD_compressBlock_lazy_generic(
         /* store sequence */
 _storeSequence:
         {   size_t const litLength = start - anchor;
-            ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
             anchor = ip = start + matchLength;
         }
 
@@ -828,7 +828,7 @@ _storeSequence:
                     const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
                     matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
                     offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
+                    ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
                     ip += matchLength;
                     anchor = ip;
                     continue;
@@ -843,7 +843,7 @@ _storeSequence:
                 /* store sequence */
                 matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
                 offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
-                ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
                 ip += matchLength;
                 anchor = ip;
                 continue;   /* faster when present ... (?) */
@@ -1051,7 +1051,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
         /* store sequence */
 _storeSequence:
         {   size_t const litLength = start - anchor;
-            ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, litLength, anchor, iend, (U32)offset, matchLength-MINMATCH);
             anchor = ip = start + matchLength;
         }
 
@@ -1066,7 +1066,7 @@ _storeSequence:
                 const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
                 matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd, prefixStart) + 4;
                 offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset history */
-                ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, 0, anchor, iend, 0, matchLength-MINMATCH);
                 ip += matchLength;
                 anchor = ip;
                 continue;   /* faster when present ... (?) */
diff --git a/thirdparty/zstd/compress/zstd_ldm.c b/thirdparty/zstd/compress/zstd_ldm.c
index 3dcf86e6e8..c3312ad3e3 100644
--- a/thirdparty/zstd/compress/zstd_ldm.c
+++ b/thirdparty/zstd/compress/zstd_ldm.c
@@ -49,9 +49,9 @@ size_t ZSTD_ldm_getTableSize(ldmParams_t params)
 {
     size_t const ldmHSize = ((size_t)1) << params.hashLog;
     size_t const ldmBucketSizeLog = MIN(params.bucketSizeLog, params.hashLog);
-    size_t const ldmBucketSize =
-        ((size_t)1) << (params.hashLog - ldmBucketSizeLog);
-    size_t const totalSize = ldmBucketSize + ldmHSize * sizeof(ldmEntry_t);
+    size_t const ldmBucketSize = ((size_t)1) << (params.hashLog - ldmBucketSizeLog);
+    size_t const totalSize = ZSTD_cwksp_alloc_size(ldmBucketSize)
+                           + ZSTD_cwksp_alloc_size(ldmHSize * sizeof(ldmEntry_t));
     return params.enableLdm ? totalSize : 0;
 }
 
@@ -583,7 +583,7 @@ size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
                 rep[i] = rep[i-1];
             rep[0] = sequence.offset;
             /* Store the sequence */
-            ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength,
+            ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength, iend,
                           sequence.offset + ZSTD_REP_MOVE,
                           sequence.matchLength - MINMATCH);
             ip += sequence.matchLength;
diff --git a/thirdparty/zstd/compress/zstd_opt.c b/thirdparty/zstd/compress/zstd_opt.c
index 2da363f93e..2e50fca6ff 100644
--- a/thirdparty/zstd/compress/zstd_opt.c
+++ b/thirdparty/zstd/compress/zstd_opt.c
@@ -1098,7 +1098,7 @@ _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
 
                     assert(anchor + llen <= iend);
                     ZSTD_updateStats(optStatePtr, llen, anchor, offCode, mlen);
-                    ZSTD_storeSeq(seqStore, llen, anchor, offCode, mlen-MINMATCH);
+                    ZSTD_storeSeq(seqStore, llen, anchor, iend, offCode, mlen-MINMATCH);
                     anchor += advance;
                     ip = anchor;
             }   }
diff --git a/thirdparty/zstd/compress/zstdmt_compress.c b/thirdparty/zstd/compress/zstdmt_compress.c
index 9e537b8848..bc3062b530 100644
--- a/thirdparty/zstd/compress/zstdmt_compress.c
+++ b/thirdparty/zstd/compress/zstdmt_compress.c
@@ -668,7 +668,7 @@ static void ZSTDMT_compressionJob(void* jobDescription)
 
     /* init */
     if (job->cdict) {
-        size_t const initError = ZSTD_compressBegin_advanced_internal(cctx, NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast, job->cdict, jobParams, job->fullFrameSize);
+        size_t const initError = ZSTD_compressBegin_advanced_internal(cctx, NULL, 0, ZSTD_dct_auto, ZSTD_dtlm_fast, job->cdict, &jobParams, job->fullFrameSize);
         assert(job->firstJob);  /* only allowed for first job */
         if (ZSTD_isError(initError)) JOB_ERROR(initError);
     } else {  /* srcStart points at reloaded section */
@@ -680,7 +680,7 @@ static void ZSTDMT_compressionJob(void* jobDescription)
                                         job->prefix.start, job->prefix.size, ZSTD_dct_rawContent, /* load dictionary in "content-only" mode (no header analysis) */
                                         ZSTD_dtlm_fast,
                                         NULL, /*cdict*/
-                                        jobParams, pledgedSrcSize);
+                                        &jobParams, pledgedSrcSize);
             if (ZSTD_isError(initError)) JOB_ERROR(initError);
     }   }
 
@@ -927,12 +927,18 @@ static void ZSTDMT_releaseAllJobResources(ZSTDMT_CCtx* mtctx)
     unsigned jobID;
     DEBUGLOG(3, "ZSTDMT_releaseAllJobResources");
     for (jobID=0; jobID <= mtctx->jobIDMask; jobID++) {
+        /* Copy the mutex/cond out */
+        ZSTD_pthread_mutex_t const mutex = mtctx->jobs[jobID].job_mutex;
+        ZSTD_pthread_cond_t const cond = mtctx->jobs[jobID].job_cond;
+
         DEBUGLOG(4, "job%02u: release dst address %08X", jobID, (U32)(size_t)mtctx->jobs[jobID].dstBuff.start);
         ZSTDMT_releaseBuffer(mtctx->bufPool, mtctx->jobs[jobID].dstBuff);
-        mtctx->jobs[jobID].dstBuff = g_nullBuffer;
-        mtctx->jobs[jobID].cSize = 0;
+
+        /* Clear the job description, but keep the mutex/cond */
+        memset(&mtctx->jobs[jobID], 0, sizeof(mtctx->jobs[jobID]));
+        mtctx->jobs[jobID].job_mutex = mutex;
+        mtctx->jobs[jobID].job_cond = cond;
     }
-    memset(mtctx->jobs, 0, (mtctx->jobIDMask+1)*sizeof(ZSTDMT_jobDescription));
     mtctx->inBuff.buffer = g_nullBuffer;
     mtctx->inBuff.filled = 0;
     mtctx->allJobsCompleted = 1;
@@ -1028,9 +1034,9 @@ size_t ZSTDMT_getMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter,
 
 /* Sets parameters relevant to the compression job,
  * initializing others to default values. */
-static ZSTD_CCtx_params ZSTDMT_initJobCCtxParams(ZSTD_CCtx_params const params)
+static ZSTD_CCtx_params ZSTDMT_initJobCCtxParams(const ZSTD_CCtx_params* params)
 {
-    ZSTD_CCtx_params jobParams = params;
+    ZSTD_CCtx_params jobParams = *params;
     /* Clear parameters related to multithreading */
     jobParams.forceWindow = 0;
     jobParams.nbWorkers = 0;
@@ -1151,16 +1157,16 @@ size_t ZSTDMT_toFlushNow(ZSTDMT_CCtx* mtctx)
 /* =====   Multi-threaded compression   ===== */
 /* ------------------------------------------ */
 
-static unsigned ZSTDMT_computeTargetJobLog(ZSTD_CCtx_params const params)
+static unsigned ZSTDMT_computeTargetJobLog(const ZSTD_CCtx_params* params)
 {
     unsigned jobLog;
-    if (params.ldmParams.enableLdm) {
+    if (params->ldmParams.enableLdm) {
         /* In Long Range Mode, the windowLog is typically oversized.
          * In which case, it's preferable to determine the jobSize
          * based on chainLog instead. */
-        jobLog = MAX(21, params.cParams.chainLog + 4);
+        jobLog = MAX(21, params->cParams.chainLog + 4);
     } else {
-        jobLog = MAX(20, params.cParams.windowLog + 2);
+        jobLog = MAX(20, params->cParams.windowLog + 2);
     }
     return MIN(jobLog, (unsigned)ZSTDMT_JOBLOG_MAX);
 }
@@ -1193,27 +1199,27 @@ static int ZSTDMT_overlapLog(int ovlog, ZSTD_strategy strat)
     return ovlog;
 }
 
