Merge pull request #3768 from ThomasWaldmann/zstd-update

zstd update

setup.py

@@ -20,7 +20,7 @@ import setup_b2
 prefer_system_liblz4 = True
 
 # True: use the shared libzstd (>= 1.3.0) from the system, False: use the bundled zstd code
-prefer_system_libzstd = True
+prefer_system_libzstd = False
 
 # True: use the shared libb2 from the system, False: use the bundled blake2 code
 prefer_system_libb2 = True

src/borg/algorithms/zstd/lib/common/bitstream.h

@@ -167,7 +167,7 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
 /*-**************************************************************
 *  Internal functions
 ****************************************************************/
-MEM_STATIC unsigned BIT_highbit32 (register U32 val)
+MEM_STATIC unsigned BIT_highbit32 (U32 val)
 {
     assert(val != 0);
     {
@@ -426,7 +426,7 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, U32 nbBits)
  *  Refill `bitD` from buffer previously set in BIT_initDStream() .
  *  This function is safe, it guarantees it will not read beyond src buffer.
  * @return : status of `BIT_DStream_t` internal register.
- *           when status == BIT_DStream_unfinished, internal register is filled with at least 25 or 57 bits */
+ *           when status == BIT_DStream_unfinished, internal register is filled with at least 25 or 57 bits */
 MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD)
 {
     if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* overflow detected, like end of stream */

src/borg/algorithms/zstd/lib/common/compiler.h

@@ -63,6 +63,31 @@
 #  endif
 #endif
 
+/* target attribute */
+#ifndef __has_attribute
+  #define __has_attribute(x) 0  /* Compatibility with non-clang compilers. */
+#endif
+#if defined(__GNUC__)
+#  define TARGET_ATTRIBUTE(target) __attribute__((__target__(target)))
+#else
+#  define TARGET_ATTRIBUTE(target)
+#endif
+
+/* Enable runtime BMI2 dispatch based on the CPU.
+ * Enabled for clang & gcc >=4.8 on x86 when BMI2 isn't enabled by default.
+ */
+#ifndef DYNAMIC_BMI2
+  #if (defined(__clang__) && __has_attribute(__target__)) \
+      || (defined(__GNUC__) \
+          && (__GNUC__ >= 5 || (__GNUC__ == 4 && __GNUC_MINOR__ >= 8))) \
+      && (defined(__x86_64__) || defined(_M_X86)) \
+      && !defined(__BMI2__)
+  #  define DYNAMIC_BMI2 1
+  #else
+  #  define DYNAMIC_BMI2 0
+  #endif
+#endif
+
 /* prefetch */
 #if defined(_MSC_VER) && (defined(_M_X64) || defined(_M_I86))  /* _mm_prefetch() is not defined outside of x86/x64 */
 #  include <mmintrin.h>   /* https://msdn.microsoft.com/fr-fr/library/84szxsww(v=vs.90).aspx */

src/borg/algorithms/zstd/lib/common/cpu.h

@@ -0,0 +1,216 @@
+/*
+ * Copyright (c) 2018-present, 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_COMMON_CPU_H
+#define ZSTD_COMMON_CPU_H
+
+/**
+ * Implementation taken from folly/CpuId.h
+ * https://github.com/facebook/folly/blob/master/folly/CpuId.h
+ */
+
+#include <string.h>
+
+#include "mem.h"
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#endif
+
+typedef struct {
+    U32 f1c;
+    U32 f1d;
+    U32 f7b;
+    U32 f7c;
+} ZSTD_cpuid_t;
+
+MEM_STATIC ZSTD_cpuid_t ZSTD_cpuid(void) {
+    U32 f1c = 0;
+    U32 f1d = 0;
+    U32 f7b = 0;
+    U32 f7c = 0;
+#ifdef _MSC_VER
+    int reg[4];
+    __cpuid((int*)reg, 0);
+    {
+        int const n = reg[0];
+        if (n >= 1) {
+            __cpuid((int*)reg, 1);
+            f1c = (U32)reg[2];
+            f1d = (U32)reg[3];
+        }
+        if (n >= 7) {
+            __cpuidex((int*)reg, 7, 0);
+            f7b = (U32)reg[1];
+            f7c = (U32)reg[2];
+        }
+    }
+#elif defined(__i386__) && defined(__PIC__) && !defined(__clang__) && defined(__GNUC__)
+    /* The following block like the normal cpuid branch below, but gcc
+     * reserves ebx for use of its pic register so we must specially
+     * handle the save and restore to avoid clobbering the register
+     */
+    U32 n;
+    __asm__(
+        "pushl %%ebx\n\t"
+        "cpuid\n\t"
+        "popl %%ebx\n\t"
+        : "=a"(n)
+        : "a"(0)
+        : "ecx", "edx");
+    if (n >= 1) {
+      U32 f1a;
+      __asm__(
+          "pushl %%ebx\n\t"
+          "cpuid\n\t"
+          "popl %%ebx\n\t"
+          : "=a"(f1a), "=c"(f1c), "=d"(f1d)
+          : "a"(1)
+          :);
+    }
+    if (n >= 7) {
+      __asm__(
+          "pushl %%ebx\n\t"
+          "cpuid\n\t"
+          "movl %%ebx, %%eax\n\r"
+          "popl %%ebx"
+          : "=a"(f7b), "=c"(f7c)
+          : "a"(7), "c"(0)
+          : "edx");
+    }
+#elif defined(__x86_64__) || defined(_M_X64) || defined(__i386__)
+    U32 n;
+    __asm__("cpuid" : "=a"(n) : "a"(0) : "ebx", "ecx", "edx");
+    if (n >= 1) {
+      U32 f1a;
+      __asm__("cpuid" : "=a"(f1a), "=c"(f1c), "=d"(f1d) : "a"(1) : "ebx");
+    }
+    if (n >= 7) {
+      U32 f7a;
+      __asm__("cpuid"
+              : "=a"(f7a), "=b"(f7b), "=c"(f7c)
+              : "a"(7), "c"(0)
+              : "edx");
+    }
+#endif
+    {
+        ZSTD_cpuid_t cpuid;
+        cpuid.f1c = f1c;
+        cpuid.f1d = f1d;
+        cpuid.f7b = f7b;
+        cpuid.f7c = f7c;
+        return cpuid;
+    }
+}
+
+#define X(name, r, bit)                                                        \
+  MEM_STATIC int ZSTD_cpuid_##name(ZSTD_cpuid_t const cpuid) {                 \
+    return ((cpuid.r) & (1U << bit)) != 0;                                     \
+  }
+
+/* cpuid(1): Processor Info and Feature Bits. */
+#define C(name, bit) X(name, f1c, bit)
+  C(sse3, 0)
+  C(pclmuldq, 1)
+  C(dtes64, 2)
+  C(monitor, 3)
+  C(dscpl, 4)
+  C(vmx, 5)
+  C(smx, 6)
+  C(eist, 7)
+  C(tm2, 8)
+  C(ssse3, 9)
+  C(cnxtid, 10)
+  C(fma, 12)
+  C(cx16, 13)
+  C(xtpr, 14)
+  C(pdcm, 15)
+  C(pcid, 17)
+  C(dca, 18)
+  C(sse41, 19)
+  C(sse42, 20)
+  C(x2apic, 21)
+  C(movbe, 22)
+  C(popcnt, 23)
+  C(tscdeadline, 24)
+  C(aes, 25)
+  C(xsave, 26)
+  C(osxsave, 27)
+  C(avx, 28)
+  C(f16c, 29)
+  C(rdrand, 30)
+#undef C
+#define D(name, bit) X(name, f1d, bit)
+  D(fpu, 0)
+  D(vme, 1)
+  D(de, 2)
+  D(pse, 3)
+  D(tsc, 4)
+  D(msr, 5)
+  D(pae, 6)
+  D(mce, 7)
+  D(cx8, 8)
+  D(apic, 9)
+  D(sep, 11)
+  D(mtrr, 12)
+  D(pge, 13)
+  D(mca, 14)
+  D(cmov, 15)
+  D(pat, 16)
+  D(pse36, 17)
+  D(psn, 18)
+  D(clfsh, 19)
+  D(ds, 21)
+  D(acpi, 22)
+  D(mmx, 23)
+  D(fxsr, 24)
+  D(sse, 25)
+  D(sse2, 26)
+  D(ss, 27)
+  D(htt, 28)
+  D(tm, 29)
+  D(pbe, 31)
+#undef D
+
+/* cpuid(7): Extended Features. */
+#define B(name, bit) X(name, f7b, bit)
+  B(bmi1, 3)
+  B(hle, 4)
+  B(avx2, 5)
+  B(smep, 7)
+  B(bmi2, 8)
+  B(erms, 9)
+  B(invpcid, 10)
+  B(rtm, 11)
+  B(mpx, 14)
+  B(avx512f, 16)
+  B(avx512dq, 17)
+  B(rdseed, 18)
+  B(adx, 19)
+  B(smap, 20)
+  B(avx512ifma, 21)
+  B(pcommit, 22)
+  B(clflushopt, 23)
+  B(clwb, 24)
+  B(avx512pf, 26)
+  B(avx512er, 27)
+  B(avx512cd, 28)
+  B(sha, 29)
+  B(avx512bw, 30)
+  B(avx512vl, 31)
+#undef B
+#define C(name, bit) X(name, f7c, bit)
+  C(prefetchwt1, 0)
+  C(avx512vbmi, 1)
+#undef C
+
+#undef X
+
+#endif /* ZSTD_COMMON_CPU_H */

src/borg/algorithms/zstd/lib/common/error_private.c

@@ -29,6 +29,7 @@ const char* ERR_getErrorString(ERR_enum code)
     case PREFIX(parameter_outOfBound): return "Parameter is out of bound";
     case PREFIX(init_missing): return "Context should be init first";
     case PREFIX(memory_allocation): return "Allocation error : not enough memory";
+    case PREFIX(workSpace_tooSmall): return "workSpace buffer is not large enough";
     case PREFIX(stage_wrong): return "Operation not authorized at current processing stage";
     case PREFIX(tableLog_tooLarge): return "tableLog requires too much memory : unsupported";
     case PREFIX(maxSymbolValue_tooLarge): return "Unsupported max Symbol Value : too large";

src/borg/algorithms/zstd/lib/common/fse.h

@@ -345,7 +345,7 @@ size_t FSE_countFast(unsigned* count, unsigned* maxSymbolValuePtr, const void* s
  */
 size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize, unsigned* workSpace);
 
-/*! FSE_count_simple
+/*! FSE_count_simple() :
  * Same as FSE_countFast(), but does not use any additional memory (not even on stack).
  * This function is unsafe, and will segfault if any value within `src` is `> *maxSymbolValuePtr` (presuming it's also the size of `count`).
 */

src/borg/algorithms/zstd/lib/common/fse_decompress.c

@@ -139,8 +139,8 @@ size_t FSE_buildDTable(FSE_DTable* dt, const short* normalizedCounter, unsigned
     {   U32 u;
         for (u=0; u<tableSize; u++) {
             FSE_FUNCTION_TYPE const symbol = (FSE_FUNCTION_TYPE)(tableDecode[u].symbol);
-            U16 nextState = symbolNext[symbol]++;
-            tableDecode[u].nbBits = (BYTE) (tableLog - BIT_highbit32 ((U32)nextState) );
+            U32 const nextState = symbolNext[symbol]++;
+            tableDecode[u].nbBits = (BYTE) (tableLog - BIT_highbit32(nextState) );
             tableDecode[u].newState = (U16) ( (nextState << tableDecode[u].nbBits) - tableSize);
     }   }
 

src/borg/algorithms/zstd/lib/common/huf.h

@@ -58,32 +58,32 @@ extern "C" {
 #endif
 
 
-/* *** simple functions *** */
-/**
-HUF_compress() :
-    Compress content from buffer 'src', of size 'srcSize', into buffer 'dst'.
-    'dst' buffer must be already allocated.
-    Compression runs faster if `dstCapacity` >= HUF_compressBound(srcSize).
-    `srcSize` must be <= `HUF_BLOCKSIZE_MAX` == 128 KB.
-    @return : size of compressed data (<= `dstCapacity`).
-    Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!!
-                     if return == 1, srcData is a single repeated byte symbol (RLE compression).
-                     if HUF_isError(return), compression failed (more details using HUF_getErrorName())
-*/
+/* ========================== */
+/* ***  simple functions  *** */
+/* ========================== */
+
+/** HUF_compress() :
+ *  Compress content from buffer 'src', of size 'srcSize', into buffer 'dst'.
+ * 'dst' buffer must be already allocated.
+ *  Compression runs faster if `dstCapacity` >= HUF_compressBound(srcSize).
+ * `srcSize` must be <= `HUF_BLOCKSIZE_MAX` == 128 KB.
+ * @return : size of compressed data (<= `dstCapacity`).
+ *  Special values : if return == 0, srcData is not compressible => Nothing is stored within dst !!!
+ *                   if HUF_isError(return), compression failed (more details using HUF_getErrorName())
+ */
 HUF_PUBLIC_API size_t HUF_compress(void* dst, size_t dstCapacity,
                              const void* src, size_t srcSize);
 
-/**
-HUF_decompress() :
-    Decompress HUF data from buffer 'cSrc', of size 'cSrcSize',
-    into already allocated buffer 'dst', of minimum size 'dstSize'.
-    `originalSize` : **must** be the ***exact*** size of original (uncompressed) data.
-    Note : in contrast with FSE, HUF_decompress can regenerate
-           RLE (cSrcSize==1) and uncompressed (cSrcSize==dstSize) data,
-           because it knows size to regenerate.
-    @return : size of regenerated data (== originalSize),
-              or an error code, which can be tested using HUF_isError()
-*/
+/** HUF_decompress() :
+ *  Decompress HUF data from buffer 'cSrc', of size 'cSrcSize',
+ *  into already allocated buffer 'dst', of minimum size 'dstSize'.
+ * `originalSize` : **must** be the ***exact*** size of original (uncompressed) data.
+ *  Note : in contrast with FSE, HUF_decompress can regenerate
+ *         RLE (cSrcSize==1) and uncompressed (cSrcSize==dstSize) data,
+ *         because it knows size to regenerate (originalSize).
+ * @return : size of regenerated data (== originalSize),
+ *           or an error code, which can be tested using HUF_isError()
+ */
 HUF_PUBLIC_API size_t HUF_decompress(void* dst,  size_t originalSize,
                                const void* cSrc, size_t cSrcSize);
 
@@ -100,39 +100,32 @@ HUF_PUBLIC_API const char* HUF_getErrorName(size_t code);  /**< provides error c
 /* ***   Advanced function   *** */
 