-static size_t ZSTDMT_computeOverlapSize(ZSTD_CCtx_params const params)
+static size_t ZSTDMT_computeOverlapSize(const ZSTD_CCtx_params* params)
 {
-    int const overlapRLog = 9 - ZSTDMT_overlapLog(params.overlapLog, params.cParams.strategy);
-    int ovLog = (overlapRLog >= 8) ? 0 : (params.cParams.windowLog - overlapRLog);
+    int const overlapRLog = 9 - ZSTDMT_overlapLog(params->overlapLog, params->cParams.strategy);
+    int ovLog = (overlapRLog >= 8) ? 0 : (params->cParams.windowLog - overlapRLog);
     assert(0 <= overlapRLog && overlapRLog <= 8);
-    if (params.ldmParams.enableLdm) {
+    if (params->ldmParams.enableLdm) {
         /* In Long Range Mode, the windowLog is typically oversized.
          * In which case, it's preferable to determine the jobSize
          * based on chainLog instead.
          * Then, ovLog becomes a fraction of the jobSize, rather than windowSize */
-        ovLog = MIN(params.cParams.windowLog, ZSTDMT_computeTargetJobLog(params) - 2)
+        ovLog = MIN(params->cParams.windowLog, ZSTDMT_computeTargetJobLog(params) - 2)
                 - overlapRLog;
     }
     assert(0 <= ovLog && ovLog <= ZSTD_WINDOWLOG_MAX);
-    DEBUGLOG(4, "overlapLog : %i", params.overlapLog);
+    DEBUGLOG(4, "overlapLog : %i", params->overlapLog);
     DEBUGLOG(4, "overlap size : %i", 1 << ovLog);
     return (ovLog==0) ? 0 : (size_t)1 << ovLog;
 }
 
 static unsigned
-ZSTDMT_computeNbJobs(ZSTD_CCtx_params params, size_t srcSize, unsigned nbWorkers)
+ZSTDMT_computeNbJobs(const ZSTD_CCtx_params* params, size_t srcSize, unsigned nbWorkers)
 {
     assert(nbWorkers>0);
     {   size_t const jobSizeTarget = (size_t)1 << ZSTDMT_computeTargetJobLog(params);
@@ -1236,9 +1242,9 @@ static size_t ZSTDMT_compress_advanced_internal(
           const ZSTD_CDict* cdict,
                 ZSTD_CCtx_params params)
 {
-    ZSTD_CCtx_params const jobParams = ZSTDMT_initJobCCtxParams(params);
-    size_t const overlapSize = ZSTDMT_computeOverlapSize(params);
-    unsigned const nbJobs = ZSTDMT_computeNbJobs(params, srcSize, params.nbWorkers);
+    ZSTD_CCtx_params const jobParams = ZSTDMT_initJobCCtxParams(&params);
+    size_t const overlapSize = ZSTDMT_computeOverlapSize(&params);
+    unsigned const nbJobs = ZSTDMT_computeNbJobs(&params, srcSize, params.nbWorkers);
     size_t const proposedJobSize = (srcSize + (nbJobs-1)) / nbJobs;
     size_t const avgJobSize = (((proposedJobSize-1) & 0x1FFFF) < 0x7FFF) ? proposedJobSize + 0xFFFF : proposedJobSize;   /* avoid too small last block */
     const char* const srcStart = (const char*)src;
@@ -1256,7 +1262,7 @@ static size_t ZSTDMT_compress_advanced_internal(
         ZSTD_CCtx* const cctx = mtctx->cctxPool->cctx[0];
         DEBUGLOG(4, "ZSTDMT_compress_advanced_internal: fallback to single-thread mode");
         if (cdict) return ZSTD_compress_usingCDict_advanced(cctx, dst, dstCapacity, src, srcSize, cdict, jobParams.fParams);
-        return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, NULL, 0, jobParams);
+        return ZSTD_compress_advanced_internal(cctx, dst, dstCapacity, src, srcSize, NULL, 0, &jobParams);
     }
 
     assert(avgJobSize >= 256 KB);  /* condition for ZSTD_compressBound(A) + ZSTD_compressBound(B) <= ZSTD_compressBound(A+B), required to compress directly into Dst (no additional buffer) */
@@ -1404,12 +1410,12 @@ size_t ZSTDMT_initCStream_internal(
 
     mtctx->singleBlockingThread = (pledgedSrcSize <= ZSTDMT_JOBSIZE_MIN);  /* do not trigger multi-threading when srcSize is too small */
     if (mtctx->singleBlockingThread) {
-        ZSTD_CCtx_params const singleThreadParams = ZSTDMT_initJobCCtxParams(params);
+        ZSTD_CCtx_params const singleThreadParams = ZSTDMT_initJobCCtxParams(&params);
         DEBUGLOG(5, "ZSTDMT_initCStream_internal: switch to single blocking thread mode");
         assert(singleThreadParams.nbWorkers == 0);
         return ZSTD_initCStream_internal(mtctx->cctxPool->cctx[0],
                                          dict, dictSize, cdict,
-                                         singleThreadParams, pledgedSrcSize);
+                                         &singleThreadParams, pledgedSrcSize);
     }
 
     DEBUGLOG(4, "ZSTDMT_initCStream_internal: %u workers", params.nbWorkers);
@@ -1435,11 +1441,11 @@ size_t ZSTDMT_initCStream_internal(
         mtctx->cdict = cdict;
     }
 
-    mtctx->targetPrefixSize = ZSTDMT_computeOverlapSize(params);
+    mtctx->targetPrefixSize = ZSTDMT_computeOverlapSize(&params);
     DEBUGLOG(4, "overlapLog=%i => %u KB", params.overlapLog, (U32)(mtctx->targetPrefixSize>>10));
     mtctx->targetSectionSize = params.jobSize;
     if (mtctx->targetSectionSize == 0) {
-        mtctx->targetSectionSize = 1ULL << ZSTDMT_computeTargetJobLog(params);
+        mtctx->targetSectionSize = 1ULL << ZSTDMT_computeTargetJobLog(&params);
     }
     assert(mtctx->targetSectionSize <= (size_t)ZSTDMT_JOBSIZE_MAX);
 
diff --git a/thirdparty/zstd/decompress/huf_decompress.c b/thirdparty/zstd/decompress/huf_decompress.c
index 3f8bd29732..bb2d0a96bc 100644
--- a/thirdparty/zstd/decompress/huf_decompress.c
+++ b/thirdparty/zstd/decompress/huf_decompress.c
@@ -61,7 +61,9 @@
 *  Error Management
 ****************************************************************/
 #define HUF_isError ERR_isError
+#ifndef CHECK_F
 #define CHECK_F(f) { size_t const err_ = (f); if (HUF_isError(err_)) return err_; }
+#endif
 
 
 /* **************************************************************
diff --git a/thirdparty/zstd/decompress/zstd_decompress.c b/thirdparty/zstd/decompress/zstd_decompress.c
index 751060b2cd..dd4591b7be 100644
--- a/thirdparty/zstd/decompress/zstd_decompress.c
+++ b/thirdparty/zstd/decompress/zstd_decompress.c
@@ -88,10 +88,7 @@ size_t ZSTD_estimateDCtxSize(void) { return sizeof(ZSTD_DCtx); }
 
 static size_t ZSTD_startingInputLength(ZSTD_format_e format)
 {
-    size_t const startingInputLength = (format==ZSTD_f_zstd1_magicless) ?
-                    ZSTD_FRAMEHEADERSIZE_PREFIX - ZSTD_FRAMEIDSIZE :
-                    ZSTD_FRAMEHEADERSIZE_PREFIX;
-    ZSTD_STATIC_ASSERT(ZSTD_FRAMEHEADERSIZE_PREFIX >= ZSTD_FRAMEIDSIZE);
+    size_t const startingInputLength = ZSTD_FRAMEHEADERSIZE_PREFIX(format);
     /* only supports formats ZSTD_f_zstd1 and ZSTD_f_zstd1_magicless */
     assert( (format == ZSTD_f_zstd1) || (format == ZSTD_f_zstd1_magicless) );
     return startingInputLength;
@@ -376,7 +373,7 @@ unsigned long long ZSTD_findDecompressedSize(const void* src, size_t srcSize)
 {
     unsigned long long totalDstSize = 0;
 
-    while (srcSize >= ZSTD_FRAMEHEADERSIZE_PREFIX) {
+    while (srcSize >= ZSTD_startingInputLength(ZSTD_f_zstd1)) {
         U32 const magicNumber = MEM_readLE32(src);
 
         if ((magicNumber & ZSTD_MAGIC_SKIPPABLE_MASK) == ZSTD_MAGIC_SKIPPABLE_START) {
@@ -629,11 +626,12 @@ static size_t ZSTD_decompressFrame(ZSTD_DCtx* dctx,
 
     /* check */
     RETURN_ERROR_IF(
-        remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN+ZSTD_blockHeaderSize,
+        remainingSrcSize < ZSTD_FRAMEHEADERSIZE_MIN(dctx->format)+ZSTD_blockHeaderSize,
         srcSize_wrong);
 
     /* Frame Header */
-    {   size_t const frameHeaderSize = ZSTD_frameHeaderSize(ip, ZSTD_FRAMEHEADERSIZE_PREFIX);
+    {   size_t const frameHeaderSize = ZSTD_frameHeaderSize_internal(
+                ip, ZSTD_FRAMEHEADERSIZE_PREFIX(dctx->format), dctx->format);
         if (ZSTD_isError(frameHeaderSize)) return frameHeaderSize;
         RETURN_ERROR_IF(remainingSrcSize < frameHeaderSize+ZSTD_blockHeaderSize,
                         srcSize_wrong);
@@ -714,7 +712,7 @@ static size_t ZSTD_decompressMultiFrame(ZSTD_DCtx* dctx,
         dictSize = ZSTD_DDict_dictSize(ddict);
     }
 
-    while (srcSize >= ZSTD_FRAMEHEADERSIZE_PREFIX) {
+    while (srcSize >= ZSTD_startingInputLength(dctx->format)) {
 
 #if defined(ZSTD_LEGACY_SUPPORT) && (ZSTD_LEGACY_SUPPORT >= 1)
         if (ZSTD_isLegacy(src, srcSize)) {
@@ -1098,7 +1096,7 @@ ZSTD_loadDEntropy(ZSTD_entropyDTables_t* entropy,
         size_t const dictContentSize = (size_t)(dictEnd - (dictPtr+12));
         for (i=0; i<3; i++) {
             U32 const rep = MEM_readLE32(dictPtr); dictPtr += 4;
-            RETURN_ERROR_IF(rep==0 || rep >= dictContentSize,
+            RETURN_ERROR_IF(rep==0 || rep > dictContentSize,
                             dictionary_corrupted);
             entropy->rep[i] = rep;
     }   }
@@ -1267,7 +1265,7 @@ size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx,
 {
     RETURN_ERROR_IF(dctx->streamStage != zdss_init, stage_wrong);
     ZSTD_clearDict(dctx);
-    if (dict && dictSize >= 8) {
+    if (dict && dictSize != 0) {
         dctx->ddictLocal = ZSTD_createDDict_advanced(dict, dictSize, dictLoadMethod, dictContentType, dctx->customMem);
         RETURN_ERROR_IF(dctx->ddictLocal == NULL, memory_allocation);
         dctx->ddict = dctx->ddictLocal;
@@ -1300,14 +1298,14 @@ size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSiz
 
 
 /* ZSTD_initDStream_usingDict() :
- * return : expected size, aka ZSTD_FRAMEHEADERSIZE_PREFIX.
+ * return : expected size, aka ZSTD_startingInputLength().
  * this function cannot fail */
 size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize)
 {
     DEBUGLOG(4, "ZSTD_initDStream_usingDict");
     FORWARD_IF_ERROR( ZSTD_DCtx_reset(zds, ZSTD_reset_session_only) );
     FORWARD_IF_ERROR( ZSTD_DCtx_loadDictionary(zds, dict, dictSize) );
-    return ZSTD_FRAMEHEADERSIZE_PREFIX;
+    return ZSTD_startingInputLength(zds->format);
 }
 
 /* note : this variant can't fail */
@@ -1324,16 +1322,16 @@ size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* dctx, const ZSTD_DDict* ddict)
 {
     FORWARD_IF_ERROR( ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only) );
     FORWARD_IF_ERROR( ZSTD_DCtx_refDDict(dctx, ddict) );
-    return ZSTD_FRAMEHEADERSIZE_PREFIX;
+    return ZSTD_startingInputLength(dctx->format);
 }
 
 /* ZSTD_resetDStream() :
- * return : expected size, aka ZSTD_FRAMEHEADERSIZE_PREFIX.
+ * return : expected size, aka ZSTD_startingInputLength().
  * this function cannot fail */
 size_t ZSTD_resetDStream(ZSTD_DStream* dctx)
 {
     FORWARD_IF_ERROR(ZSTD_DCtx_reset(dctx, ZSTD_reset_session_only));
-    return ZSTD_FRAMEHEADERSIZE_PREFIX;
+    return ZSTD_startingInputLength(dctx->format);
 }
 
 
@@ -1564,7 +1562,7 @@ size_t ZSTD_decompressStream(ZSTD_DStream* zds, ZSTD_outBuffer* output, ZSTD_inB
                             zds->lhSize += remainingInput;
                         }
                         input->pos = input->size;
-                        return (MAX(ZSTD_FRAMEHEADERSIZE_MIN, hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
+                        return (MAX((size_t)ZSTD_FRAMEHEADERSIZE_MIN(zds->format), hSize) - zds->lhSize) + ZSTD_blockHeaderSize;   /* remaining header bytes + next block header */
                     }
                     assert(ip != NULL);
                     memcpy(zds->headerBuffer + zds->lhSize, ip, toLoad); zds->lhSize = hSize; ip += toLoad;
diff --git a/thirdparty/zstd/decompress/zstd_decompress_block.c b/thirdparty/zstd/decompress/zstd_decompress_block.c
index cbcfc08406..767e5f9a0b 100644
--- a/thirdparty/zstd/decompress/zstd_decompress_block.c
+++ b/thirdparty/zstd/decompress/zstd_decompress_block.c
@@ -573,38 +573,118 @@ typedef struct {
     size_t pos;
 } seqState_t;
 
+/*! ZSTD_overlapCopy8() :
+ *  Copies 8 bytes from ip to op and updates op and ip where ip <= op.
+ *  If the offset is < 8 then the offset is spread to at least 8 bytes.
+ *
+ *  Precondition: *ip <= *op
+ *  Postcondition: *op - *op >= 8
+ */
+static void ZSTD_overlapCopy8(BYTE** op, BYTE const** ip, size_t offset) {
+    assert(*ip <= *op);
+    if (offset < 8) {
+        /* close range match, overlap */
+        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
+        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
+        int const sub2 = dec64table[offset];
+        (*op)[0] = (*ip)[0];
+        (*op)[1] = (*ip)[1];
+        (*op)[2] = (*ip)[2];
+        (*op)[3] = (*ip)[3];
+        *ip += dec32table[offset];
+        ZSTD_copy4(*op+4, *ip);
+        *ip -= sub2;
+    } else {
+        ZSTD_copy8(*op, *ip);
+    }
+    *ip += 8;
+    *op += 8;
+    assert(*op - *ip >= 8);
+}
+
+/*! ZSTD_safecopy() :
+ *  Specialized version of memcpy() that is allowed to READ up to WILDCOPY_OVERLENGTH past the input buffer
+ *  and write up to 16 bytes past oend_w (op >= oend_w is allowed).
+ *  This function is only called in the uncommon case where the sequence is near the end of the block. It
+ *  should be fast for a single long sequence, but can be slow for several short sequences.
+ *
+ *  @param ovtype controls the overlap detection
+ *         - ZSTD_no_overlap: The source and destination are guaranteed to be at least WILDCOPY_VECLEN bytes apart.
+ *         - ZSTD_overlap_src_before_dst: The src and dst may overlap and may be any distance apart.
+ *           The src buffer must be before the dst buffer.
+ */
+static void ZSTD_safecopy(BYTE* op, BYTE* const oend_w, BYTE const* ip, ptrdiff_t length, ZSTD_overlap_e ovtype) {
+    ptrdiff_t const diff = op - ip;
+    BYTE* const oend = op + length;
 