 /** HUF_compress2() :
- *  Same as HUF_compress(), but offers direct control over `maxSymbolValue` and `tableLog`.
- *  `tableLog` must be `<= HUF_TABLELOG_MAX` . */
-HUF_PUBLIC_API size_t HUF_compress2 (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
+ *  Same as HUF_compress(), but offers control over `maxSymbolValue` and `tableLog`.
+ * `maxSymbolValue` must be <= HUF_SYMBOLVALUE_MAX .
+ * `tableLog` must be `<= HUF_TABLELOG_MAX` . */
+HUF_PUBLIC_API size_t HUF_compress2 (void* dst, size_t dstCapacity,
+                               const void* src, size_t srcSize,
+                               unsigned maxSymbolValue, unsigned tableLog);
 
 /** HUF_compress4X_wksp() :
  *  Same as HUF_compress2(), but uses externally allocated `workSpace`.
- *  `workspace` must have minimum alignment of 4, and be at least as large as following macro */
+ * `workspace` must have minimum alignment of 4, and be at least as large as HUF_WORKSPACE_SIZE */
 #define HUF_WORKSPACE_SIZE (6 << 10)
 #define HUF_WORKSPACE_SIZE_U32 (HUF_WORKSPACE_SIZE / sizeof(U32))
-HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);
-
-/**
- *  The minimum workspace size for the `workSpace` used in
- *  HUF_readDTableX2_wksp() and HUF_readDTableX4_wksp().
- *
- *  The space used depends on HUF_TABLELOG_MAX, ranging from ~1500 bytes when
- *  HUF_TABLE_LOG_MAX=12 to ~1850 bytes when HUF_TABLE_LOG_MAX=15.
- *  Buffer overflow errors may potentially occur if code modifications result in
- *  a required workspace size greater than that specified in the following
- *  macro.
- */
-#define HUF_DECOMPRESS_WORKSPACE_SIZE (2 << 10)
-#define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32))
+HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity,
+                                     const void* src, size_t srcSize,
+                                     unsigned maxSymbolValue, unsigned tableLog,
+                                     void* workSpace, size_t wkspSize);
 
 #endif   /* HUF_H_298734234 */
 
 /* ******************************************************************
  *  WARNING !!
  *  The following section contains advanced and experimental definitions
- *  which shall never be used in the context of dll
+ *  which shall never be used in the context of a dynamic library,
  *  because they are not guaranteed to remain stable in the future.
  *  Only consider them in association with static linking.
- *******************************************************************/
+ * *****************************************************************/
 #if defined(HUF_STATIC_LINKING_ONLY) && !defined(HUF_H_HUF_STATIC_LINKING_ONLY)
 #define HUF_H_HUF_STATIC_LINKING_ONLY
 
@@ -141,11 +134,11 @@ HUF_PUBLIC_API size_t HUF_compress4X_wksp (void* dst, size_t dstCapacity, const
 
 
 /* *** Constants *** */
-#define HUF_TABLELOG_MAX      12       /* max configured tableLog (for static allocation); can be modified up to HUF_ABSOLUTEMAX_TABLELOG */
-#define HUF_TABLELOG_DEFAULT  11       /* tableLog by default, when not specified */
+#define HUF_TABLELOG_MAX      12      /* max runtime value of tableLog (due to static allocation); can be modified up to HUF_ABSOLUTEMAX_TABLELOG */
+#define HUF_TABLELOG_DEFAULT  11      /* default tableLog value when none specified */
 #define HUF_SYMBOLVALUE_MAX  255
 
-#define HUF_TABLELOG_ABSOLUTEMAX  15   /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
+#define HUF_TABLELOG_ABSOLUTEMAX  15  /* absolute limit of HUF_MAX_TABLELOG. Beyond that value, code does not work */
 #if (HUF_TABLELOG_MAX > HUF_TABLELOG_ABSOLUTEMAX)
 #  error "HUF_TABLELOG_MAX is too large !"
 #endif
@@ -192,24 +185,23 @@ size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
 
 
 /* ****************************************
-*  HUF detailed API
-******************************************/
-/*!
-HUF_compress() does the following:
-1. count symbol occurrence from source[] into table count[] using FSE_count()
-2. (optional) refine tableLog using HUF_optimalTableLog()
-3. build Huffman table from count using HUF_buildCTable()
-4. save Huffman table to memory buffer using HUF_writeCTable()
-5. encode the data stream using HUF_compress4X_usingCTable()
-
-The following API allows targeting specific sub-functions for advanced tasks.
-For example, it's possible to compress several blocks using the same 'CTable',
-or to save and regenerate 'CTable' using external methods.
-*/
-/* FSE_count() : find it within "fse.h" */
+ *  HUF detailed API
+ * ****************************************/
+
+/*! HUF_compress() does the following:
+ *  1. count symbol occurrence from source[] into table count[] using FSE_count() (exposed within "fse.h")
+ *  2. (optional) refine tableLog using HUF_optimalTableLog()
+ *  3. build Huffman table from count using HUF_buildCTable()
+ *  4. save Huffman table to memory buffer using HUF_writeCTable()
+ *  5. encode the data stream using HUF_compress4X_usingCTable()
+ *
+ *  The following API allows targeting specific sub-functions for advanced tasks.
+ *  For example, it's possible to compress several blocks using the same 'CTable',
+ *  or to save and regenerate 'CTable' using external methods.
+ */
 unsigned HUF_optimalTableLog(unsigned maxTableLog, size_t srcSize, unsigned maxSymbolValue);
 typedef struct HUF_CElt_s HUF_CElt;   /* incomplete type */
-size_t HUF_buildCTable (HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits);
+size_t HUF_buildCTable (HUF_CElt* CTable, const unsigned* count, unsigned maxSymbolValue, unsigned maxNbBits);   /* @return : maxNbBits; CTable and count can overlap. In which case, CTable will overwrite count content */
 size_t HUF_writeCTable (void* dst, size_t maxDstSize, const HUF_CElt* CTable, unsigned maxSymbolValue, unsigned huffLog);
 size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
 
@@ -219,46 +211,65 @@ typedef enum {
    HUF_repeat_valid  /**< Can use the previous table and it is asumed to be valid */
  } HUF_repeat;
 /** HUF_compress4X_repeat() :
-*   Same as HUF_compress4X_wksp(), but considers using hufTable if *repeat != HUF_repeat_none.
-*   If it uses hufTable it does not modify hufTable or repeat.
-*   If it doesn't, it sets *repeat = HUF_repeat_none, and it sets hufTable to the table used.
-*   If preferRepeat then the old table will always be used if valid. */
-size_t HUF_compress4X_repeat(void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize, HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat);  /**< `workSpace` must be a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
+ *  Same as HUF_compress4X_wksp(), but considers using hufTable if *repeat != HUF_repeat_none.
+ *  If it uses hufTable it does not modify hufTable or repeat.
+ *  If it doesn't, it sets *repeat = HUF_repeat_none, and it sets hufTable to the table used.
+ *  If preferRepeat then the old table will always be used if valid. */
+size_t HUF_compress4X_repeat(void* dst, size_t dstSize,
+                       const void* src, size_t srcSize,
+                       unsigned maxSymbolValue, unsigned tableLog,
+                       void* workSpace, size_t wkspSize,    /**< `workSpace` must be aligned on 4-bytes boundaries, `wkspSize` must be >= HUF_WORKSPACE_SIZE */
+                       HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2);
 
 /** HUF_buildCTable_wksp() :
  *  Same as HUF_buildCTable(), but using externally allocated scratch buffer.
- *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of 1024 unsigned.
+ * `workSpace` must be aligned on 4-bytes boundaries, and its size must be >= HUF_CTABLE_WORKSPACE_SIZE.
  */
+#define HUF_CTABLE_WORKSPACE_SIZE_U32 (2*HUF_SYMBOLVALUE_MAX +1 +1)
+#define HUF_CTABLE_WORKSPACE_SIZE (HUF_CTABLE_WORKSPACE_SIZE_U32 * sizeof(unsigned))
 size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize);
 
 /*! HUF_readStats() :
-    Read compact Huffman tree, saved by HUF_writeCTable().
-    `huffWeight` is destination buffer.
-    @return : size read from `src` , or an error Code .
-    Note : Needed by HUF_readCTable() and HUF_readDTableXn() . */
-size_t HUF_readStats(BYTE* huffWeight, size_t hwSize, U32* rankStats,
-                     U32* nbSymbolsPtr, U32* tableLogPtr,
+ *  Read compact Huffman tree, saved by HUF_writeCTable().
+ * `huffWeight` is destination buffer.
+ * @return : size read from `src` , or an error Code .
+ *  Note : Needed by HUF_readCTable() and HUF_readDTableXn() . */
+size_t HUF_readStats(BYTE* huffWeight, size_t hwSize,
+                     U32* rankStats, U32* nbSymbolsPtr, U32* tableLogPtr,
                      const void* src, size_t srcSize);
 
 /** HUF_readCTable() :
-*   Loading a CTable saved with HUF_writeCTable() */
+ *  Loading a CTable saved with HUF_writeCTable() */
 size_t HUF_readCTable (HUF_CElt* CTable, unsigned* maxSymbolValuePtr, const void* src, size_t srcSize);
 
 
 /*
-HUF_decompress() does the following:
-1. select the decompression algorithm (X2, X4) based on pre-computed heuristics
-2. build Huffman table from save, using HUF_readDTableXn()
-3. decode 1 or 4 segments in parallel using HUF_decompressSXn_usingDTable
-*/
+ * HUF_decompress() does the following:
+ * 1. select the decompression algorithm (X2, X4) based on pre-computed heuristics
+ * 2. build Huffman table from save, using HUF_readDTableX?()
+ * 3. decode 1 or 4 segments in parallel using HUF_decompress?X?_usingDTable()
+ */
 
 /** HUF_selectDecoder() :
-*   Tells which decoder is likely to decode faster,
-*   based on a set of pre-determined metrics.
-*   @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
-*   Assumption : 0 < cSrcSize < dstSize <= 128 KB */
+ *  Tells which decoder is likely to decode faster,
+ *  based on a set of pre-computed metrics.
+ * @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
+ *  Assumption : 0 < dstSize <= 128 KB */
 U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize);
 
+/**
+ *  The minimum workspace size for the `workSpace` used in
+ *  HUF_readDTableX2_wksp() and HUF_readDTableX4_wksp().
+ *
+ *  The space used depends on HUF_TABLELOG_MAX, ranging from ~1500 bytes when
+ *  HUF_TABLE_LOG_MAX=12 to ~1850 bytes when HUF_TABLE_LOG_MAX=15.
+ *  Buffer overflow errors may potentially occur if code modifications result in
+ *  a required workspace size greater than that specified in the following
+ *  macro.
+ */
+#define HUF_DECOMPRESS_WORKSPACE_SIZE (2 << 10)
+#define HUF_DECOMPRESS_WORKSPACE_SIZE_U32 (HUF_DECOMPRESS_WORKSPACE_SIZE / sizeof(U32))
+
 size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize);
 size_t HUF_readDTableX2_wksp (HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize);
 size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize);
@@ -269,17 +280,23 @@ size_t HUF_decompress4X2_usingDTable(void* dst, size_t maxDstSize, const void* c
 size_t HUF_decompress4X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 
 
+/* ====================== */
 /* single stream variants */
+/* ====================== */
 
 size_t HUF_compress1X (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog);
 size_t HUF_compress1X_wksp (void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize);  /**< `workSpace` must be a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
 size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable);
 /** HUF_compress1X_repeat() :
-*   Same as HUF_compress1X_wksp(), but considers using hufTable if *repeat != HUF_repeat_none.
-*   If it uses hufTable it does not modify hufTable or repeat.
-*   If it doesn't, it sets *repeat = HUF_repeat_none, and it sets hufTable to the table used.
-*   If preferRepeat then the old table will always be used if valid. */
-size_t HUF_compress1X_repeat(void* dst, size_t dstSize, const void* src, size_t srcSize, unsigned maxSymbolValue, unsigned tableLog, void* workSpace, size_t wkspSize, HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat);  /**< `workSpace` must be a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
+ *  Same as HUF_compress1X_wksp(), but considers using hufTable if *repeat != HUF_repeat_none.
+ *  If it uses hufTable it does not modify hufTable or repeat.
+ *  If it doesn't, it sets *repeat = HUF_repeat_none, and it sets hufTable to the table used.
+ *  If preferRepeat then the old table will always be used if valid. */
+size_t HUF_compress1X_repeat(void* dst, size_t dstSize,
+                       const void* src, size_t srcSize,
+                       unsigned maxSymbolValue, unsigned tableLog,
+                       void* workSpace, size_t wkspSize,   /**< `workSpace` must be aligned on 4-bytes boundaries, `wkspSize` must be >= HUF_WORKSPACE_SIZE */
+                       HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2);
 
 size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* single-symbol decoder */
 size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);   /* double-symbol decoder */
@@ -295,6 +312,14 @@ size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize, const void* cS
 size_t HUF_decompress1X2_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 size_t HUF_decompress1X4_usingDTable(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable);
 
+/* BMI2 variants.
+ * If the CPU has BMI2 support, pass bmi2=1, otherwise pass bmi2=0.
+ */
+size_t HUF_decompress1X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2);
+size_t HUF_decompress1X2_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2);
+size_t HUF_decompress4X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2);
+size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2);
+
 #endif /* HUF_STATIC_LINKING_ONLY */
 
 #if defined (__cplusplus)

src/borg/algorithms/zstd/lib/common/mem.h

@@ -56,8 +56,6 @@ MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (size
   typedef   int32_t S32;
   typedef  uint64_t U64;
   typedef   int64_t S64;
-  typedef  intptr_t iPtrDiff;
-  typedef uintptr_t uPtrDiff;
 #else
   typedef unsigned char      BYTE;
   typedef unsigned short      U16;
@@ -66,8 +64,6 @@ MEM_STATIC void MEM_check(void) { MEM_STATIC_ASSERT((sizeof(size_t)==4) || (size
   typedef   signed int        S32;
   typedef unsigned long long  U64;
   typedef   signed long long  S64;
-  typedef ptrdiff_t      iPtrDiff;
-  typedef size_t         uPtrDiff;
 #endif
 
 
@@ -123,20 +119,26 @@ MEM_STATIC void MEM_write64(void* memPtr, U64 value) { *(U64*)memPtr = value; }
 /* currently only defined for gcc and icc */
 #if defined(_MSC_VER) || (defined(__INTEL_COMPILER) && defined(WIN32))
     __pragma( pack(push, 1) )
-    typedef union { U16 u16; U32 u32; U64 u64; size_t st; } unalign;
+    typedef struct { U16 v; } unalign16;
+    typedef struct { U32 v; } unalign32;
+    typedef struct { U64 v; } unalign64;
+    typedef struct { size_t v; } unalignArch;
     __pragma( pack(pop) )
 #else
-    typedef union { U16 u16; U32 u32; U64 u64; size_t st; } __attribute__((packed)) unalign;
+    typedef struct { U16 v; } __attribute__((packed)) unalign16;
+    typedef struct { U32 v; } __attribute__((packed)) unalign32;
+    typedef struct { U64 v; } __attribute__((packed)) unalign64;
+    typedef struct { size_t v; } __attribute__((packed)) unalignArch;
 #endif
 
-MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign*)ptr)->u16; }
-MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign*)ptr)->u32; }
-MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign*)ptr)->u64; }
-MEM_STATIC size_t MEM_readST(const void* ptr) { return ((const unalign*)ptr)->st; }
+MEM_STATIC U16 MEM_read16(const void* ptr) { return ((const unalign16*)ptr)->v; }
+MEM_STATIC U32 MEM_read32(const void* ptr) { return ((const unalign32*)ptr)->v; }
+MEM_STATIC U64 MEM_read64(const void* ptr) { return ((const unalign64*)ptr)->v; }
+MEM_STATIC size_t MEM_readST(const void* ptr) { return ((const unalignArch*)ptr)->v; }
 
-MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign*)memPtr)->u16 = value; }
-MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign*)memPtr)->u32 = value; }
-MEM_STATIC void MEM_write64(void* memPtr, U64 value) { ((unalign*)memPtr)->u64 = value; }
+MEM_STATIC void MEM_write16(void* memPtr, U16 value) { ((unalign16*)memPtr)->v = value; }
+MEM_STATIC void MEM_write32(void* memPtr, U32 value) { ((unalign32*)memPtr)->v = value; }
+MEM_STATIC void MEM_write64(void* memPtr, U64 value) { ((unalign64*)memPtr)->v = value; }
 
 #else
 

src/borg/algorithms/zstd/lib/common/pool.c

@@ -11,8 +11,8 @@
 
 /* ======   Dependencies   ======= */
 #include <stddef.h>  /* size_t */
-#include <stdlib.h>  /* malloc, calloc, free */
 #include "pool.h"
+#include "zstd_internal.h"  /* ZSTD_malloc, ZSTD_free */
 
 /* ======   Compiler specifics   ====== */
 #if defined(_MSC_VER)
@@ -115,7 +115,7 @@ POOL_ctx* POOL_create_advanced(size_t numThreads, size_t queueSize, ZSTD_customM
      * and full queues.
      */
     ctx->queueSize = queueSize + 1;
-    ctx->queue = (POOL_job*) malloc(ctx->queueSize * sizeof(POOL_job));
+    ctx->queue = (POOL_job*)ZSTD_malloc(ctx->queueSize * sizeof(POOL_job), customMem);
     ctx->queueHead = 0;
     ctx->queueTail = 0;
     ctx->numThreadsBusy = 0;
@@ -194,32 +194,54 @@ static int isQueueFull(POOL_ctx const* ctx) {
     }
 }
 
-void POOL_add(void* ctxVoid, POOL_function function, void *opaque) {
-    POOL_ctx* const ctx = (POOL_ctx*)ctxVoid;
-    if (!ctx) { return; }
 
+static void POOL_add_internal(POOL_ctx* ctx, POOL_function function, void *opaque)
+{
+    POOL_job const job = {function, opaque};
+    assert(ctx != NULL);
+    if (ctx->shutdown) return;
+
+    ctx->queueEmpty = 0;
+    ctx->queue[ctx->queueTail] = job;
+    ctx->queueTail = (ctx->queueTail + 1) % ctx->queueSize;
+    ZSTD_pthread_cond_signal(&ctx->queuePopCond);
+}
+
+void POOL_add(POOL_ctx* ctx, POOL_function function, void* opaque)
+{
+    assert(ctx != NULL);
     ZSTD_pthread_mutex_lock(&ctx->queueMutex);
-    {   POOL_job const job = {function, opaque};
+    /* Wait until there is space in the queue for the new job */
+    while (isQueueFull(ctx) && (!ctx->shutdown)) {
+        ZSTD_pthread_cond_wait(&ctx->queuePushCond, &ctx->queueMutex);
+    }
+    POOL_add_internal(ctx, function, opaque);
+    ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
+}
 
-        /* Wait until there is space in the queue for the new job */
-        while (isQueueFull(ctx) && !ctx->shutdown) {
-          ZSTD_pthread_cond_wait(&ctx->queuePushCond, &ctx->queueMutex);
-        }
-        /* The queue is still going => there is space */
-        if (!ctx->shutdown) {
-            ctx->queueEmpty = 0;
-            ctx->queue[ctx->queueTail] = job;
-            ctx->queueTail = (ctx->queueTail + 1) % ctx->queueSize;
-        }
+
+int POOL_tryAdd(POOL_ctx* ctx, POOL_function function, void* opaque)
+{
+    assert(ctx != NULL);
+    ZSTD_pthread_mutex_lock(&ctx->queueMutex);
+    if (isQueueFull(ctx)) {
+        ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
+        return 0;
     }
+    POOL_add_internal(ctx, function, opaque);
     ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
-    ZSTD_pthread_cond_signal(&ctx->queuePopCond);
+    return 1;
 }
 
+
 #else  /* ZSTD_MULTITHREAD  not defined */
+
+/* ========================== */
 /* No multi-threading support */
+/* ========================== */
 
-/* We don't need any data, but if it is empty malloc() might return NULL. */
+
+/* We don't need any data, but if it is empty, malloc() might return NULL. */
 struct POOL_ctx_s {
     int dummy;
 };
@@ -241,9 +263,15 @@ void POOL_free(POOL_ctx* ctx) {
     (void)ctx;
 }
 
-void POOL_add(void* ctx, POOL_function function, void* opaque) {
+void POOL_add(POOL_ctx* ctx, POOL_function function, void* opaque) {
+    (void)ctx;
+    function(opaque);
+}
+
+int POOL_tryAdd(POOL_ctx* ctx, POOL_function function, void* opaque) {
     (void)ctx;
     function(opaque);
+    return 1;
 }
 
 size_t POOL_sizeof(POOL_ctx* ctx) {

src/borg/algorithms/zstd/lib/common/pool.h

@@ -17,7 +17,8 @@ extern "C" {
 
 
 #include <stddef.h>   /* size_t */
-#include "zstd_internal.h"   /* ZSTD_customMem */
+#define ZSTD_STATIC_LINKING_ONLY   /* ZSTD_customMem */
+#include "zstd.h"
 
 typedef struct POOL_ctx_s POOL_ctx;
 
@@ -27,35 +28,43 @@ typedef struct POOL_ctx_s POOL_ctx;
  *  The maximum number of queued jobs before blocking is `queueSize`.
  * @return : POOL_ctx pointer on success, else NULL.
 */
-POOL_ctx *POOL_create(size_t numThreads, size_t queueSize);
+POOL_ctx* POOL_create(size_t numThreads, size_t queueSize);
 
-POOL_ctx *POOL_create_advanced(size_t numThreads, size_t queueSize, ZSTD_customMem customMem);
+POOL_ctx* POOL_create_advanced(size_t numThreads, size_t queueSize, ZSTD_customMem customMem);
 
 /*! POOL_free() :
     Free a thread pool returned by POOL_create().
 */
-void POOL_free(POOL_ctx *ctx);
+void POOL_free(POOL_ctx* ctx);
 
 /*! POOL_sizeof() :
     return memory usage of pool returned by POOL_create().
 */
-size_t POOL_sizeof(POOL_ctx *ctx);
+size_t POOL_sizeof(POOL_ctx* ctx);
 
 /*! POOL_function :
     The function type that can be added to a thread pool.
 */
-typedef void (*POOL_function)(void *);
+typedef void (*POOL_function)(void*);
 /*! POOL_add_function :
     The function type for a generic thread pool add function.
 */
-typedef void (*POOL_add_function)(void *, POOL_function, void *);
+typedef void (*POOL_add_function)(void*, POOL_function, void*);
 
 /*! POOL_add() :
-    Add the job `function(opaque)` to the thread pool.
+    Add the job `function(opaque)` to the thread pool. `ctx` must be valid.
     Possibly blocks until there is room in the queue.
     Note : The function may be executed asynchronously, so `opaque` must live until the function has been completed.
 */
-void POOL_add(void *ctx, POOL_function function, void *opaque);
+void POOL_add(POOL_ctx* ctx, POOL_function function, void* opaque);
+
+
+/*! POOL_tryAdd() :
+    Add the job `function(opaque)` to the thread pool if a worker is available.
+    return immediately otherwise.
+   @return : 1 if successful, 0 if not.
+*/
+int POOL_tryAdd(POOL_ctx* ctx, POOL_function function, void* opaque);
 
 
 #if defined (__cplusplus)

src/borg/algorithms/zstd/lib/common/threading.h

@@ -45,15 +45,15 @@ extern "C" {
 
 /* mutex */
 #define ZSTD_pthread_mutex_t           CRITICAL_SECTION
-#define ZSTD_pthread_mutex_init(a, b)  (InitializeCriticalSection((a)), 0)
+#define ZSTD_pthread_mutex_init(a, b)  ((void)(b), InitializeCriticalSection((a)), 0)
 #define ZSTD_pthread_mutex_destroy(a)  DeleteCriticalSection((a))
 #define ZSTD_pthread_mutex_lock(a)     EnterCriticalSection((a))
 #define ZSTD_pthread_mutex_unlock(a)   LeaveCriticalSection((a))
 
 /* condition variable */
 #define ZSTD_pthread_cond_t             CONDITION_VARIABLE
-#define ZSTD_pthread_cond_init(a, b)    (InitializeConditionVariable((a)), 0)
-#define ZSTD_pthread_cond_destroy(a)    /* No delete */
+#define ZSTD_pthread_cond_init(a, b)    ((void)(b), InitializeConditionVariable((a)), 0)
+#define ZSTD_pthread_cond_destroy(a)    ((void)(a))
 #define ZSTD_pthread_cond_wait(a, b)    SleepConditionVariableCS((a), (b), INFINITE)
 #define ZSTD_pthread_cond_signal(a)     WakeConditionVariable((a))
 #define ZSTD_pthread_cond_broadcast(a)  WakeAllConditionVariable((a))
@@ -100,17 +100,17 @@ int ZSTD_pthread_join(ZSTD_pthread_t thread, void** value_ptr);
 /* No multithreading support */
 
 typedef int ZSTD_pthread_mutex_t;
-#define ZSTD_pthread_mutex_init(a, b)   ((void)a, 0)
-#define ZSTD_pthread_mutex_destroy(a)
-#define ZSTD_pthread_mutex_lock(a)
-#define ZSTD_pthread_mutex_unlock(a)
+#define ZSTD_pthread_mutex_init(a, b)   ((void)(a), (void)(b), 0)
+#define ZSTD_pthread_mutex_destroy(a)   ((void)(a))
+#define ZSTD_pthread_mutex_lock(a)      ((void)(a))
+#define ZSTD_pthread_mutex_unlock(a)    ((void)(a))
 
 typedef int ZSTD_pthread_cond_t;
-#define ZSTD_pthread_cond_init(a, b)    ((void)a, 0)
-#define ZSTD_pthread_cond_destroy(a)
-#define ZSTD_pthread_cond_wait(a, b)
-#define ZSTD_pthread_cond_signal(a)
-#define ZSTD_pthread_cond_broadcast(a)
+#define ZSTD_pthread_cond_init(a, b)    ((void)(a), (void)(b), 0)
+#define ZSTD_pthread_cond_destroy(a)    ((void)(a))
+#define ZSTD_pthread_cond_wait(a, b)    ((void)(a), (void)(b))
+#define ZSTD_pthread_cond_signal(a)     ((void)(a))
+#define ZSTD_pthread_cond_broadcast(a)  ((void)(a))
 
 /* do not use ZSTD_pthread_t */
 

src/borg/algorithms/zstd/lib/common/zstd_common.c

@@ -31,21 +31,27 @@ const char* ZSTD_versionString(void) { return ZSTD_VERSION_STRING; }
 *  ZSTD Error Management
 ******************************************/
 /*! ZSTD_isError() :
-*   tells if a return value is an error code */
+ *  tells if a return value is an error code */
 unsigned ZSTD_isError(size_t code) { return ERR_isError(code); }
 
 /*! ZSTD_getErrorName() :
-*   provides error code string from function result (useful for debugging) */
+ *  provides error code string from function result (useful for debugging) */
 const char* ZSTD_getErrorName(size_t code) { return ERR_getErrorName(code); }
 
 /*! ZSTD_getError() :
-*   convert a `size_t` function result into a proper ZSTD_errorCode enum */
+ *  convert a `size_t` function result into a proper ZSTD_errorCode enum */
 ZSTD_ErrorCode ZSTD_getErrorCode(size_t code) { return ERR_getErrorCode(code); }
 
 /*! ZSTD_getErrorString() :
-*   provides error code string from enum */
+ *  provides error code string from enum */
 const char* ZSTD_getErrorString(ZSTD_ErrorCode code) { return ERR_getErrorString(code); }
 
+/*! g_debuglog_enable :
+ *  turn on/off debug traces (global switch) */
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG >= 2)
+int g_debuglog_enable = 1;
+#endif
+
 
 /*=**************************************************************
 *  Custom allocator

src/borg/algorithms/zstd/lib/common/zstd_errors.h

@@ -35,12 +35,20 @@ extern "C" {
 #  define ZSTDERRORLIB_API ZSTDERRORLIB_VISIBILITY
 #endif
 
-/*-****************************************
- *  error codes list
- *  note : this API is still considered unstable
- *         and shall not be used with a dynamic library.
- *         only static linking is allowed
- ******************************************/
+/*-*********************************************
+ *  Error codes list
+ *-*********************************************
+ *  Error codes _values_ are pinned down since v1.3.1 only.
+ *  Therefore, don't rely on values if you may link to any version < v1.3.1.
+ *
+ *  Only values < 100 are considered stable.
+ *
+ *  note 1 : this API shall be used with static linking only.
+ *           dynamic linking is not yet officially supported.
+ *  note 2 : Prefer relying on the enum than on its value whenever possible
+ *           This is the only supported way to use the error list < v1.3.1
+ *  note 3 : ZSTD_isError() is always correct, whatever the library version.
+ **********************************************/
 typedef enum {
   ZSTD_error_no_error = 0,
   ZSTD_error_GENERIC  = 1,
@@ -61,9 +69,10 @@ typedef enum {
   ZSTD_error_stage_wrong       = 60,
   ZSTD_error_init_missing      = 62,
   ZSTD_error_memory_allocation = 64,
+  ZSTD_error_workSpace_tooSmall= 66,
   ZSTD_error_dstSize_tooSmall = 70,
   ZSTD_error_srcSize_wrong    = 72,
-  /* following error codes are not stable and may be removed or changed in a future version */
+  /* following error codes are __NOT STABLE__, they can be removed or changed in future versions */
   ZSTD_error_frameIndex_tooLarge = 100,
   ZSTD_error_seekableIO          = 102,
   ZSTD_error_maxCode = 120  /* never EVER use this value directly, it can change in future versions! Use ZSTD_isError() instead */

src/borg/algorithms/zstd/lib/common/zstd_internal.h

@@ -11,6 +11,10 @@
 #ifndef ZSTD_CCOMMON_H_MODULE
 #define ZSTD_CCOMMON_H_MODULE
 
+/* this module contains definitions which must be identical
+ * across compression, decompression and dictBuilder.
+ * It also contains a few functions useful to at least 2 of them
+ * and which benefit from being inlined */
 