-/* ZSTD_execSequenceLast7():
- * exceptional case : decompress a match starting within last 7 bytes of output buffer.
- * requires more careful checks, to ensure there is no overflow.
- * performance does not matter though.
- * note : this case is supposed to be never generated "naturally" by reference encoder,
- *        since in most cases it needs at least 8 bytes to look for a match.
- *        but it's allowed by the specification. */
+    assert((ovtype == ZSTD_no_overlap && (diff <= -8 || diff >= 8 || op >= oend_w)) ||
+           (ovtype == ZSTD_overlap_src_before_dst && diff >= 0));
+
+    if (length < 8) {
+        /* Handle short lengths. */
+        while (op < oend) *op++ = *ip++;
+        return;
+    }
+    if (ovtype == ZSTD_overlap_src_before_dst) {
+        /* Copy 8 bytes and ensure the offset >= 8 when there can be overlap. */
+        assert(length >= 8);
+        ZSTD_overlapCopy8(&op, &ip, diff);
+        assert(op - ip >= 8);
+        assert(op <= oend);
+    }
+
+    if (oend <= oend_w) {
+        /* No risk of overwrite. */
+        ZSTD_wildcopy(op, ip, length, ovtype);
+        return;
+    }
+    if (op <= oend_w) {
+        /* Wildcopy until we get close to the end. */
+        assert(oend > oend_w);
+        ZSTD_wildcopy(op, ip, oend_w - op, ovtype);
+        ip += oend_w - op;
+        op = oend_w;
+    }
+    /* Handle the leftovers. */
+    while (op < oend) *op++ = *ip++;
+}
+
+/* ZSTD_execSequenceEnd():
+ * This version handles cases that are near the end of the output buffer. It requires
+ * more careful checks to make sure there is no overflow. By separating out these hard
+ * and unlikely cases, we can speed up the common cases.
+ *
+ * NOTE: This function needs to be fast for a single long sequence, but doesn't need
+ * to be optimized for many small sequences, since those fall into ZSTD_execSequence().
+ */
 FORCE_NOINLINE
-size_t ZSTD_execSequenceLast7(BYTE* op,
-                              BYTE* const oend, seq_t sequence,
-                              const BYTE** litPtr, const BYTE* const litLimit,
-                              const BYTE* const base, const BYTE* const vBase, const BYTE* const dictEnd)
+size_t ZSTD_execSequenceEnd(BYTE* op,
+                            BYTE* const oend, seq_t sequence,
+                            const BYTE** litPtr, const BYTE* const litLimit,
+                            const BYTE* const prefixStart, const BYTE* const virtualStart, const BYTE* const dictEnd)
 {
     BYTE* const oLitEnd = op + sequence.litLength;
     size_t const sequenceLength = sequence.litLength + sequence.matchLength;
     BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
     const BYTE* const iLitEnd = *litPtr + sequence.litLength;
     const BYTE* match = oLitEnd - sequence.offset;
+    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
 
-    /* check */
-    RETURN_ERROR_IF(oMatchEnd>oend, dstSize_tooSmall, "last match must fit within dstBuffer");
+    /* bounds checks */
+    assert(oLitEnd < oMatchEnd);
+    RETURN_ERROR_IF(oMatchEnd > oend, dstSize_tooSmall, "last match must fit within dstBuffer");
     RETURN_ERROR_IF(iLitEnd > litLimit, corruption_detected, "try to read beyond literal buffer");
 
     /* copy literals */
-    while (op < oLitEnd) *op++ = *(*litPtr)++;
+    ZSTD_safecopy(op, oend_w, *litPtr, sequence.litLength, ZSTD_no_overlap);
+    op = oLitEnd;
+    *litPtr = iLitEnd;
 
     /* copy Match */
-    if (sequence.offset > (size_t)(oLitEnd - base)) {
+    if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
         /* offset beyond prefix */
-        RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - vBase),corruption_detected);
-        match = dictEnd - (base-match);
+        RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected);
+        match = dictEnd - (prefixStart-match);
         if (match + sequence.matchLength <= dictEnd) {
             memmove(oLitEnd, match, sequence.matchLength);
             return sequenceLength;
@@ -614,13 +694,12 @@ size_t ZSTD_execSequenceLast7(BYTE* op,
             memmove(oLitEnd, match, length1);
             op = oLitEnd + length1;
             sequence.matchLength -= length1;
-            match = base;
+            match = prefixStart;
     }   }
-    while (op < oMatchEnd) *op++ = *match++;
+    ZSTD_safecopy(op, oend_w, match, sequence.matchLength, ZSTD_overlap_src_before_dst);
     return sequenceLength;
 }
 
-
 HINT_INLINE
 size_t ZSTD_execSequence(BYTE* op,
                          BYTE* const oend, seq_t sequence,
@@ -634,20 +713,29 @@ size_t ZSTD_execSequence(BYTE* op,
     const BYTE* const iLitEnd = *litPtr + sequence.litLength;
     const BYTE* match = oLitEnd - sequence.offset;
 
-    /* check */
-    RETURN_ERROR_IF(oMatchEnd>oend, dstSize_tooSmall, "last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend");
-    RETURN_ERROR_IF(iLitEnd > litLimit, corruption_detected, "over-read beyond lit buffer");
-    if (oLitEnd>oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
-
-    /* copy Literals */
-    if (sequence.litLength > 8)
-        ZSTD_wildcopy_16min(op, (*litPtr), sequence.litLength, ZSTD_no_overlap);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
-    else
-        ZSTD_copy8(op, *litPtr);
+    /* Errors and uncommon cases handled here. */
+    assert(oLitEnd < oMatchEnd);
+    if (iLitEnd > litLimit || oMatchEnd > oend_w)
+        return ZSTD_execSequenceEnd(op, oend, sequence, litPtr, litLimit, prefixStart, virtualStart, dictEnd);
+
+    /* Assumptions (everything else goes into ZSTD_execSequenceEnd()) */
+    assert(iLitEnd <= litLimit /* Literal length is in bounds */);
+    assert(oLitEnd <= oend_w /* Can wildcopy literals */);
+    assert(oMatchEnd <= oend_w /* Can wildcopy matches */);
+
+    /* Copy Literals:
+     * Split out litLength <= 16 since it is nearly always true. +1.6% on gcc-9.
+     * We likely don't need the full 32-byte wildcopy.
+     */
+    assert(WILDCOPY_OVERLENGTH >= 16);
+    ZSTD_copy16(op, (*litPtr));
+    if (sequence.litLength > 16) {
+        ZSTD_wildcopy(op+16, (*litPtr)+16, sequence.litLength-16, ZSTD_no_overlap);
+    }
     op = oLitEnd;
     *litPtr = iLitEnd;   /* update for next sequence */
 
-    /* copy Match */
+    /* Copy Match */
     if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
         /* offset beyond prefix -> go into extDict */
         RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - virtualStart), corruption_detected);
@@ -662,123 +750,33 @@ size_t ZSTD_execSequence(BYTE* op,
             op = oLitEnd + length1;
             sequence.matchLength -= length1;
             match = prefixStart;
-            if (op > oend_w || sequence.matchLength < MINMATCH) {
-              U32 i;
-              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
-              return sequenceLength;
-            }
     }   }
-    /* Requirement: op <= oend_w && sequence.matchLength >= MINMATCH */
-
-    /* match within prefix */
-    if (sequence.offset < 8) {
-        /* close range match, overlap */
-        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
-        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
-        int const sub2 = dec64table[sequence.offset];
-        op[0] = match[0];
-        op[1] = match[1];
-        op[2] = match[2];
-        op[3] = match[3];
-        match += dec32table[sequence.offset];
-        ZSTD_copy4(op+4, match);
-        match -= sub2;
-    } else {
-        ZSTD_copy8(op, match);
-    }
-    op += 8; match += 8;
-
-    if (oMatchEnd > oend-(16-MINMATCH)) {
-        if (op < oend_w) {
-            ZSTD_wildcopy(op, match, oend_w - op, ZSTD_overlap_src_before_dst);
-            match += oend_w - op;
-            op = oend_w;
-        }
-        while (op < oMatchEnd) *op++ = *match++;
-    } else {
-        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8, ZSTD_overlap_src_before_dst);   /* works even if matchLength < 8 */
+    /* Match within prefix of 1 or more bytes */
+    assert(op <= oMatchEnd);
+    assert(oMatchEnd <= oend_w);
+    assert(match >= prefixStart);
+    assert(sequence.matchLength >= 1);
+
+    /* Nearly all offsets are >= WILDCOPY_VECLEN bytes, which means we can use wildcopy
+     * without overlap checking.
+     */
+    if (sequence.offset >= WILDCOPY_VECLEN) {
+        /* We bet on a full wildcopy for matches, since we expect matches to be
+         * longer than literals (in general). In silesia, ~10% of matches are longer
+         * than 16 bytes.
+         */
+        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength, ZSTD_no_overlap);
+        return sequenceLength;
     }
-    return sequenceLength;
-}
-
-
-HINT_INLINE
-size_t ZSTD_execSequenceLong(BYTE* op,
-                             BYTE* const oend, seq_t sequence,
-                             const BYTE** litPtr, const BYTE* const litLimit,
-                             const BYTE* const prefixStart, const BYTE* const dictStart, const BYTE* const dictEnd)
-{
-    BYTE* const oLitEnd = op + sequence.litLength;
-    size_t const sequenceLength = sequence.litLength + sequence.matchLength;
-    BYTE* const oMatchEnd = op + sequenceLength;   /* risk : address space overflow (32-bits) */
-    BYTE* const oend_w = oend - WILDCOPY_OVERLENGTH;
-    const BYTE* const iLitEnd = *litPtr + sequence.litLength;
-    const BYTE* match = sequence.match;
+    assert(sequence.offset < WILDCOPY_VECLEN);
 
-    /* check */
-    RETURN_ERROR_IF(oMatchEnd > oend, dstSize_tooSmall, "last match must start at a minimum distance of WILDCOPY_OVERLENGTH from oend");
-    RETURN_ERROR_IF(iLitEnd > litLimit, corruption_detected, "over-read beyond lit buffer");
-    if (oLitEnd > oend_w) return ZSTD_execSequenceLast7(op, oend, sequence, litPtr, litLimit, prefixStart, dictStart, dictEnd);
-
-    /* copy Literals */
-    if (sequence.litLength > 8)
-        ZSTD_wildcopy_16min(op, *litPtr, sequence.litLength, ZSTD_no_overlap);   /* note : since oLitEnd <= oend-WILDCOPY_OVERLENGTH, no risk of overwrite beyond oend */
-    else
-        ZSTD_copy8(op, *litPtr);  /* note : op <= oLitEnd <= oend_w == oend - 8 */
+    /* Copy 8 bytes and spread the offset to be >= 8. */
+    ZSTD_overlapCopy8(&op, &match, sequence.offset);
 
-    op = oLitEnd;
-    *litPtr = iLitEnd;   /* update for next sequence */
-
-    /* copy Match */
-    if (sequence.offset > (size_t)(oLitEnd - prefixStart)) {
-        /* offset beyond prefix */
-        RETURN_ERROR_IF(sequence.offset > (size_t)(oLitEnd - dictStart), corruption_detected);
-        if (match + sequence.matchLength <= dictEnd) {
-            memmove(oLitEnd, match, sequence.matchLength);
-            return sequenceLength;
-        }
-        /* span extDict & currentPrefixSegment */
-        {   size_t const length1 = dictEnd - match;
-            memmove(oLitEnd, match, length1);
-            op = oLitEnd + length1;
-            sequence.matchLength -= length1;
-            match = prefixStart;
-            if (op > oend_w || sequence.matchLength < MINMATCH) {
-              U32 i;
-              for (i = 0; i < sequence.matchLength; ++i) op[i] = match[i];
-              return sequenceLength;
-            }
-    }   }
-    assert(op <= oend_w);
-    assert(sequence.matchLength >= MINMATCH);
-
-    /* match within prefix */
-    if (sequence.offset < 8) {
-        /* close range match, overlap */
-        static const U32 dec32table[] = { 0, 1, 2, 1, 4, 4, 4, 4 };   /* added */
-        static const int dec64table[] = { 8, 8, 8, 7, 8, 9,10,11 };   /* subtracted */
-        int const sub2 = dec64table[sequence.offset];
-        op[0] = match[0];
-        op[1] = match[1];
-        op[2] = match[2];
-        op[3] = match[3];
-        match += dec32table[sequence.offset];
-        ZSTD_copy4(op+4, match);
-        match -= sub2;
-    } else {
-        ZSTD_copy8(op, match);
-    }
-    op += 8; match += 8;
-
-    if (oMatchEnd > oend-(16-MINMATCH)) {
-        if (op < oend_w) {
-            ZSTD_wildcopy(op, match, oend_w - op, ZSTD_overlap_src_before_dst);
-            match += oend_w - op;
-            op = oend_w;
-        }
-        while (op < oMatchEnd) *op++ = *match++;
-    } else {
-        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8, ZSTD_overlap_src_before_dst);   /* works even if matchLength < 8 */
+    /* If the match length is > 8 bytes, then continue with the wildcopy. */
+    if (sequence.matchLength > 8) {
+        assert(op < oMatchEnd);
+        ZSTD_wildcopy(op, match, (ptrdiff_t)sequence.matchLength-8, ZSTD_overlap_src_before_dst);
     }
     return sequenceLength;
 }
@@ -1098,7 +1096,7 @@ ZSTD_decompressSequencesLong_body(
         /* decode and decompress */
         for ( ; (BIT_reloadDStream(&(seqState.DStream)) <= BIT_DStream_completed) && (seqNb<nbSeq) ; seqNb++) {
             seq_t const sequence = ZSTD_decodeSequenceLong(&seqState, isLongOffset);
-            size_t const oneSeqSize = ZSTD_execSequenceLong(op, oend, sequences[(seqNb-ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd);
+            size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequences[(seqNb-ADVANCED_SEQS) & STORED_SEQS_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd);
             if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
             PREFETCH_L1(sequence.match); PREFETCH_L1(sequence.match + sequence.matchLength - 1); /* note : it's safe to invoke PREFETCH() on any memory address, including invalid ones */
             sequences[seqNb & STORED_SEQS_MASK] = sequence;
@@ -1109,7 +1107,7 @@ ZSTD_decompressSequencesLong_body(
         /* finish queue */
         seqNb -= seqAdvance;
         for ( ; seqNb<nbSeq ; seqNb++) {
-            size_t const oneSeqSize = ZSTD_execSequenceLong(op, oend, sequences[seqNb&STORED_SEQS_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd);
+            size_t const oneSeqSize = ZSTD_execSequence(op, oend, sequences[seqNb&STORED_SEQS_MASK], &litPtr, litEnd, prefixStart, dictStart, dictEnd);
             if (ZSTD_isError(oneSeqSize)) return oneSeqSize;
             op += oneSeqSize;
         }
diff --git a/thirdparty/zstd/zstd.h b/thirdparty/zstd/zstd.h
index f8e95f2283..72080ea87e 100644
--- a/thirdparty/zstd/zstd.h
+++ b/thirdparty/zstd/zstd.h
@@ -15,6 +15,7 @@ extern "C" {
 #define ZSTD_H_235446
 
 /* ======   Dependency   ======*/
+#include <limits.h>   /* INT_MAX */
 #include <stddef.h>   /* size_t */
 
 
@@ -71,7 +72,7 @@ extern "C" {
 /*------   Version   ------*/
 #define ZSTD_VERSION_MAJOR    1
 #define ZSTD_VERSION_MINOR    4
-#define ZSTD_VERSION_RELEASE  3
+#define ZSTD_VERSION_RELEASE  4
 
 #define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
 ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< to check runtime library version */
@@ -196,9 +197,13 @@ ZSTDLIB_API ZSTD_CCtx* ZSTD_createCCtx(void);
 ZSTDLIB_API size_t     ZSTD_freeCCtx(ZSTD_CCtx* cctx);
 
 /*! ZSTD_compressCCtx() :
- *  Same as ZSTD_compress(), using an explicit ZSTD_CCtx
- *  The function will compress at requested compression level,
- *  ignoring any other parameter */
+ *  Same as ZSTD_compress(), using an explicit ZSTD_CCtx.
+ *  Important : in order to behave similarly to `ZSTD_compress()`,
+ *  this function compresses at requested compression level,
+ *  __ignoring any other parameter__ .
+ *  If any advanced parameter was set using the advanced API,
+ *  they will all be reset. Only `compressionLevel` remains.
+ */
 ZSTDLIB_API size_t ZSTD_compressCCtx(ZSTD_CCtx* cctx,
                                      void* dst, size_t dstCapacity,
                                const void* src, size_t srcSize,
@@ -233,7 +238,7 @@ ZSTDLIB_API size_t ZSTD_decompressDCtx(ZSTD_DCtx* dctx,
  *   using ZSTD_CCtx_set*() functions.
  *   Pushed parameters are sticky : they are valid for next compressed frame, and any subsequent frame.
  *   "sticky" parameters are applicable to `ZSTD_compress2()` and `ZSTD_compressStream*()` !
- *   They do not apply to "simple" one-shot variants such as ZSTD_compressCCtx()
+ *   __They do not apply to "simple" one-shot variants such as ZSTD_compressCCtx()__ .
  *
  *   It's possible to reset all parameters to "default" using ZSTD_CCtx_reset().
  *
@@ -261,18 +266,26 @@ typedef enum {
 