 /*-*************************************
 *  Dependencies
@@ -50,21 +54,26 @@ extern "C" {
 
 #if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
 #  include <stdio.h>
+extern int g_debuglog_enable;
 /* recommended values for ZSTD_DEBUG display levels :
  * 1 : no display, enables assert() only
- * 2 : reserved for currently active debugging path
- * 3 : events once per object lifetime (CCtx, CDict)
+ * 2 : reserved for currently active debug path
+ * 3 : events once per object lifetime (CCtx, CDict, etc.)
  * 4 : events once per frame
  * 5 : events once per block
  * 6 : events once per sequence (*very* verbose) */
-#  define DEBUGLOG(l, ...) {                         \
-                if (l<=ZSTD_DEBUG) {                 \
-                    fprintf(stderr, __FILE__ ": ");  \
-                    fprintf(stderr, __VA_ARGS__);    \
-                    fprintf(stderr, " \n");          \
+#  define RAWLOG(l, ...) {                                      \
+                if ((g_debuglog_enable) & (l<=ZSTD_DEBUG)) {    \
+                    fprintf(stderr, __VA_ARGS__);               \
+            }   }
+#  define DEBUGLOG(l, ...) {                                    \
+                if ((g_debuglog_enable) & (l<=ZSTD_DEBUG)) {    \
+                    fprintf(stderr, __FILE__ ": " __VA_ARGS__); \
+                    fprintf(stderr, " \n");                     \
             }   }
 #else
-#  define DEBUGLOG(l, ...)      {}    /* disabled */
+#  define RAWLOG(l, ...)      {}    /* disabled */
+#  define DEBUGLOG(l, ...)    {}    /* disabled */
 #endif
 
 
@@ -85,9 +94,7 @@ extern "C" {
 #define ZSTD_OPT_NUM    (1<<12)
 
 #define ZSTD_REP_NUM      3                 /* number of repcodes */
-#define ZSTD_REP_CHECK    (ZSTD_REP_NUM)    /* number of repcodes to check by the optimal parser */
 #define ZSTD_REP_MOVE     (ZSTD_REP_NUM-1)
-#define ZSTD_REP_MOVE_OPT (ZSTD_REP_NUM)
 static const U32 repStartValue[ZSTD_REP_NUM] = { 1, 4, 8 };
 
 #define KB *(1 <<10)
@@ -125,37 +132,50 @@ typedef enum { set_basic, set_rle, set_compressed, set_repeat } symbolEncodingTy
 
 #define Litbits  8
 #define MaxLit ((1<<Litbits) - 1)
-#define MaxML  52
-#define MaxLL  35
+#define MaxML   52
+#define MaxLL   35
 #define DefaultMaxOff 28
-#define MaxOff 31
+#define MaxOff  31
 #define MaxSeq MAX(MaxLL, MaxML)   /* Assumption : MaxOff < MaxLL,MaxML */
 #define MLFSELog    9
 #define LLFSELog    9
 #define OffFSELog   8
+#define MaxFSELog  MAX(MAX(MLFSELog, LLFSELog), OffFSELog)
 
-static const U32 LL_bits[MaxLL+1] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-                                      1, 1, 1, 1, 2, 2, 3, 3, 4, 6, 7, 8, 9,10,11,12,
+static const U32 LL_bits[MaxLL+1] = { 0, 0, 0, 0, 0, 0, 0, 0,
+                                      0, 0, 0, 0, 0, 0, 0, 0,
+                                      1, 1, 1, 1, 2, 2, 3, 3,
+                                      4, 6, 7, 8, 9,10,11,12,
                                      13,14,15,16 };
-static const S16 LL_defaultNorm[MaxLL+1] = { 4, 3, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 1, 1, 1,
-                                             2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 2, 1, 1, 1, 1, 1,
+static const S16 LL_defaultNorm[MaxLL+1] = { 4, 3, 2, 2, 2, 2, 2, 2,
+                                             2, 2, 2, 2, 2, 1, 1, 1,
+                                             2, 2, 2, 2, 2, 2, 2, 2,
+                                             2, 3, 2, 1, 1, 1, 1, 1,
                                             -1,-1,-1,-1 };
 #define LL_DEFAULTNORMLOG 6  /* for static allocation */
 static const U32 LL_defaultNormLog = LL_DEFAULTNORMLOG;
 
-static const U32 ML_bits[MaxML+1] = { 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-                                      0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
-                                      1, 1, 1, 1, 2, 2, 3, 3, 4, 4, 5, 7, 8, 9,10,11,
+static const U32 ML_bits[MaxML+1] = { 0, 0, 0, 0, 0, 0, 0, 0,
+                                      0, 0, 0, 0, 0, 0, 0, 0,
+                                      0, 0, 0, 0, 0, 0, 0, 0,
+                                      0, 0, 0, 0, 0, 0, 0, 0,
+                                      1, 1, 1, 1, 2, 2, 3, 3,
+                                      4, 4, 5, 7, 8, 9,10,11,
                                      12,13,14,15,16 };
-static const S16 ML_defaultNorm[MaxML+1] = { 1, 4, 3, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1,
-                                             1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
-                                             1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,-1,-1,
+static const S16 ML_defaultNorm[MaxML+1] = { 1, 4, 3, 2, 2, 2, 2, 2,
+                                             2, 1, 1, 1, 1, 1, 1, 1,
+                                             1, 1, 1, 1, 1, 1, 1, 1,
+                                             1, 1, 1, 1, 1, 1, 1, 1,
+                                             1, 1, 1, 1, 1, 1, 1, 1,
+                                             1, 1, 1, 1, 1, 1,-1,-1,
                                             -1,-1,-1,-1,-1 };
 #define ML_DEFAULTNORMLOG 6  /* for static allocation */
 static const U32 ML_defaultNormLog = ML_DEFAULTNORMLOG;
 
-static const S16 OF_defaultNorm[DefaultMaxOff+1] = { 1, 1, 1, 1, 1, 1, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1,
-                                                     1, 1, 1, 1, 1, 1, 1, 1,-1,-1,-1,-1,-1 };
+static const S16 OF_defaultNorm[DefaultMaxOff+1] = { 1, 1, 1, 1, 1, 1, 2, 2,
+                                                     2, 1, 1, 1, 1, 1, 1, 1,
+                                                     1, 1, 1, 1, 1, 1, 1, 1,
+                                                    -1,-1,-1,-1,-1 };
 #define OF_DEFAULTNORMLOG 5  /* for static allocation */
 static const U32 OF_defaultNormLog = OF_DEFAULTNORMLOG;
 
@@ -167,7 +187,7 @@ static void ZSTD_copy8(void* dst, const void* src) { memcpy(dst, src, 8); }
 #define COPY8(d,s) { ZSTD_copy8(d,s); d+=8; s+=8; }
 
 /*! ZSTD_wildcopy() :
-*   custom version of memcpy(), can copy up to 7 bytes too many (8 bytes if length==0) */
+ *  custom version of memcpy(), can overwrite up to WILDCOPY_OVERLENGTH bytes (if length==0) */
 #define WILDCOPY_OVERLENGTH 8
 MEM_STATIC void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length)
 {
@@ -191,17 +211,14 @@ MEM_STATIC void ZSTD_wildcopy_e(void* dst, const void* src, void* dstEnd)   /* s
 
 
 /*-*******************************************
-*  Private interfaces
+*  Private declarations
 *********************************************/
-typedef struct ZSTD_stats_s ZSTD_stats_t;
-
 typedef struct seqDef_s {
     U32 offset;
     U16 litLength;
     U16 matchLength;
 } seqDef;
 
-
 typedef struct {
     seqDef* sequencesStart;
     seqDef* sequences;
@@ -212,104 +229,10 @@ typedef struct {
     BYTE* ofCode;
     U32   longLengthID;   /* 0 == no longLength; 1 == Lit.longLength; 2 == Match.longLength; */
     U32   longLengthPos;
-    U32   rep[ZSTD_REP_NUM];
-    U32   repToConfirm[ZSTD_REP_NUM];
 } seqStore_t;
 
-typedef struct {
-    U32 off;
-    U32 len;
-} ZSTD_match_t;
-
-typedef struct {
-    U32 price;
-    U32 off;
-    U32 mlen;
-    U32 litlen;
-    U32 rep[ZSTD_REP_NUM];
-} ZSTD_optimal_t;
-
-typedef struct {
-    U32* litFreq;
-    U32* litLengthFreq;
-    U32* matchLengthFreq;
-    U32* offCodeFreq;
-    ZSTD_match_t* matchTable;
-    ZSTD_optimal_t* priceTable;
-
-    U32  matchLengthSum;
-    U32  matchSum;
-    U32  litLengthSum;
-    U32  litSum;
-    U32  offCodeSum;
-    U32  log2matchLengthSum;
-    U32  log2matchSum;
-    U32  log2litLengthSum;
-    U32  log2litSum;
-    U32  log2offCodeSum;
-    U32  factor;
-    U32  staticPrices;
-    U32  cachedPrice;
-    U32  cachedLitLength;
-    const BYTE* cachedLiterals;
-} optState_t;
-
-typedef struct {
-    U32 offset;
-    U32 checksum;
-} ldmEntry_t;
-
-typedef struct {
-    ldmEntry_t* hashTable;
-    BYTE* bucketOffsets;    /* Next position in bucket to insert entry */
-    U64 hashPower;          /* Used to compute the rolling hash.
-                             * Depends on ldmParams.minMatchLength */
-} ldmState_t;
-
-typedef struct {
-    U32 enableLdm;          /* 1 if enable long distance matching */
-    U32 hashLog;            /* Log size of hashTable */
-    U32 bucketSizeLog;      /* Log bucket size for collision resolution, at most 8 */
-    U32 minMatchLength;     /* Minimum match length */
-    U32 hashEveryLog;       /* Log number of entries to skip */
-} ldmParams_t;
-
-typedef struct {
-    U32 hufCTable[HUF_CTABLE_SIZE_U32(255)];
-    FSE_CTable offcodeCTable[FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)];
-    FSE_CTable matchlengthCTable[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)];
-    FSE_CTable litlengthCTable[FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)];
-    U32 workspace[HUF_WORKSPACE_SIZE_U32];
-    HUF_repeat hufCTable_repeatMode;
-    FSE_repeat offcode_repeatMode;
-    FSE_repeat matchlength_repeatMode;
-    FSE_repeat litlength_repeatMode;
-} ZSTD_entropyCTables_t;
-
-struct ZSTD_CCtx_params_s {
-    ZSTD_format_e format;
-    ZSTD_compressionParameters cParams;
-    ZSTD_frameParameters fParams;
-
-    int compressionLevel;
-    U32 forceWindow;           /* force back-references to respect limit of
-                                * 1<<wLog, even for dictionary */
-
-    /* Multithreading: used to pass parameters to mtctx */
-    U32 nbThreads;
-    unsigned jobSize;
-    unsigned overlapSizeLog;
-
-    /* Long distance matching parameters */
-    ldmParams_t ldmParams;
-
-    /* For use with createCCtxParams() and freeCCtxParams() only */
-    ZSTD_customMem customMem;
-
-};  /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
-
-const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx);
-void ZSTD_seqToCodes(const seqStore_t* seqStorePtr);
+const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx);   /* compress & dictBuilder */
+void ZSTD_seqToCodes(const seqStore_t* seqStorePtr);   /* compress, dictBuilder, decodeCorpus (shouldn't get its definition from here) */
 
 /* custom memory allocation functions */
 void* ZSTD_malloc(size_t size, ZSTD_customMem customMem);
@@ -317,9 +240,7 @@ void* ZSTD_calloc(size_t size, ZSTD_customMem customMem);
 void ZSTD_free(void* ptr, ZSTD_customMem customMem);
 
 
-/*======  common function  ======*/
-
-MEM_STATIC U32 ZSTD_highbit32(U32 val)
+MEM_STATIC U32 ZSTD_highbit32(U32 val)   /* compress, dictBuilder, decodeCorpus */
 {
     assert(val != 0);
     {
@@ -330,66 +251,25 @@ MEM_STATIC U32 ZSTD_highbit32(U32 val)
 #   elif defined(__GNUC__) && (__GNUC__ >= 3)   /* GCC Intrinsic */
         return 31 - __builtin_clz(val);
 #   else   /* Software version */
-        static const int DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
+        static const U32 DeBruijnClz[32] = { 0, 9, 1, 10, 13, 21, 2, 29, 11, 14, 16, 18, 22, 25, 3, 30, 8, 12, 20, 28, 15, 17, 24, 7, 19, 27, 23, 6, 26, 5, 4, 31 };
         U32 v = val;
-        int r;
         v |= v >> 1;
         v |= v >> 2;
         v |= v >> 4;
         v |= v >> 8;
         v |= v >> 16;
-        r = DeBruijnClz[(U32)(v * 0x07C4ACDDU) >> 27];
-        return r;
+        return DeBruijnClz[(v * 0x07C4ACDDU) >> 27];
 #   endif
     }
 }
 
 
-/* hidden functions */
-
 /* ZSTD_invalidateRepCodes() :
  * ensures next compression will not use repcodes from previous block.
  * Note : only works with regular variant;
  *        do not use with extDict variant ! */
-void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx);
-
-
-/*! ZSTD_initCStream_internal() :
- *  Private use only. Init streaming operation.
- *  expects params to be valid.
- *  must receive dict, or cdict, or none, but not both.
- *  @return : 0, or an error code */
-size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
-                     const void* dict, size_t dictSize,
-                     const ZSTD_CDict* cdict,
-                     ZSTD_CCtx_params  params, unsigned long long pledgedSrcSize);
-
-/*! ZSTD_compressStream_generic() :
- *  Private use only. To be called from zstdmt_compress.c in single-thread mode. */
-size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
-                                   ZSTD_outBuffer* output,
-                                   ZSTD_inBuffer* input,
-                                   ZSTD_EndDirective const flushMode);
-
-/*! ZSTD_getCParamsFromCDict() :
- *  as the name implies */
-ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict);
-
-/* ZSTD_compressBegin_advanced_internal() :
- * Private use only. To be called from zstdmt_compress.c. */
-size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
-                                    const void* dict, size_t dictSize,
-                                    ZSTD_dictMode_e dictMode,
-                                    ZSTD_CCtx_params params,
-                                    unsigned long long pledgedSrcSize);
-
-/* ZSTD_compress_advanced_internal() :
- * Private use only. To be called from zstdmt_compress.c. */
-size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
-                                       void* dst, size_t dstCapacity,
-                                 const void* src, size_t srcSize,
-                                 const void* dict,size_t dictSize,
-                                 ZSTD_CCtx_params params);
+void ZSTD_invalidateRepCodes(ZSTD_CCtx* cctx);   /* zstdmt, adaptive_compression (shouldn't get this definition from here) */
+
 
 typedef struct {
     blockType_e blockType;
@@ -398,7 +278,8 @@ typedef struct {
 } blockProperties_t;
 