     /* compression parameters
      * Note: When compressing with a ZSTD_CDict these parameters are superseded
-     * by the parameters used to construct the ZSTD_CDict. See ZSTD_CCtx_refCDict()
-     * for more info (superseded-by-cdict). */
-    ZSTD_c_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
+     * by the parameters used to construct the ZSTD_CDict.
+     * See ZSTD_CCtx_refCDict() for more info (superseded-by-cdict). */
+    ZSTD_c_compressionLevel=100, /* Set compression parameters according to pre-defined cLevel table.
+                              * Note that exact compression parameters are dynamically determined,
+                              * depending on both compression level and srcSize (when known).
                               * Default level is ZSTD_CLEVEL_DEFAULT==3.
                               * Special: value 0 means default, which is controlled by ZSTD_CLEVEL_DEFAULT.
                               * Note 1 : it's possible to pass a negative compression level.
-                              * Note 2 : setting a level sets all default values of other compression parameters */
+                              * Note 2 : setting a level resets all other compression parameters to default */
+    /* Advanced compression parameters :
+     * It's possible to pin down compression parameters to some specific values.
+     * In which case, these values are no longer dynamically selected by the compressor */
     ZSTD_c_windowLog=101,    /* Maximum allowed back-reference distance, expressed as power of 2.
+                              * This will set a memory budget for streaming decompression,
+                              * with larger values requiring more memory
+                              * and typically compressing more.
                               * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
                               * Special: value 0 means "use default windowLog".
                               * Note: Using a windowLog greater than ZSTD_WINDOWLOG_LIMIT_DEFAULT
-                              *       requires explicitly allowing such window size at decompression stage if using streaming. */
+                              *       requires explicitly allowing such size at streaming decompression stage. */
     ZSTD_c_hashLog=102,      /* Size of the initial probe table, as a power of 2.
                               * Resulting memory usage is (1 << (hashLog+2)).
                               * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
@@ -283,13 +296,13 @@ typedef enum {
                               * Resulting memory usage is (1 << (chainLog+2)).
                               * Must be clamped between ZSTD_CHAINLOG_MIN and ZSTD_CHAINLOG_MAX.
                               * Larger tables result in better and slower compression.
-                              * This parameter is useless when using "fast" strategy.
+                              * This parameter is useless for "fast" strategy.
                               * It's still useful when using "dfast" strategy,
                               * in which case it defines a secondary probe table.
                               * Special: value 0 means "use default chainLog". */
     ZSTD_c_searchLog=104,    /* Number of search attempts, as a power of 2.
                               * More attempts result in better and slower compression.
-                              * This parameter is useless when using "fast" and "dFast" strategies.
+                              * This parameter is useless for "fast" and "dFast" strategies.
                               * Special: value 0 means "use default searchLog". */
     ZSTD_c_minMatch=105,     /* Minimum size of searched matches.
                               * Note that Zstandard can still find matches of smaller size,
@@ -344,7 +357,7 @@ typedef enum {
     ZSTD_c_contentSizeFlag=200, /* Content size will be written into frame header _whenever known_ (default:1)
                               * Content size must be known at the beginning of compression.
                               * This is automatically the case when using ZSTD_compress2(),
-                              * For streaming variants, content size must be provided with ZSTD_CCtx_setPledgedSrcSize() */
+                              * For streaming scenarios, content size must be provided with ZSTD_CCtx_setPledgedSrcSize() */
     ZSTD_c_checksumFlag=201, /* A 32-bits checksum of content is written at end of frame (default:0) */
     ZSTD_c_dictIDFlag=202,   /* When applicable, dictionary's ID is written into frame header (default:1) */
 
@@ -363,7 +376,7 @@ typedef enum {
                               * Each compression job is completed in parallel, so this value can indirectly impact the nb of active threads.
                               * 0 means default, which is dynamically determined based on compression parameters.
                               * Job size must be a minimum of overlap size, or 1 MB, whichever is largest.
-                              * The minimum size is automatically and transparently enforced */
+                              * The minimum size is automatically and transparently enforced. */
     ZSTD_c_overlapLog=402,   /* Control the overlap size, as a fraction of window size.
                               * The overlap size is an amount of data reloaded from previous job at the beginning of a new job.
                               * It helps preserve compression ratio, while each job is compressed in parallel.
@@ -386,6 +399,7 @@ typedef enum {
      * ZSTD_c_forceAttachDict
      * ZSTD_c_literalCompressionMode
      * ZSTD_c_targetCBlockSize
+     * ZSTD_c_srcSizeHint
      * Because they are not stable, it's necessary to define ZSTD_STATIC_LINKING_ONLY to access them.
      * note : never ever use experimentalParam? names directly;
      *        also, the enums values themselves are unstable and can still change.
@@ -396,6 +410,7 @@ typedef enum {
      ZSTD_c_experimentalParam4=1001,
      ZSTD_c_experimentalParam5=1002,
      ZSTD_c_experimentalParam6=1003,
+     ZSTD_c_experimentalParam7=1004
 } ZSTD_cParameter;
 
 typedef struct {
@@ -793,12 +808,17 @@ ZSTDLIB_API size_t ZSTD_decompress_usingDict(ZSTD_DCtx* dctx,
 typedef struct ZSTD_CDict_s ZSTD_CDict;
 
 /*! ZSTD_createCDict() :
- *  When compressing multiple messages / blocks using the same dictionary, it's recommended to load it only once.
- *  ZSTD_createCDict() will create a digested dictionary, ready to start future compression operations without startup cost.
+ *  When compressing multiple messages or blocks using the same dictionary,
+ *  it's recommended to digest the dictionary only once, since it's a costly operation.
+ *  ZSTD_createCDict() will create a state from digesting a dictionary.
+ *  The resulting state can be used for future compression operations with very limited startup cost.
  *  ZSTD_CDict can be created once and shared by multiple threads concurrently, since its usage is read-only.
- * `dictBuffer` can be released after ZSTD_CDict creation, because its content is copied within CDict.
- *  Consider experimental function `ZSTD_createCDict_byReference()` if you prefer to not duplicate `dictBuffer` content.
- *  Note : A ZSTD_CDict can be created from an empty dictBuffer, but it is inefficient when used to compress small data. */
+ * @dictBuffer can be released after ZSTD_CDict creation, because its content is copied within CDict.
+ *  Note 1 : Consider experimental function `ZSTD_createCDict_byReference()` if you prefer to not duplicate @dictBuffer content.
+ *  Note 2 : A ZSTD_CDict can be created from an empty @dictBuffer,
+ *      in which case the only thing that it transports is the @compressionLevel.
+ *      This can be useful in a pipeline featuring ZSTD_compress_usingCDict() exclusively,
+ *      expecting a ZSTD_CDict parameter with any data, including those without a known dictionary. */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict(const void* dictBuffer, size_t dictSize,
                                          int compressionLevel);
 
@@ -925,7 +945,7 @@ ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
  *  Note 3 : Referencing a prefix involves building tables, which are dependent on compression parameters.
  *           It's a CPU consuming operation, with non-negligible impact on latency.
  *           If there is a need to use the same prefix multiple times, consider loadDictionary instead.
- *  Note 4 : By default, the prefix is interpreted as raw content (ZSTD_dm_rawContent).
+ *  Note 4 : By default, the prefix is interpreted as raw content (ZSTD_dct_rawContent).
  *           Use experimental ZSTD_CCtx_refPrefix_advanced() to alter dictionary interpretation. */
 ZSTDLIB_API size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx,
                                  const void* prefix, size_t prefixSize);
@@ -969,7 +989,7 @@ ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
  *  Note 2 : Prefix buffer is referenced. It **must** outlive decompression.
  *           Prefix buffer must remain unmodified up to the end of frame,
  *           reached when ZSTD_decompressStream() returns 0.
- *  Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
+ *  Note 3 : By default, the prefix is treated as raw content (ZSTD_dct_rawContent).
  *           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode (Experimental section)
  *  Note 4 : Referencing a raw content prefix has almost no cpu nor memory cost.
  *           A full dictionary is more costly, as it requires building tables.
@@ -1014,8 +1034,8 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
  * Some of them might be removed in the future (especially when redundant with existing stable functions)
  * ***************************************************************************************/
 
-#define ZSTD_FRAMEHEADERSIZE_PREFIX 5   /* minimum input size required to query frame header size */
-#define ZSTD_FRAMEHEADERSIZE_MIN    6
+#define ZSTD_FRAMEHEADERSIZE_PREFIX(format) ((format) == ZSTD_f_zstd1 ? 5 : 1)   /* minimum input size required to query frame header size */
+#define ZSTD_FRAMEHEADERSIZE_MIN(format)    ((format) == ZSTD_f_zstd1 ? 6 : 2)
 #define ZSTD_FRAMEHEADERSIZE_MAX   18   /* can be useful for static allocation */
 #define ZSTD_SKIPPABLEHEADERSIZE    8
 
@@ -1063,6 +1083,8 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
 /* Advanced parameter bounds */
 #define ZSTD_TARGETCBLOCKSIZE_MIN   64
 #define ZSTD_TARGETCBLOCKSIZE_MAX   ZSTD_BLOCKSIZE_MAX
+#define ZSTD_SRCSIZEHINT_MIN        0
+#define ZSTD_SRCSIZEHINT_MAX        INT_MAX
 
 /* internal */
 #define ZSTD_HASHLOG3_MAX           17
@@ -1073,6 +1095,24 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
 typedef struct ZSTD_CCtx_params_s ZSTD_CCtx_params;
 
 typedef struct {
+    unsigned int matchPos; /* Match pos in dst */
+    /* If seqDef.offset > 3, then this is seqDef.offset - 3
+     * If seqDef.offset < 3, then this is the corresponding repeat offset
+     * But if seqDef.offset < 3 and litLength == 0, this is the
+     *   repeat offset before the corresponding repeat offset
+     * And if seqDef.offset == 3 and litLength == 0, this is the
+     *   most recent repeat offset - 1
+     */
+    unsigned int offset;
+    unsigned int litLength; /* Literal length */
+    unsigned int matchLength; /* Match length */
+    /* 0 when seq not rep and seqDef.offset otherwise
+     * when litLength == 0 this will be <= 4, otherwise <= 3 like normal
+     */
+    unsigned int rep;
+} ZSTD_Sequence;
+
+typedef struct {
     unsigned windowLog;       /**< largest match distance : larger == more compression, more memory needed during decompression */
     unsigned chainLog;        /**< fully searched segment : larger == more compression, slower, more memory (useless for fast) */
     unsigned hashLog;         /**< dispatch table : larger == faster, more memory */
@@ -1101,21 +1141,12 @@ typedef enum {
 
 typedef enum {
     ZSTD_dlm_byCopy = 0,  /**< Copy dictionary content internally */
-    ZSTD_dlm_byRef = 1,   /**< Reference dictionary content -- the dictionary buffer must outlive its users. */
+    ZSTD_dlm_byRef = 1    /**< Reference dictionary content -- the dictionary buffer must outlive its users. */
 } ZSTD_dictLoadMethod_e;
 
 typedef enum {
-    /* Opened question : should we have a format ZSTD_f_auto ?
-     * Today, it would mean exactly the same as ZSTD_f_zstd1.
-     * But, in the future, should several formats become supported,
-     * on the compression side, it would mean "default format".
-     * On the decompression side, it would mean "automatic format detection",
-     * so that ZSTD_f_zstd1 would mean "accept *only* zstd frames".
-     * Since meaning is a little different, another option could be to define different enums for compression and decompression.
-     * This question could be kept for later, when there are actually multiple formats to support,
-     * but there is also the question of pinning enum values, and pinning value `0` is especially important */
     ZSTD_f_zstd1 = 0,           /* zstd frame format, specified in zstd_compression_format.md (default) */
-    ZSTD_f_zstd1_magicless = 1, /* Variant of zstd frame format, without initial 4-bytes magic number.
+    ZSTD_f_zstd1_magicless = 1  /* Variant of zstd frame format, without initial 4-bytes magic number.
                                  * Useful to save 4 bytes per generated frame.
                                  * Decoder cannot recognise automatically this format, requiring this instruction. */
 } ZSTD_format_e;
@@ -1126,7 +1157,7 @@ typedef enum {
      * to evolve and should be considered only in the context of extremely
      * advanced performance tuning.
      *
-     * Zstd currently supports the use of a CDict in two ways:
+     * Zstd currently supports the use of a CDict in three ways:
      *
      * - The contents of the CDict can be copied into the working context. This
      *   means that the compression can search both the dictionary and input
@@ -1142,6 +1173,12 @@ typedef enum {
      *   working context's tables can be reused). For small inputs, this can be
      *   faster than copying the CDict's tables.
      *
+     * - The CDict's tables are not used at all, and instead we use the working
+     *   context alone to reload the dictionary and use params based on the source
+     *   size. See ZSTD_compress_insertDictionary() and ZSTD_compress_usingDict().
+     *   This method is effective when the dictionary sizes are very small relative
+     *   to the input size, and the input size is fairly large to begin with.
+     *
      * Zstd has a simple internal heuristic that selects which strategy to use
      * at the beginning of a compression. However, if experimentation shows that
      * Zstd is making poor choices, it is possible to override that choice with
@@ -1150,6 +1187,7 @@ typedef enum {
     ZSTD_dictDefaultAttach = 0, /* Use the default heuristic. */
     ZSTD_dictForceAttach   = 1, /* Never copy the dictionary. */
     ZSTD_dictForceCopy     = 2, /* Always copy the dictionary. */
+    ZSTD_dictForceLoad     = 3  /* Always reload the dictionary */
 } ZSTD_dictAttachPref_e;
 
 typedef enum {
@@ -1158,7 +1196,7 @@ typedef enum {
                                *   levels will be compressed. */
   ZSTD_lcm_huffman = 1,       /**< Always attempt Huffman compression. Uncompressed literals will still be
                                *   emitted if Huffman compression is not profitable. */
-  ZSTD_lcm_uncompressed = 2,  /**< Always emit uncompressed literals. */
+  ZSTD_lcm_uncompressed = 2   /**< Always emit uncompressed literals. */
 } ZSTD_literalCompressionMode_e;
 
 
@@ -1210,20 +1248,38 @@ ZSTDLIB_API unsigned long long ZSTD_decompressBound(const void* src, size_t srcS
  *           or an error code (if srcSize is too small) */
 ZSTDLIB_API size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
 
+/*! ZSTD_getSequences() :
+ * Extract sequences from the sequence store
+ * zc can be used to insert custom compression params.
+ * This function invokes ZSTD_compress2
+ * @return : number of sequences extracted
+ */
+ZSTDLIB_API size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
+    size_t outSeqsSize, const void* src, size_t srcSize);
+
 
 /***************************************
 *  Memory management
 ***************************************/
 
 /*! ZSTD_estimate*() :
- *  These functions make it possible to estimate memory usage
- *  of a future {D,C}Ctx, before its creation.
- *  ZSTD_estimateCCtxSize() will provide a budget large enough for any compression level up to selected one.
- *  It will also consider src size to be arbitrarily "large", which is worst case.
- *  If srcSize is known to always be small, ZSTD_estimateCCtxSize_usingCParams() can provide a tighter estimation.
- *  ZSTD_estimateCCtxSize_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
- *  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParams_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_c_nbWorkers is >= 1.
- *  Note : CCtx size estimation is only correct for single-threaded compression. */
+ *  These functions make it possible to estimate memory usage of a future
+ *  {D,C}Ctx, before its creation.
+ *
+ *  ZSTD_estimateCCtxSize() will provide a budget large enough for any
+ *  compression level up to selected one. Unlike ZSTD_estimateCStreamSize*(),
+ *  this estimate does not include space for a window buffer, so this estimate
+ *  is guaranteed to be enough for single-shot compressions, but not streaming
+ *  compressions. It will however assume the input may be arbitrarily large,
+ *  which is the worst case. If srcSize is known to always be small,
+ *  ZSTD_estimateCCtxSize_usingCParams() can provide a tighter estimation.
+ *  ZSTD_estimateCCtxSize_usingCParams() can be used in tandem with
+ *  ZSTD_getCParams() to create cParams from compressionLevel.
+ *  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with
+ *  ZSTD_CCtxParams_setParameter().
+ *
+ *  Note: only single-threaded compression is supported. This function will
+ *  return an error code if ZSTD_c_nbWorkers is >= 1. */
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize(int compressionLevel);
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize_usingCParams(ZSTD_compressionParameters cParams);
 ZSTDLIB_API size_t ZSTD_estimateCCtxSize_usingCCtxParams(const ZSTD_CCtx_params* params);
@@ -1334,7 +1390,8 @@ ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictS
  *  Create a digested dictionary for compression
  *  Dictionary content is just referenced, not duplicated.
  *  As a consequence, `dictBuffer` **must** outlive CDict,
- *  and its content must remain unmodified throughout the lifetime of CDict. */
+ *  and its content must remain unmodified throughout the lifetime of CDict.
+ *  note: equivalent to ZSTD_createCDict_advanced(), with dictLoadMethod==ZSTD_dlm_byRef */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
 
 /*! ZSTD_getCParams() :
@@ -1361,7 +1418,9 @@ ZSTDLIB_API size_t ZSTD_checkCParams(ZSTD_compressionParameters params);
 ZSTDLIB_API ZSTD_compressionParameters ZSTD_adjustCParams(ZSTD_compressionParameters cPar, unsigned long long srcSize, size_t dictSize);
 