 /*! ZSTD_getcBlockSize() :
-*   Provides the size of compressed block from block header `src` */
+ *  Provides the size of compressed block from block header `src` */
+/* Used by: decompress, fullbench (does not get its definition from here) */
 size_t ZSTD_getcBlockSize(const void* src, size_t srcSize,
                           blockProperties_t* bpPtr);
 

src/borg/algorithms/zstd/lib/compress/fse_compress.c

@@ -248,7 +248,7 @@ static size_t FSE_writeNCount_generic (void* header, size_t headerBufferSize,
             bitCount  -= (count<max);
             previous0  = (count==1);
             if (remaining<1) return ERROR(GENERIC);
-            while (remaining<threshold) nbBits--, threshold>>=1;
+            while (remaining<threshold) { nbBits--; threshold>>=1; }
         }
         if (bitCount>16) {
             if ((!writeIsSafe) && (out > oend - 2)) return ERROR(dstSize_tooSmall);   /* Buffer overflow */
@@ -292,7 +292,7 @@ size_t FSE_writeNCount (void* buffer, size_t bufferSize, const short* normalized
     It doesn't use any additional memory.
     But this function is unsafe : it doesn't check that all values within `src` can fit into `count`.
     For this reason, prefer using a table `count` with 256 elements.
-    @return : count of most numerous element
+    @return : count of most numerous element.
 */
 size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
                         const void* src, size_t srcSize)
@@ -305,7 +305,10 @@ size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
     memset(count, 0, (maxSymbolValue+1)*sizeof(*count));
     if (srcSize==0) { *maxSymbolValuePtr = 0; return 0; }
 
-    while (ip<end) count[*ip++]++;
+    while (ip<end) {
+        assert(*ip <= maxSymbolValue);
+        count[*ip++]++;
+    }
 
     while (!count[maxSymbolValue]) maxSymbolValue--;
     *maxSymbolValuePtr = maxSymbolValue;
@@ -318,7 +321,8 @@ size_t FSE_count_simple(unsigned* count, unsigned* maxSymbolValuePtr,
 
 /* FSE_count_parallel_wksp() :
  * Same as FSE_count_parallel(), but using an externally provided scratch buffer.
- * `workSpace` size must be a minimum of `1024 * sizeof(unsigned)`` */
+ * `workSpace` size must be a minimum of `1024 * sizeof(unsigned)`.
+ * @return : largest histogram frequency, or an error code (notably when histogram would be larger than *maxSymbolValuePtr). */
 static size_t FSE_count_parallel_wksp(
                                 unsigned* count, unsigned* maxSymbolValuePtr,
                                 const void* source, size_t sourceSize,
@@ -333,7 +337,7 @@ static size_t FSE_count_parallel_wksp(
     U32* const Counting3 = Counting2 + 256;
     U32* const Counting4 = Counting3 + 256;
 
-    memset(Counting1, 0, 4*256*sizeof(unsigned));
+    memset(workSpace, 0, 4*256*sizeof(unsigned));
 
     /* safety checks */
     if (!sourceSize) {
@@ -379,7 +383,9 @@ static size_t FSE_count_parallel_wksp(
             if (Counting1[s]) return ERROR(maxSymbolValue_tooSmall);
     }   }
 
-    {   U32 s; for (s=0; s<=maxSymbolValue; s++) {
+    {   U32 s;
+        if (maxSymbolValue > 255) maxSymbolValue = 255;
+        for (s=0; s<=maxSymbolValue; s++) {
             count[s] = Counting1[s] + Counting2[s] + Counting3[s] + Counting4[s];
             if (count[s] > max) max = count[s];
     }   }
@@ -393,9 +399,11 @@ static size_t FSE_count_parallel_wksp(
  * Same as FSE_countFast(), but using an externally provided scratch buffer.
  * `workSpace` size must be table of >= `1024` unsigned */
 size_t FSE_countFast_wksp(unsigned* count, unsigned* maxSymbolValuePtr,
-                     const void* source, size_t sourceSize, unsigned* workSpace)
+                          const void* source, size_t sourceSize,
+                          unsigned* workSpace)
 {
-    if (sourceSize < 1500) return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
+    if (sourceSize < 1500) /* heuristic threshold */
+        return FSE_count_simple(count, maxSymbolValuePtr, source, sourceSize);
     return FSE_count_parallel_wksp(count, maxSymbolValuePtr, source, sourceSize, 0, workSpace);
 }
 
@@ -540,7 +548,7 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
            find max, then give all remaining points to max */
         U32 maxV = 0, maxC = 0;
         for (s=0; s<=maxSymbolValue; s++)
-            if (count[s] > maxC) maxV=s, maxC=count[s];
+            if (count[s] > maxC) { maxV=s; maxC=count[s]; }
         norm[maxV] += (short)ToDistribute;
         return 0;
     }
@@ -548,7 +556,7 @@ static size_t FSE_normalizeM2(short* norm, U32 tableLog, const unsigned* count,
     if (total == 0) {
         /* all of the symbols were low enough for the lowOne or lowThreshold */
         for (s=0; ToDistribute > 0; s = (s+1)%(maxSymbolValue+1))
-            if (norm[s] > 0) ToDistribute--, norm[s]++;
+            if (norm[s] > 0) { ToDistribute--; norm[s]++; }
         return 0;
     }
 
@@ -604,7 +612,7 @@ size_t FSE_normalizeCount (short* normalizedCounter, unsigned tableLog,
                     U64 restToBeat = vStep * rtbTable[proba];
                     proba += (count[s]*step) - ((U64)proba<<scale) > restToBeat;
                 }
-                if (proba > largestP) largestP=proba, largest=s;
+                if (proba > largestP) { largestP=proba; largest=s; }
                 normalizedCounter[s] = proba;
                 stillToDistribute -= proba;
         }   }

src/borg/algorithms/zstd/lib/compress/huf_compress.c

@@ -46,6 +46,7 @@
 #include <string.h>     /* memcpy, memset */
 #include <stdio.h>      /* printf (debug) */
 #include "bitstream.h"
+#include "compiler.h"
 #define FSE_STATIC_LINKING_ONLY   /* FSE_optimalTableLog_internal */
 #include "fse.h"        /* header compression */
 #define HUF_STATIC_LINKING_ONLY
@@ -322,7 +323,10 @@ static void HUF_sort(nodeElt* huffNode, const U32* count, U32 maxSymbolValue)
         U32 const c = count[n];
         U32 const r = BIT_highbit32(c+1) + 1;
         U32 pos = rank[r].current++;
-        while ((pos > rank[r].base) && (c > huffNode[pos-1].count)) huffNode[pos]=huffNode[pos-1], pos--;
+        while ((pos > rank[r].base) && (c > huffNode[pos-1].count)) {
+            huffNode[pos] = huffNode[pos-1];
+            pos--;
+        }
         huffNode[pos].count = c;
         huffNode[pos].byte  = (BYTE)n;
     }
@@ -331,10 +335,10 @@ static void HUF_sort(nodeElt* huffNode, const U32* count, U32 maxSymbolValue)
 
 /** HUF_buildCTable_wksp() :
  *  Same as HUF_buildCTable(), but using externally allocated scratch buffer.
- *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of 1024 unsigned.
+ *  `workSpace` must be aligned on 4-bytes boundaries, and be at least as large as a table of HUF_CTABLE_WORKSPACE_SIZE_U32 unsigned.
  */
 #define STARTNODE (HUF_SYMBOLVALUE_MAX+1)
-typedef nodeElt huffNodeTable[2*HUF_SYMBOLVALUE_MAX+1 +1];
+typedef nodeElt huffNodeTable[HUF_CTABLE_WORKSPACE_SIZE_U32];
 size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits, void* workSpace, size_t wkspSize)
 {
     nodeElt* const huffNode0 = (nodeElt*)workSpace;
@@ -345,9 +349,10 @@ size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValu
     U32 nodeRoot;
 
     /* safety checks */
-    if (wkspSize < sizeof(huffNodeTable)) return ERROR(GENERIC);   /* workSpace is not large enough */
+    if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
+    if (wkspSize < sizeof(huffNodeTable)) return ERROR(workSpace_tooSmall);
     if (maxNbBits == 0) maxNbBits = HUF_TABLELOG_DEFAULT;
-    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(GENERIC);
+    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
     memset(huffNode0, 0, sizeof(huffNodeTable));
 
     /* sort, decreasing order */
@@ -405,6 +410,7 @@ size_t HUF_buildCTable_wksp (HUF_CElt* tree, const U32* count, U32 maxSymbolValu
 }
 
 /** HUF_buildCTable() :
+ * @return : maxNbBits
  *  Note : count is used before tree is written, so they can safely overlap
  */
 size_t HUF_buildCTable (HUF_CElt* tree, const U32* count, U32 maxSymbolValue, U32 maxNbBits)
@@ -432,13 +438,14 @@ static int HUF_validateCTable(const HUF_CElt* CTable, const unsigned* count, uns
   return !bad;
 }
 
-static void HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
+size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
+
+FORCE_INLINE_TEMPLATE void
+HUF_encodeSymbol(BIT_CStream_t* bitCPtr, U32 symbol, const HUF_CElt* CTable)
 {
     BIT_addBitsFast(bitCPtr, CTable[symbol].val, CTable[symbol].nbBits);
 }
 
-size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
-
 #define HUF_FLUSHBITS(s)  BIT_flushBits(s)
 
 #define HUF_FLUSHBITS_1(stream) \
@@ -447,7 +454,10 @@ size_t HUF_compressBound(size_t size) { return HUF_COMPRESSBOUND(size); }
 #define HUF_FLUSHBITS_2(stream) \
     if (sizeof((stream)->bitContainer)*8 < HUF_TABLELOG_MAX*4+7) HUF_FLUSHBITS(stream)
 
-size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+FORCE_INLINE_TEMPLATE size_t
+HUF_compress1X_usingCTable_internal_body(void* dst, size_t dstSize,
+                                   const void* src, size_t srcSize,
+                                   const HUF_CElt* CTable)
 {
     const BYTE* ip = (const BYTE*) src;
     BYTE* const ostart = (BYTE*)dst;
@@ -491,8 +501,58 @@ size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, si
     return BIT_closeCStream(&bitC);
 }
 
+#if DYNAMIC_BMI2
 
-size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+static TARGET_ATTRIBUTE("bmi2") size_t
+HUF_compress1X_usingCTable_internal_bmi2(void* dst, size_t dstSize,
+                                   const void* src, size_t srcSize,
+                                   const HUF_CElt* CTable)
+{
+    return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
+}
+
+static size_t
+HUF_compress1X_usingCTable_internal_default(void* dst, size_t dstSize,
+                                      const void* src, size_t srcSize,
+                                      const HUF_CElt* CTable)
+{
+    return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
+}
+
+static size_t
+HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
+                              const void* src, size_t srcSize,
+                              const HUF_CElt* CTable, const int bmi2)
+{
+    if (bmi2) {
+        return HUF_compress1X_usingCTable_internal_bmi2(dst, dstSize, src, srcSize, CTable);
+    }
+    return HUF_compress1X_usingCTable_internal_default(dst, dstSize, src, srcSize, CTable);
+}
+
+#else
+
+static size_t
+HUF_compress1X_usingCTable_internal(void* dst, size_t dstSize,
+                              const void* src, size_t srcSize,
+                              const HUF_CElt* CTable, const int bmi2)
+{
+    (void)bmi2;
+    return HUF_compress1X_usingCTable_internal_body(dst, dstSize, src, srcSize, CTable);
+}
+
+#endif
+
+size_t HUF_compress1X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+{
+    return HUF_compress1X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
+}
+
+
+static size_t
+HUF_compress4X_usingCTable_internal(void* dst, size_t dstSize,
+                              const void* src, size_t srcSize,
+                              const HUF_CElt* CTable, int bmi2)
 {
     size_t const segmentSize = (srcSize+3)/4;   /* first 3 segments */
     const BYTE* ip = (const BYTE*) src;
@@ -505,28 +565,31 @@ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, si
     if (srcSize < 12) return 0;   /* no saving possible : too small input */
     op += 6;   /* jumpTable */
 
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
         if (cSize==0) return 0;
+        assert(cSize <= 65535);
         MEM_writeLE16(ostart, (U16)cSize);
         op += cSize;
     }
 
     ip += segmentSize;
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
         if (cSize==0) return 0;
+        assert(cSize <= 65535);
         MEM_writeLE16(ostart+2, (U16)cSize);
         op += cSize;
     }
 
     ip += segmentSize;
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, segmentSize, CTable) );
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, segmentSize, CTable, bmi2) );
         if (cSize==0) return 0;
+        assert(cSize <= 65535);
         MEM_writeLE16(ostart+4, (U16)cSize);
         op += cSize;
     }
 
     ip += segmentSize;
-    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable(op, oend-op, ip, iend-ip, CTable) );
+    {   CHECK_V_F(cSize, HUF_compress1X_usingCTable_internal(op, oend-op, ip, iend-ip, CTable, bmi2) );
         if (cSize==0) return 0;
         op += cSize;
     }
@@ -534,15 +597,20 @@ size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, si
     return op-ostart;
 }
 
+size_t HUF_compress4X_usingCTable(void* dst, size_t dstSize, const void* src, size_t srcSize, const HUF_CElt* CTable)
+{
+    return HUF_compress4X_usingCTable_internal(dst, dstSize, src, srcSize, CTable, /* bmi2 */ 0);
+}
+
 
 static size_t HUF_compressCTable_internal(
                 BYTE* const ostart, BYTE* op, BYTE* const oend,
                 const void* src, size_t srcSize,
-                unsigned singleStream, const HUF_CElt* CTable)
+                unsigned singleStream, const HUF_CElt* CTable, const int bmi2)
 {
     size_t const cSize = singleStream ?
-                         HUF_compress1X_usingCTable(op, oend - op, src, srcSize, CTable) :
-                         HUF_compress4X_usingCTable(op, oend - op, src, srcSize, CTable);
+                         HUF_compress1X_usingCTable_internal(op, oend - op, src, srcSize, CTable, bmi2) :
+                         HUF_compress4X_usingCTable_internal(op, oend - op, src, srcSize, CTable, bmi2);
     if (HUF_isError(cSize)) { return cSize; }
     if (cSize==0) { return 0; }   /* uncompressible */
     op += cSize;
@@ -551,86 +619,98 @@ static size_t HUF_compressCTable_internal(
     return op-ostart;
 }
 
+typedef struct {
+    U32 count[HUF_SYMBOLVALUE_MAX + 1];
+    HUF_CElt CTable[HUF_SYMBOLVALUE_MAX + 1];
+    huffNodeTable nodeTable;
+} HUF_compress_tables_t;
 
-/* `workSpace` must a table of at least 1024 unsigned */
+/* HUF_compress_internal() :
+ * `workSpace` must a table of at least HUF_WORKSPACE_SIZE_U32 unsigned */
 static size_t HUF_compress_internal (
                 void* dst, size_t dstSize,
                 const void* src, size_t srcSize,
                 unsigned maxSymbolValue, unsigned huffLog,
                 unsigned singleStream,
                 void* workSpace, size_t wkspSize,
-                HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat)
+                HUF_CElt* oldHufTable, HUF_repeat* repeat, int preferRepeat,
+                const int bmi2)
 {
+    HUF_compress_tables_t* const table = (HUF_compress_tables_t*)workSpace;
     BYTE* const ostart = (BYTE*)dst;
     BYTE* const oend = ostart + dstSize;
     BYTE* op = ostart;
 
-    U32* count;
-    size_t const countSize = sizeof(U32) * (HUF_SYMBOLVALUE_MAX + 1);
-    HUF_CElt* CTable;
-    size_t const CTableSize = sizeof(HUF_CElt) * (HUF_SYMBOLVALUE_MAX + 1);
-
     /* checks & inits */
-    if (wkspSize < sizeof(huffNodeTable) + countSize + CTableSize) return ERROR(GENERIC);
-    if (!srcSize) return 0;  /* Uncompressed (note : 1 means rle, so first byte must be correct) */
-    if (!dstSize) return 0;  /* cannot fit within dst budget */
+    if (((size_t)workSpace & 3) != 0) return ERROR(GENERIC);  /* must be aligned on 4-bytes boundaries */
+    if (wkspSize < sizeof(*table)) return ERROR(workSpace_tooSmall);
+    if (!srcSize) return 0;  /* Uncompressed */
+    if (!dstSize) return 0;  /* cannot fit anything within dst budget */
     if (srcSize > HUF_BLOCKSIZE_MAX) return ERROR(srcSize_wrong);   /* current block size limit */
     if (huffLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
+    if (maxSymbolValue > HUF_SYMBOLVALUE_MAX) return ERROR(maxSymbolValue_tooLarge);
     if (!maxSymbolValue) maxSymbolValue = HUF_SYMBOLVALUE_MAX;
     if (!huffLog) huffLog = HUF_TABLELOG_DEFAULT;
 
-    count = (U32*)workSpace;
-    workSpace = (BYTE*)workSpace + countSize;
-    wkspSize -= countSize;
-    CTable = (HUF_CElt*)workSpace;
-    workSpace = (BYTE*)workSpace + CTableSize;
-    wkspSize -= CTableSize;
-
-    /* Heuristic : If we don't need to check the validity of the old table use the old table for small inputs */
+    /* Heuristic : If old table is valid, use it for small inputs */
     if (preferRepeat && repeat && *repeat == HUF_repeat_valid) {
-        return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
+        return HUF_compressCTable_internal(ostart, op, oend,
+                                           src, srcSize,
+                                           singleStream, oldHufTable, bmi2);
     }
 
     /* Scan input and build symbol stats */
-    {   CHECK_V_F(largest, FSE_count_wksp (count, &maxSymbolValue, (const BYTE*)src, srcSize, (U32*)workSpace) );
+    {   CHECK_V_F(largest, FSE_count_wksp (table->count, &maxSymbolValue, (const BYTE*)src, srcSize, table->count) );
         if (largest == srcSize) { *ostart = ((const BYTE*)src)[0]; return 1; }   /* single symbol, rle */
-        if (largest <= (srcSize >> 7)+1) return 0;   /* Fast heuristic : not compressible enough */
+        if (largest <= (srcSize >> 7)+1) return 0;   /* heuristic : probably not compressible enough */
     }
 
     /* Check validity of previous table */
-    if (repeat && *repeat == HUF_repeat_check && !HUF_validateCTable(oldHufTable, count, maxSymbolValue)) {
+    if ( repeat
+      && *repeat == HUF_repeat_check
+      && !HUF_validateCTable(oldHufTable, table->count, maxSymbolValue)) {
         *repeat = HUF_repeat_none;
     }
     /* Heuristic : use existing table for small inputs */
     if (preferRepeat && repeat && *repeat != HUF_repeat_none) {
-        return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
+        return HUF_compressCTable_internal(ostart, op, oend,
+                                           src, srcSize,
+                                           singleStream, oldHufTable, bmi2);
     }
 
     /* Build Huffman Tree */
     huffLog = HUF_optimalTableLog(huffLog, srcSize, maxSymbolValue);
-    {   CHECK_V_F(maxBits, HUF_buildCTable_wksp (CTable, count, maxSymbolValue, huffLog, workSpace, wkspSize) );
+    {   CHECK_V_F(maxBits, HUF_buildCTable_wksp(table->CTable, table->count,
+                                                maxSymbolValue, huffLog,
+                                                table->nodeTable, sizeof(table->nodeTable)) );
         huffLog = (U32)maxBits;
-        /* Zero the unused symbols so we can check it for validity */
-        memset(CTable + maxSymbolValue + 1, 0, CTableSize - (maxSymbolValue + 1) * sizeof(HUF_CElt));
+        /* Zero unused symbols in CTable, so we can check it for validity */
+        memset(table->CTable + (maxSymbolValue + 1), 0,
+               sizeof(table->CTable) - ((maxSymbolValue + 1) * sizeof(HUF_CElt)));
     }
 
     /* Write table description header */
-    {   CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, CTable, maxSymbolValue, huffLog) );
-        /* Check if using the previous table will be beneficial */
+    {   CHECK_V_F(hSize, HUF_writeCTable (op, dstSize, table->CTable, maxSymbolValue, huffLog) );
+        /* Check if using previous huffman table is beneficial */
         if (repeat && *repeat != HUF_repeat_none) {
-            size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, count, maxSymbolValue);
-            size_t const newSize = HUF_estimateCompressedSize(CTable, count, maxSymbolValue);
+            size_t const oldSize = HUF_estimateCompressedSize(oldHufTable, table->count, maxSymbolValue);
+            size_t const newSize = HUF_estimateCompressedSize(table->CTable, table->count, maxSymbolValue);
             if (oldSize <= hSize + newSize || hSize + 12 >= srcSize) {
-                return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, oldHufTable);
-            }
-        }
-        /* Use the new table */
+                return HUF_compressCTable_internal(ostart, op, oend,
+                                                   src, srcSize,
+                                                   singleStream, oldHufTable, bmi2);
+        }   }
+
+        /* Use the new huffman table */
         if (hSize + 12ul >= srcSize) { return 0; }
         op += hSize;
         if (repeat) { *repeat = HUF_repeat_none; }
-        if (oldHufTable) { memcpy(oldHufTable, CTable, CTableSize); } /* Save the new table */
+        if (oldHufTable)
+            memcpy(oldHufTable, table->CTable, sizeof(table->CTable));  /* Save new table */
     }
-    return HUF_compressCTable_internal(ostart, op, oend, src, srcSize, singleStream, CTable);
+    return HUF_compressCTable_internal(ostart, op, oend,
+                                       src, srcSize,
+                                       singleStream, table->CTable, bmi2);
 }
 
 
@@ -639,52 +719,70 @@ size_t HUF_compress1X_wksp (void* dst, size_t dstSize,
                       unsigned maxSymbolValue, unsigned huffLog,
                       void* workSpace, size_t wkspSize)
 {
-    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, NULL, NULL, 0);
+    return HUF_compress_internal(dst, dstSize, src, srcSize,
+                                 maxSymbolValue, huffLog, 1 /*single stream*/,
+                                 workSpace, wkspSize,
+                                 NULL, NULL, 0, 0 /*bmi2*/);
 }
 
 size_t HUF_compress1X_repeat (void* dst, size_t dstSize,
                       const void* src, size_t srcSize,
                       unsigned maxSymbolValue, unsigned huffLog,
                       void* workSpace, size_t wkspSize,
-                      HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat)
+                      HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
 {
-    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 1 /* single stream */, workSpace, wkspSize, hufTable, repeat, preferRepeat);
+    return HUF_compress_internal(dst, dstSize, src, srcSize,
+                                 maxSymbolValue, huffLog, 1 /*single stream*/,
+                                 workSpace, wkspSize, hufTable,
+                                 repeat, preferRepeat, bmi2);
 }
 
 size_t HUF_compress1X (void* dst, size_t dstSize,
                  const void* src, size_t srcSize,
                  unsigned maxSymbolValue, unsigned huffLog)
 {
-    unsigned workSpace[1024];
+    unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
     return HUF_compress1X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
 }
 
+/* HUF_compress4X_repeat():
+ * compress input using 4 streams.
+ * provide workspace to generate compression tables */
 size_t HUF_compress4X_wksp (void* dst, size_t dstSize,
                       const void* src, size_t srcSize,
                       unsigned maxSymbolValue, unsigned huffLog,
                       void* workSpace, size_t wkspSize)
 {
-    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, NULL, NULL, 0);
+    return HUF_compress_internal(dst, dstSize, src, srcSize,
+                                 maxSymbolValue, huffLog, 0 /*4 streams*/,
+                                 workSpace, wkspSize,
+                                 NULL, NULL, 0, 0 /*bmi2*/);
 }
 
+/* HUF_compress4X_repeat():
+ * compress input using 4 streams.
+ * re-use an existing huffman compression table */
 size_t HUF_compress4X_repeat (void* dst, size_t dstSize,
                       const void* src, size_t srcSize,
                       unsigned maxSymbolValue, unsigned huffLog,
                       void* workSpace, size_t wkspSize,
-                      HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat)
+                      HUF_CElt* hufTable, HUF_repeat* repeat, int preferRepeat, int bmi2)
 {
-    return HUF_compress_internal(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, 0 /* 4 streams */, workSpace, wkspSize, hufTable, repeat, preferRepeat);
+    return HUF_compress_internal(dst, dstSize, src, srcSize,
+                                 maxSymbolValue, huffLog, 0 /* 4 streams */,
+                                 workSpace, wkspSize,
+                                 hufTable, repeat, preferRepeat, bmi2);
 }
 
 size_t HUF_compress2 (void* dst, size_t dstSize,
                 const void* src, size_t srcSize,
                 unsigned maxSymbolValue, unsigned huffLog)
 {
-    unsigned workSpace[1024];
+    unsigned workSpace[HUF_WORKSPACE_SIZE_U32];
     return HUF_compress4X_wksp(dst, dstSize, src, srcSize, maxSymbolValue, huffLog, workSpace, sizeof(workSpace));
 }
 
 size_t HUF_compress (void* dst, size_t maxDstSize, const void* src, size_t srcSize)
 {
-    return HUF_compress2(dst, maxDstSize, src, (U32)srcSize, 255, HUF_TABLELOG_DEFAULT);
+    return HUF_compress2(dst, maxDstSize, src, srcSize, 255, HUF_TABLELOG_DEFAULT);
 }