 /*! ZSTD_compress_advanced() :
- *  Same as ZSTD_compress_usingDict(), with fine-tune control over compression parameters (by structure) */
+ *  Note : this function is now DEPRECATED.
+ *         It can be replaced by ZSTD_compress2(), in combination with ZSTD_CCtx_setParameter() and other parameter setters.
+ *  This prototype will be marked as deprecated and generate compilation warning on reaching v1.5.x */
 ZSTDLIB_API size_t ZSTD_compress_advanced(ZSTD_CCtx* cctx,
                                           void* dst, size_t dstCapacity,
                                     const void* src, size_t srcSize,
@@ -1369,7 +1428,9 @@ ZSTDLIB_API size_t ZSTD_compress_advanced(ZSTD_CCtx* cctx,
                                           ZSTD_parameters params);
 
 /*! ZSTD_compress_usingCDict_advanced() :
- *  Same as ZSTD_compress_usingCDict(), with fine-tune control over frame parameters */
+ *  Note : this function is now REDUNDANT.
+ *         It can be replaced by ZSTD_compress2(), in combination with ZSTD_CCtx_loadDictionary() and other parameter setters.
+ *  This prototype will be marked as deprecated and generate compilation warning in some future version */
 ZSTDLIB_API size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
                                               void* dst, size_t dstCapacity,
                                         const void* src, size_t srcSize,
@@ -1441,6 +1502,12 @@ ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* pre
  * There is no guarantee on compressed block size (default:0) */
 #define ZSTD_c_targetCBlockSize ZSTD_c_experimentalParam6
 
+/* User's best guess of source size.
+ * Hint is not valid when srcSizeHint == 0.
+ * There is no guarantee that hint is close to actual source size,
+ * but compression ratio may regress significantly if guess considerably underestimates */
+#define ZSTD_c_srcSizeHint ZSTD_c_experimentalParam7
+
 /*! ZSTD_CCtx_getParameter() :
  *  Get the requested compression parameter value, selected by enum ZSTD_cParameter,
  *  and store it into int* value.
@@ -1613,8 +1680,13 @@ ZSTDLIB_API size_t ZSTD_decompressStream_simpleArgs (
  * pledgedSrcSize must be correct. If it is not known at init time, use
  * ZSTD_CONTENTSIZE_UNKNOWN. Note that, for compatibility with older programs,
  * "0" also disables frame content size field. It may be enabled in the future.
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
-ZSTDLIB_API size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize);
+ZSTDLIB_API size_t
+ZSTD_initCStream_srcSize(ZSTD_CStream* zcs,
+                         int compressionLevel,
+                         unsigned long long pledgedSrcSize);
+
 /**! ZSTD_initCStream_usingDict() :
  * This function is deprecated, and is equivalent to:
  *     ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
@@ -1623,42 +1695,66 @@ ZSTDLIB_API size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLe
  *
  * Creates of an internal CDict (incompatible with static CCtx), except if
  * dict == NULL or dictSize < 8, in which case no dict is used.
- * Note: dict is loaded with ZSTD_dm_auto (treated as a full zstd dictionary if
+ * Note: dict is loaded with ZSTD_dct_auto (treated as a full zstd dictionary if
  * it begins with ZSTD_MAGIC_DICTIONARY, else as raw content) and ZSTD_dlm_byCopy.
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
-ZSTDLIB_API size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel);
+ZSTDLIB_API size_t
+ZSTD_initCStream_usingDict(ZSTD_CStream* zcs,
+                     const void* dict, size_t dictSize,
+                           int compressionLevel);
+
 /**! ZSTD_initCStream_advanced() :
  * This function is deprecated, and is approximately equivalent to:
  *     ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
- *     ZSTD_CCtx_setZstdParams(zcs, params); // Set the zstd params and leave the rest as-is
+ *     // Pseudocode: Set each zstd parameter and leave the rest as-is.
+ *     for ((param, value) : params) {
+ *         ZSTD_CCtx_setParameter(zcs, param, value);
+ *     }
  *     ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize);
  *     ZSTD_CCtx_loadDictionary(zcs, dict, dictSize);
  *
- * pledgedSrcSize must be correct. If srcSize is not known at init time, use
- * value ZSTD_CONTENTSIZE_UNKNOWN. dict is loaded with ZSTD_dm_auto and ZSTD_dlm_byCopy.
+ * dict is loaded with ZSTD_dct_auto and ZSTD_dlm_byCopy.
+ * pledgedSrcSize must be correct.
+ * If srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN.
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
-ZSTDLIB_API size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs, const void* dict, size_t dictSize,
-                                             ZSTD_parameters params, unsigned long long pledgedSrcSize);
+ZSTDLIB_API size_t
+ZSTD_initCStream_advanced(ZSTD_CStream* zcs,
+                    const void* dict, size_t dictSize,
+                          ZSTD_parameters params,
+                          unsigned long long pledgedSrcSize);
+
 /**! ZSTD_initCStream_usingCDict() :
  * This function is deprecated, and equivalent to:
  *     ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
  *     ZSTD_CCtx_refCDict(zcs, cdict);
  *
  * note : cdict will just be referenced, and must outlive compression session
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
 ZSTDLIB_API size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict);
+
 /**! ZSTD_initCStream_usingCDict_advanced() :
- * This function is deprecated, and is approximately equivalent to:
+ *   This function is DEPRECATED, and is approximately equivalent to:
  *     ZSTD_CCtx_reset(zcs, ZSTD_reset_session_only);
- *     ZSTD_CCtx_setZstdFrameParams(zcs, fParams); // Set the zstd frame params and leave the rest as-is
+ *     // Pseudocode: Set each zstd frame parameter and leave the rest as-is.
+ *     for ((fParam, value) : fParams) {
+ *         ZSTD_CCtx_setParameter(zcs, fParam, value);
+ *     }
  *     ZSTD_CCtx_setPledgedSrcSize(zcs, pledgedSrcSize);
  *     ZSTD_CCtx_refCDict(zcs, cdict);
  *
  * same as ZSTD_initCStream_usingCDict(), with control over frame parameters.
  * pledgedSrcSize must be correct. If srcSize is not known at init time, use
  * value ZSTD_CONTENTSIZE_UNKNOWN.
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
-ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize);
+ZSTDLIB_API size_t
+ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs,
+                               const ZSTD_CDict* cdict,
+                                     ZSTD_frameParameters fParams,
+                                     unsigned long long pledgedSrcSize);
 
 /*! ZSTD_resetCStream() :
  * This function is deprecated, and is equivalent to:
@@ -1673,6 +1769,7 @@ ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const
  *  For the time being, pledgedSrcSize==0 is interpreted as "srcSize unknown" for compatibility with older programs,
  *  but it will change to mean "empty" in future version, so use macro ZSTD_CONTENTSIZE_UNKNOWN instead.
  * @return : 0, or an error code (which can be tested using ZSTD_isError())
+ *  Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
 ZSTDLIB_API size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize);
 
@@ -1718,8 +1815,10 @@ ZSTDLIB_API size_t ZSTD_toFlushNow(ZSTD_CCtx* cctx);
  *     ZSTD_DCtx_loadDictionary(zds, dict, dictSize);
  *
  * note: no dictionary will be used if dict == NULL or dictSize < 8
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
 ZSTDLIB_API size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize);
+
 /**
  * This function is deprecated, and is equivalent to:
  *
@@ -1727,14 +1826,17 @@ ZSTDLIB_API size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dic
  *     ZSTD_DCtx_refDDict(zds, ddict);
  *
  * note : ddict is referenced, it must outlive decompression session
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
 ZSTDLIB_API size_t ZSTD_initDStream_usingDDict(ZSTD_DStream* zds, const ZSTD_DDict* ddict);
+
 /**
  * This function is deprecated, and is equivalent to:
  *
  *     ZSTD_DCtx_reset(zds, ZSTD_reset_session_only);
  *
  * re-use decompression parameters from previous init; saves dictionary loading
+ * Note : this prototype will be marked as deprecated and generate compilation warnings on reaching v1.5.x
  */
 ZSTDLIB_API size_t ZSTD_resetDStream(ZSTD_DStream* zds);
 
@@ -1908,7 +2010,7 @@ ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
 
 /*!
     Block functions produce and decode raw zstd blocks, without frame metadata.
-    Frame metadata cost is typically ~18 bytes, which can be non-negligible for very small blocks (< 100 bytes).
+    Frame metadata cost is typically ~12 bytes, which can be non-negligible for very small blocks (< 100 bytes).
     But users will have to take in charge needed metadata to regenerate data, such as compressed and content sizes.
 
     A few rules to respect :