src/borg/algorithms/zstd/lib/compress/zstd_compress.h

@@ -1,307 +0,0 @@
-/*
- * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
- * All rights reserved.
- *
- * This source code is licensed under both the BSD-style license (found in the
- * LICENSE file in the root directory of this source tree) and the GPLv2 (found
- * in the COPYING file in the root directory of this source tree).
- * You may select, at your option, one of the above-listed licenses.
- */
-
-
-#ifndef ZSTD_COMPRESS_H
-#define ZSTD_COMPRESS_H
-
-/*-*************************************
-*  Dependencies
-***************************************/
-#include "zstd_internal.h"
-#ifdef ZSTD_MULTITHREAD
-#  include "zstdmt_compress.h"
-#endif
-
-#if defined (__cplusplus)
-extern "C" {
-#endif
-
-/*-*************************************
-*  Constants
-***************************************/
-static const U32 g_searchStrength = 8;
-#define HASH_READ_SIZE 8
-
-
-/*-*************************************
-*  Context memory management
-***************************************/
-typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
-typedef enum { zcss_init=0, zcss_load, zcss_flush } ZSTD_cStreamStage;
-
-typedef struct ZSTD_prefixDict_s {
-    const void* dict;
-    size_t dictSize;
-    ZSTD_dictMode_e dictMode;
-} ZSTD_prefixDict;
-
-struct ZSTD_CCtx_s {
-    const BYTE* nextSrc;    /* next block here to continue on current prefix */
-    const BYTE* base;       /* All regular indexes relative to this position */
-    const BYTE* dictBase;   /* extDict indexes relative to this position */
-    U32   dictLimit;        /* below that point, need extDict */
-    U32   lowLimit;         /* below that point, no more data */
-    U32   nextToUpdate;     /* index from which to continue dictionary update */
-    U32   nextToUpdate3;    /* index from which to continue dictionary update */
-    U32   hashLog3;         /* dispatch table : larger == faster, more memory */
-    U32   loadedDictEnd;    /* index of end of dictionary */
-    ZSTD_compressionStage_e stage;
-    U32   dictID;
-    ZSTD_CCtx_params requestedParams;
-    ZSTD_CCtx_params appliedParams;
-    void* workSpace;
-    size_t workSpaceSize;
-    size_t blockSize;
-    U64 pledgedSrcSizePlusOne;  /* this way, 0 (default) == unknown */
-    U64 consumedSrcSize;
-    XXH64_state_t xxhState;
-    ZSTD_customMem customMem;
-    size_t staticSize;
-
-    seqStore_t seqStore;    /* sequences storage ptrs */
-    optState_t optState;
-    ldmState_t ldmState;    /* long distance matching state */
-    U32* hashTable;
-    U32* hashTable3;
-    U32* chainTable;
-    ZSTD_entropyCTables_t* entropy;
-
-    /* streaming */
-    char*  inBuff;
-    size_t inBuffSize;
-    size_t inToCompress;
-    size_t inBuffPos;
-    size_t inBuffTarget;
-    char*  outBuff;
-    size_t outBuffSize;
-    size_t outBuffContentSize;
-    size_t outBuffFlushedSize;
-    ZSTD_cStreamStage streamStage;
-    U32    frameEnded;
-
-    /* Dictionary */
-    ZSTD_CDict* cdictLocal;
-    const ZSTD_CDict* cdict;
-    ZSTD_prefixDict prefixDict;   /* single-usage dictionary */
-
-    /* Multi-threading */
-#ifdef ZSTD_MULTITHREAD
-    ZSTDMT_CCtx* mtctx;
-#endif
-};
-
-
-static const BYTE LL_Code[64] = {  0,  1,  2,  3,  4,  5,  6,  7,
-                                   8,  9, 10, 11, 12, 13, 14, 15,
-                                  16, 16, 17, 17, 18, 18, 19, 19,
-                                  20, 20, 20, 20, 21, 21, 21, 21,
-                                  22, 22, 22, 22, 22, 22, 22, 22,
-                                  23, 23, 23, 23, 23, 23, 23, 23,
-                                  24, 24, 24, 24, 24, 24, 24, 24,
-                                  24, 24, 24, 24, 24, 24, 24, 24 };
-
-static const BYTE ML_Code[128] = { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
-                                  16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
-                                  32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
-                                  38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39,
-                                  40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
-                                  41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
-                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
-                                  42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42 };
-
-/*! ZSTD_storeSeq() :
-    Store a sequence (literal length, literals, offset code and match length code) into seqStore_t.
-    `offsetCode` : distance to match, or 0 == repCode.
-    `matchCode` : matchLength - MINMATCH
-*/
-MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const void* literals, U32 offsetCode, size_t matchCode)
-{
-#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG >= 6)
-    static const BYTE* g_start = NULL;
-    U32 const pos = (U32)((const BYTE*)literals - g_start);
-    if (g_start==NULL) g_start = (const BYTE*)literals;
-    if ((pos > 0) && (pos < 1000000000))
-        DEBUGLOG(6, "Cpos %6u :%5u literals & match %3u bytes at distance %6u",
-               pos, (U32)litLength, (U32)matchCode+MINMATCH, (U32)offsetCode);
-#endif
-    /* copy Literals */
-    assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + 128 KB);
-    ZSTD_wildcopy(seqStorePtr->lit, literals, litLength);
-    seqStorePtr->lit += litLength;
-
-    /* literal Length */
-    if (litLength>0xFFFF) {
-        seqStorePtr->longLengthID = 1;
-        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
-    }
-    seqStorePtr->sequences[0].litLength = (U16)litLength;
-
-    /* match offset */
-    seqStorePtr->sequences[0].offset = offsetCode + 1;
-
-    /* match Length */
-    if (matchCode>0xFFFF) {
-        seqStorePtr->longLengthID = 2;
-        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
-    }
-    seqStorePtr->sequences[0].matchLength = (U16)matchCode;
-
-    seqStorePtr->sequences++;
-}
-
-
-/*-*************************************
-*  Match length counter
-***************************************/
-static unsigned ZSTD_NbCommonBytes (register size_t val)
-{
-    if (MEM_isLittleEndian()) {
-        if (MEM_64bits()) {
-#       if defined(_MSC_VER) && defined(_WIN64)
-            unsigned long r = 0;
-            _BitScanForward64( &r, (U64)val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 4)
-            return (__builtin_ctzll((U64)val) >> 3);
-#       else
-            static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2,
-                                                     0, 3, 1, 3, 1, 4, 2, 7,
-                                                     0, 2, 3, 6, 1, 5, 3, 5,
-                                                     1, 3, 4, 4, 2, 5, 6, 7,
-                                                     7, 0, 1, 2, 3, 3, 4, 6,
-                                                     2, 6, 5, 5, 3, 4, 5, 6,
-                                                     7, 1, 2, 4, 6, 4, 4, 5,
-                                                     7, 2, 6, 5, 7, 6, 7, 7 };
-            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
-#       endif
-        } else { /* 32 bits */
-#       if defined(_MSC_VER)
-            unsigned long r=0;
-            _BitScanForward( &r, (U32)val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 3)
-            return (__builtin_ctz((U32)val) >> 3);
-#       else
-            static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0,
-                                                     3, 2, 2, 1, 3, 2, 0, 1,
-                                                     3, 3, 1, 2, 2, 2, 2, 0,
-                                                     3, 1, 2, 0, 1, 0, 1, 1 };
-            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
-#       endif
-        }
-    } else {  /* Big Endian CPU */
-        if (MEM_64bits()) {
-#       if defined(_MSC_VER) && defined(_WIN64)
-            unsigned long r = 0;
-            _BitScanReverse64( &r, val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 4)
-            return (__builtin_clzll(val) >> 3);
-#       else
-            unsigned r;
-            const unsigned n32 = sizeof(size_t)*4;   /* calculate this way due to compiler complaining in 32-bits mode */
-            if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
-            if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
-            r += (!val);
-            return r;
-#       endif
-        } else { /* 32 bits */
-#       if defined(_MSC_VER)
-            unsigned long r = 0;
-            _BitScanReverse( &r, (unsigned long)val );
-            return (unsigned)(r>>3);
-#       elif defined(__GNUC__) && (__GNUC__ >= 3)
-            return (__builtin_clz((U32)val) >> 3);
-#       else
-            unsigned r;
-            if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
-            r += (!val);
-            return r;
-#       endif
-    }   }
-}
-
-
-MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* const pInLimit)
-{
-    const BYTE* const pStart = pIn;
-    const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
-
-    while (pIn < pInLoopLimit) {
-        size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
-        if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
-        pIn += ZSTD_NbCommonBytes(diff);
-        return (size_t)(pIn - pStart);
-    }
-    if (MEM_64bits()) if ((pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
-    if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
-    if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
-    return (size_t)(pIn - pStart);
-}
-
-/** ZSTD_count_2segments() :
-*   can count match length with `ip` & `match` in 2 different segments.
-*   convention : on reaching mEnd, match count continue starting from iStart
-*/
-MEM_STATIC size_t ZSTD_count_2segments(const BYTE* ip, const BYTE* match, const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
-{
-    const BYTE* const vEnd = MIN( ip + (mEnd - match), iEnd);
-    size_t const matchLength = ZSTD_count(ip, match, vEnd);
-    if (match + matchLength != mEnd) return matchLength;
-    return matchLength + ZSTD_count(ip+matchLength, iStart, iEnd);
-}
-
-
-/*-*************************************
-*  Hashes
-***************************************/
-static const U32 prime3bytes = 506832829U;
-static U32    ZSTD_hash3(U32 u, U32 h) { return ((u << (32-24)) * prime3bytes)  >> (32-h) ; }
-MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); } /* only in zstd_opt.h */
-
-static const U32 prime4bytes = 2654435761U;
-static U32    ZSTD_hash4(U32 u, U32 h) { return (u * prime4bytes) >> (32-h) ; }
-static size_t ZSTD_hash4Ptr(const void* ptr, U32 h) { return ZSTD_hash4(MEM_read32(ptr), h); }
-
-static const U64 prime5bytes = 889523592379ULL;
-static size_t ZSTD_hash5(U64 u, U32 h) { return (size_t)(((u  << (64-40)) * prime5bytes) >> (64-h)) ; }
-static size_t ZSTD_hash5Ptr(const void* p, U32 h) { return ZSTD_hash5(MEM_readLE64(p), h); }
-
-static const U64 prime6bytes = 227718039650203ULL;
-static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u  << (64-48)) * prime6bytes) >> (64-h)) ; }
-static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
-
-static const U64 prime7bytes = 58295818150454627ULL;
-static size_t ZSTD_hash7(U64 u, U32 h) { return (size_t)(((u  << (64-56)) * prime7bytes) >> (64-h)) ; }
-static size_t ZSTD_hash7Ptr(const void* p, U32 h) { return ZSTD_hash7(MEM_readLE64(p), h); }
-
-static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
-static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
-static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
-
-MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
-{
-    switch(mls)
-    {
-    default:
-    case 4: return ZSTD_hash4Ptr(p, hBits);
-    case 5: return ZSTD_hash5Ptr(p, hBits);
-    case 6: return ZSTD_hash6Ptr(p, hBits);
-    case 7: return ZSTD_hash7Ptr(p, hBits);
-    case 8: return ZSTD_hash8Ptr(p, hBits);
-    }
-}
-
-#if defined (__cplusplus)
-}
-#endif
-
-#endif /* ZSTD_COMPRESS_H */

src/borg/algorithms/zstd/lib/compress/zstd_compress_internal.h

@@ -0,0 +1,709 @@
+/*
+ * 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.
+ */
+
+/* This header contains definitions
+ * that shall **only** be used by modules within lib/compress.
+ */
+
+#ifndef ZSTD_COMPRESS_H
+#define ZSTD_COMPRESS_H
+
+/*-*************************************
+*  Dependencies
+***************************************/
+#include "zstd_internal.h"
+#ifdef ZSTD_MULTITHREAD
+#  include "zstdmt_compress.h"
+#endif
+
+#if defined (__cplusplus)
+extern "C" {
+#endif
+
+/*-*************************************
+*  Constants
+***************************************/
+#define kSearchStrength      8
+#define HASH_READ_SIZE       8
+#define ZSTD_DUBT_UNSORTED_MARK 1   /* For btlazy2 strategy, index 1 now means "unsorted".
+                                       It could be confused for a real successor at index "1", if sorted as larger than its predecessor.
+                                       It's not a big deal though : candidate will just be sorted again.
+                                       Additionnally, candidate position 1 will be lost.
+                                       But candidate 1 cannot hide a large tree of candidates, so it's a minimal loss.
+                                       The benefit is that ZSTD_DUBT_UNSORTED_MARK cannot be misdhandled after table re-use with a different strategy */
+
+
+/*-*************************************
+*  Context memory management
+***************************************/
+typedef enum { ZSTDcs_created=0, ZSTDcs_init, ZSTDcs_ongoing, ZSTDcs_ending } ZSTD_compressionStage_e;
+typedef enum { zcss_init=0, zcss_load, zcss_flush } ZSTD_cStreamStage;
+
+typedef struct ZSTD_prefixDict_s {
+    const void* dict;
+    size_t dictSize;
+    ZSTD_dictContentType_e dictContentType;
+} ZSTD_prefixDict;
+
+typedef struct {
+    U32 hufCTable[HUF_CTABLE_SIZE_U32(255)];
+    FSE_CTable offcodeCTable[FSE_CTABLE_SIZE_U32(OffFSELog, MaxOff)];
+    FSE_CTable matchlengthCTable[FSE_CTABLE_SIZE_U32(MLFSELog, MaxML)];
+    FSE_CTable litlengthCTable[FSE_CTABLE_SIZE_U32(LLFSELog, MaxLL)];
+    HUF_repeat hufCTable_repeatMode;
+    FSE_repeat offcode_repeatMode;
+    FSE_repeat matchlength_repeatMode;
+    FSE_repeat litlength_repeatMode;
+} ZSTD_entropyCTables_t;
+
+typedef struct {
+    U32 off;
+    U32 len;
+} ZSTD_match_t;
+
+typedef struct {
+    int price;
+    U32 off;
+    U32 mlen;
+    U32 litlen;
+    U32 rep[ZSTD_REP_NUM];
+} ZSTD_optimal_t;
+
+typedef struct {
+    /* All tables are allocated inside cctx->workspace by ZSTD_resetCCtx_internal() */
+    U32* litFreq;               /* table of literals statistics, of size 256 */
+    U32* litLengthFreq;         /* table of litLength statistics, of size (MaxLL+1) */
+    U32* matchLengthFreq;       /* table of matchLength statistics, of size (MaxML+1) */
+    U32* offCodeFreq;           /* table of offCode statistics, of size (MaxOff+1) */
+    ZSTD_match_t* matchTable;   /* list of found matches, of size ZSTD_OPT_NUM+1 */
+    ZSTD_optimal_t* priceTable; /* All positions tracked by optimal parser, of size ZSTD_OPT_NUM+1 */
+
+    U32  litSum;                 /* nb of literals */
+    U32  litLengthSum;           /* nb of litLength codes */
+    U32  matchLengthSum;         /* nb of matchLength codes */
+    U32  offCodeSum;             /* nb of offset codes */
+    /* begin updated by ZSTD_setLog2Prices */
+    U32  log2litSum;             /* pow2 to compare log2(litfreq) to */
+    U32  log2litLengthSum;       /* pow2 to compare log2(llfreq) to */
+    U32  log2matchLengthSum;     /* pow2 to compare log2(mlfreq) to */
+    U32  log2offCodeSum;         /* pow2 to compare log2(offreq) to */
+    /* end : updated by ZSTD_setLog2Prices */
+    U32  staticPrices;           /* prices follow a pre-defined cost structure, statistics are irrelevant */
+} optState_t;
+
+typedef struct {
+  ZSTD_entropyCTables_t entropy;
+  U32 rep[ZSTD_REP_NUM];
+} ZSTD_compressedBlockState_t;
+
+typedef struct {
+    BYTE const* nextSrc;    /* next block here to continue on current prefix */
+    BYTE const* base;       /* All regular indexes relative to this position */
+    BYTE const* dictBase;   /* extDict indexes relative to this position */
+    U32 dictLimit;          /* below that point, need extDict */
+    U32 lowLimit;           /* below that point, no more data */
+} ZSTD_window_t;
+
+typedef struct {
+    ZSTD_window_t window;      /* State for window round buffer management */
+    U32 loadedDictEnd;         /* index of end of dictionary */
+    U32 nextToUpdate;          /* index from which to continue table update */
+    U32 nextToUpdate3;         /* index from which to continue table update */
+    U32 hashLog3;              /* dispatch table : larger == faster, more memory */
+    U32* hashTable;
+    U32* hashTable3;
+    U32* chainTable;
+    optState_t opt;         /* optimal parser state */
+} ZSTD_matchState_t;
+
+typedef struct {
+    ZSTD_compressedBlockState_t* prevCBlock;
+    ZSTD_compressedBlockState_t* nextCBlock;
+    ZSTD_matchState_t matchState;
+} ZSTD_blockState_t;
+
+typedef struct {
+    U32 offset;
+    U32 checksum;
+} ldmEntry_t;
+
+typedef struct {
+    ZSTD_window_t window;   /* State for the window round buffer management */
+    ldmEntry_t* hashTable;
+    BYTE* bucketOffsets;    /* Next position in bucket to insert entry */
+    U64 hashPower;          /* Used to compute the rolling hash.
+                             * Depends on ldmParams.minMatchLength */
+} ldmState_t;
+
+typedef struct {
+    U32 enableLdm;          /* 1 if enable long distance matching */
+    U32 hashLog;            /* Log size of hashTable */
+    U32 bucketSizeLog;      /* Log bucket size for collision resolution, at most 8 */
+    U32 minMatchLength;     /* Minimum match length */
+    U32 hashEveryLog;       /* Log number of entries to skip */
+    U32 windowLog;          /* Window log for the LDM */
+} ldmParams_t;
+
+typedef struct {
+    U32 offset;
+    U32 litLength;
+    U32 matchLength;
+} rawSeq;
+
+typedef struct {
+  rawSeq* seq;     /* The start of the sequences */
+  size_t pos;      /* The position where reading stopped. <= size. */
+  size_t size;     /* The number of sequences. <= capacity. */
+  size_t capacity; /* The capacity of the `seq` pointer */
+} rawSeqStore_t;
+
+struct ZSTD_CCtx_params_s {
+    ZSTD_format_e format;
+    ZSTD_compressionParameters cParams;
+    ZSTD_frameParameters fParams;
+
+    int compressionLevel;
+    int disableLiteralCompression;
+    int forceWindow;           /* force back-references to respect limit of
+                                * 1<<wLog, even for dictionary */
+
+    /* Multithreading: used to pass parameters to mtctx */
+    unsigned nbWorkers;
+    unsigned jobSize;
+    unsigned overlapSizeLog;
+
+    /* Long distance matching parameters */
+    ldmParams_t ldmParams;
+
+    /* Internal use, for createCCtxParams() and freeCCtxParams() only */
+    ZSTD_customMem customMem;
+};  /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
+
+struct ZSTD_CCtx_s {
+    ZSTD_compressionStage_e stage;
+    int cParamsChanged;                  /* == 1 if cParams(except wlog) or compression level are changed in requestedParams. Triggers transmission of new params to ZSTDMT (if available) then reset to 0. */
+    int bmi2;                            /* == 1 if the CPU supports BMI2 and 0 otherwise. CPU support is determined dynamically once per context lifetime. */
+    ZSTD_CCtx_params requestedParams;
+    ZSTD_CCtx_params appliedParams;
+    U32   dictID;
+    void* workSpace;
+    size_t workSpaceSize;
+    size_t blockSize;
+    unsigned long long pledgedSrcSizePlusOne;  /* this way, 0 (default) == unknown */
+    unsigned long long consumedSrcSize;
+    unsigned long long producedCSize;
+    XXH64_state_t xxhState;
+    ZSTD_customMem customMem;
+    size_t staticSize;
+
+    seqStore_t seqStore;      /* sequences storage ptrs */
+    ldmState_t ldmState;      /* long distance matching state */
+    rawSeq* ldmSequences;     /* Storage for the ldm output sequences */
+    size_t maxNbLdmSequences;
+    rawSeqStore_t externSeqStore; /* Mutable reference to external sequences */
+    ZSTD_blockState_t blockState;
+    U32* entropyWorkspace;  /* entropy workspace of HUF_WORKSPACE_SIZE bytes */
+
+    /* streaming */
+    char*  inBuff;
+    size_t inBuffSize;
+    size_t inToCompress;
+    size_t inBuffPos;
+    size_t inBuffTarget;
+    char*  outBuff;
+    size_t outBuffSize;
+    size_t outBuffContentSize;
+    size_t outBuffFlushedSize;
+    ZSTD_cStreamStage streamStage;
+    U32    frameEnded;
+
+    /* Dictionary */
+    ZSTD_CDict* cdictLocal;
+    const ZSTD_CDict* cdict;
+    ZSTD_prefixDict prefixDict;   /* single-usage dictionary */
+
+    /* Multi-threading */
+#ifdef ZSTD_MULTITHREAD
+    ZSTDMT_CCtx* mtctx;
+#endif
+};
+
+
+typedef size_t (*ZSTD_blockCompressor) (
+        ZSTD_matchState_t* bs, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict);
+
+
+MEM_STATIC U32 ZSTD_LLcode(U32 litLength)
+{
+    static const BYTE LL_Code[64] = {  0,  1,  2,  3,  4,  5,  6,  7,
+                                       8,  9, 10, 11, 12, 13, 14, 15,
+                                      16, 16, 17, 17, 18, 18, 19, 19,
+                                      20, 20, 20, 20, 21, 21, 21, 21,
+                                      22, 22, 22, 22, 22, 22, 22, 22,
+                                      23, 23, 23, 23, 23, 23, 23, 23,
+                                      24, 24, 24, 24, 24, 24, 24, 24,
+                                      24, 24, 24, 24, 24, 24, 24, 24 };
+    static const U32 LL_deltaCode = 19;
+    return (litLength > 63) ? ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+}
+
+/* ZSTD_MLcode() :
+ * note : mlBase = matchLength - MINMATCH;
+ *        because it's the format it's stored in seqStore->sequences */
+MEM_STATIC U32 ZSTD_MLcode(U32 mlBase)
+{
+    static const BYTE ML_Code[128] = { 0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15,
+                                      16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31,
+                                      32, 32, 33, 33, 34, 34, 35, 35, 36, 36, 36, 36, 37, 37, 37, 37,
+                                      38, 38, 38, 38, 38, 38, 38, 38, 39, 39, 39, 39, 39, 39, 39, 39,
+                                      40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40,
+                                      41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41, 41,
+                                      42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42,
+                                      42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42, 42 };
+    static const U32 ML_deltaCode = 36;
+    return (mlBase > 127) ? ZSTD_highbit32(mlBase) + ML_deltaCode : ML_Code[mlBase];
+}
+
+/*! 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).
+ *  `mlBase` : matchLength - MINMATCH
+*/
+MEM_STATIC void ZSTD_storeSeq(seqStore_t* seqStorePtr, size_t litLength, const void* literals, U32 offsetCode, size_t mlBase)
+{
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG >= 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%3u bytes at dist.code%7u",
+               pos, (U32)litLength, (U32)mlBase+MINMATCH, (U32)offsetCode);
+    }
+#endif
+    /* copy Literals */
+    assert(seqStorePtr->lit + litLength <= seqStorePtr->litStart + 128 KB);
+    ZSTD_wildcopy(seqStorePtr->lit, literals, litLength);
+    seqStorePtr->lit += litLength;
+
+    /* literal Length */
+    if (litLength>0xFFFF) {
+        assert(seqStorePtr->longLengthID == 0); /* there can only be a single long length */
+        seqStorePtr->longLengthID = 1;
+        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+    }
+    seqStorePtr->sequences[0].litLength = (U16)litLength;
+
+    /* match offset */
+    seqStorePtr->sequences[0].offset = offsetCode + 1;
+
+    /* match Length */
+    if (mlBase>0xFFFF) {
+        assert(seqStorePtr->longLengthID == 0); /* there can only be a single long length */
+        seqStorePtr->longLengthID = 2;
+        seqStorePtr->longLengthPos = (U32)(seqStorePtr->sequences - seqStorePtr->sequencesStart);
+    }
+    seqStorePtr->sequences[0].matchLength = (U16)mlBase;
+
+    seqStorePtr->sequences++;
+}
+
+
+/*-*************************************
+*  Match length counter
+***************************************/
+static unsigned ZSTD_NbCommonBytes (size_t val)
+{
+    if (MEM_isLittleEndian()) {
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanForward64( &r, (U64)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 4)
+            return (__builtin_ctzll((U64)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[64] = { 0, 0, 0, 0, 0, 1, 1, 2,
+                                                     0, 3, 1, 3, 1, 4, 2, 7,
+                                                     0, 2, 3, 6, 1, 5, 3, 5,
+                                                     1, 3, 4, 4, 2, 5, 6, 7,
+                                                     7, 0, 1, 2, 3, 3, 4, 6,
+                                                     2, 6, 5, 5, 3, 4, 5, 6,
+                                                     7, 1, 2, 4, 6, 4, 4, 5,
+                                                     7, 2, 6, 5, 7, 6, 7, 7 };
+            return DeBruijnBytePos[((U64)((val & -(long long)val) * 0x0218A392CDABBD3FULL)) >> 58];
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r=0;
+            _BitScanForward( &r, (U32)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_ctz((U32)val) >> 3);
+#       else
+            static const int DeBruijnBytePos[32] = { 0, 0, 3, 0, 3, 1, 3, 0,
+                                                     3, 2, 2, 1, 3, 2, 0, 1,
+                                                     3, 3, 1, 2, 2, 2, 2, 0,
+                                                     3, 1, 2, 0, 1, 0, 1, 1 };
+            return DeBruijnBytePos[((U32)((val & -(S32)val) * 0x077CB531U)) >> 27];
+#       endif
+        }
+    } else {  /* Big Endian CPU */
+        if (MEM_64bits()) {
+#       if defined(_MSC_VER) && defined(_WIN64)
+            unsigned long r = 0;
+            _BitScanReverse64( &r, val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 4)
+            return (__builtin_clzll(val) >> 3);
+#       else
+            unsigned r;
+            const unsigned n32 = sizeof(size_t)*4;   /* calculate this way due to compiler complaining in 32-bits mode */
+            if (!(val>>n32)) { r=4; } else { r=0; val>>=n32; }
+            if (!(val>>16)) { r+=2; val>>=8; } else { val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+        } else { /* 32 bits */
+#       if defined(_MSC_VER)
+            unsigned long r = 0;
+            _BitScanReverse( &r, (unsigned long)val );
+            return (unsigned)(r>>3);
+#       elif defined(__GNUC__) && (__GNUC__ >= 3)
+            return (__builtin_clz((U32)val) >> 3);
+#       else
+            unsigned r;
+            if (!(val>>16)) { r=2; val>>=8; } else { r=0; val>>=24; }
+            r += (!val);
+            return r;
+#       endif
+    }   }
+}
+
+
+MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* const pInLimit)
+{
+    const BYTE* const pStart = pIn;
+    const BYTE* const pInLoopLimit = pInLimit - (sizeof(size_t)-1);
+
+    if (pIn < pInLoopLimit) {
+        { size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+          if (diff) return ZSTD_NbCommonBytes(diff); }
+        pIn+=sizeof(size_t); pMatch+=sizeof(size_t);
+        while (pIn < pInLoopLimit) {
+            size_t const diff = MEM_readST(pMatch) ^ MEM_readST(pIn);
+            if (!diff) { pIn+=sizeof(size_t); pMatch+=sizeof(size_t); continue; }
+            pIn += ZSTD_NbCommonBytes(diff);
+            return (size_t)(pIn - pStart);
+    }   }
+    if (MEM_64bits() && (pIn<(pInLimit-3)) && (MEM_read32(pMatch) == MEM_read32(pIn))) { pIn+=4; pMatch+=4; }
+    if ((pIn<(pInLimit-1)) && (MEM_read16(pMatch) == MEM_read16(pIn))) { pIn+=2; pMatch+=2; }
+    if ((pIn<pInLimit) && (*pMatch == *pIn)) pIn++;
+    return (size_t)(pIn - pStart);
+}
+
+/** ZSTD_count_2segments() :
+ *  can count match length with `ip` & `match` in 2 different segments.
+ *  convention : on reaching mEnd, match count continue starting from iStart
+ */
+MEM_STATIC size_t
+ZSTD_count_2segments(const BYTE* ip, const BYTE* match,
+                     const BYTE* iEnd, const BYTE* mEnd, const BYTE* iStart)
+{
+    const BYTE* const vEnd = MIN( ip + (mEnd - match), iEnd);
+    size_t const matchLength = ZSTD_count(ip, match, vEnd);
+    if (match + matchLength != mEnd) return matchLength;
+    return matchLength + ZSTD_count(ip+matchLength, iStart, iEnd);
+}
+
+
+/*-*************************************
+ *  Hashes
+ ***************************************/
+static const U32 prime3bytes = 506832829U;
+static U32    ZSTD_hash3(U32 u, U32 h) { return ((u << (32-24)) * prime3bytes)  >> (32-h) ; }
+MEM_STATIC size_t ZSTD_hash3Ptr(const void* ptr, U32 h) { return ZSTD_hash3(MEM_readLE32(ptr), h); } /* only in zstd_opt.h */
+
+static const U32 prime4bytes = 2654435761U;
+static U32    ZSTD_hash4(U32 u, U32 h) { return (u * prime4bytes) >> (32-h) ; }
+static size_t ZSTD_hash4Ptr(const void* ptr, U32 h) { return ZSTD_hash4(MEM_read32(ptr), h); }
+
+static const U64 prime5bytes = 889523592379ULL;
+static size_t ZSTD_hash5(U64 u, U32 h) { return (size_t)(((u  << (64-40)) * prime5bytes) >> (64-h)) ; }
+static size_t ZSTD_hash5Ptr(const void* p, U32 h) { return ZSTD_hash5(MEM_readLE64(p), h); }
+
+static const U64 prime6bytes = 227718039650203ULL;
+static size_t ZSTD_hash6(U64 u, U32 h) { return (size_t)(((u  << (64-48)) * prime6bytes) >> (64-h)) ; }
+static size_t ZSTD_hash6Ptr(const void* p, U32 h) { return ZSTD_hash6(MEM_readLE64(p), h); }
+
+static const U64 prime7bytes = 58295818150454627ULL;
+static size_t ZSTD_hash7(U64 u, U32 h) { return (size_t)(((u  << (64-56)) * prime7bytes) >> (64-h)) ; }
+static size_t ZSTD_hash7Ptr(const void* p, U32 h) { return ZSTD_hash7(MEM_readLE64(p), h); }
+
+static const U64 prime8bytes = 0xCF1BBCDCB7A56463ULL;
+static size_t ZSTD_hash8(U64 u, U32 h) { return (size_t)(((u) * prime8bytes) >> (64-h)) ; }
+static size_t ZSTD_hash8Ptr(const void* p, U32 h) { return ZSTD_hash8(MEM_readLE64(p), h); }
+
+MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
+{
+    switch(mls)
+    {
+    default:
+    case 4: return ZSTD_hash4Ptr(p, hBits);
+    case 5: return ZSTD_hash5Ptr(p, hBits);
+    case 6: return ZSTD_hash6Ptr(p, hBits);
+    case 7: return ZSTD_hash7Ptr(p, hBits);
+    case 8: return ZSTD_hash8Ptr(p, hBits);
+    }
+}
+
+/*-*************************************
+*  Round buffer management
+***************************************/
+/* Max current allowed */
+#define ZSTD_CURRENT_MAX ((3U << 29) + (1U << ZSTD_WINDOWLOG_MAX))
+/* Maximum chunk size before overflow correction needs to be called again */
+#define ZSTD_CHUNKSIZE_MAX                                                     \
+    ( ((U32)-1)                  /* Maximum ending current index */            \
+    - ZSTD_CURRENT_MAX)          /* Maximum beginning lowLimit */
+
+/**
+ * ZSTD_window_clear():
+ * Clears the window containing the history by simply setting it to empty.
+ */
+MEM_STATIC void ZSTD_window_clear(ZSTD_window_t* window)
+{
+    size_t const endT = (size_t)(window->nextSrc - window->base);
+    U32 const end = (U32)endT;
+
+    window->lowLimit = end;
+    window->dictLimit = end;
+}
+
+/**
+ * ZSTD_window_hasExtDict():
+ * Returns non-zero if the window has a non-empty extDict.
+ */
+MEM_STATIC U32 ZSTD_window_hasExtDict(ZSTD_window_t const window)
+{
+    return window.lowLimit < window.dictLimit;
+}
+
+/**
+ * ZSTD_window_needOverflowCorrection():
+ * Returns non-zero if the indices are getting too large and need overflow
+ * protection.
+ */
+MEM_STATIC U32 ZSTD_window_needOverflowCorrection(ZSTD_window_t const window,
+                                                  void const* srcEnd)
+{
+    U32 const current = (U32)((BYTE const*)srcEnd - window.base);
+    return current > ZSTD_CURRENT_MAX;
+}
+
+/**
+ * ZSTD_window_correctOverflow():
+ * Reduces the indices to protect from index overflow.
+ * Returns the correction made to the indices, which must be applied to every
+ * stored index.
+ *
+ * The least significant cycleLog bits of the indices must remain the same,
+ * which may be 0. Every index up to maxDist in the past must be valid.
+ * NOTE: (maxDist & cycleMask) must be zero.
+ */
+MEM_STATIC U32 ZSTD_window_correctOverflow(ZSTD_window_t* window, U32 cycleLog,
+                                           U32 maxDist, void const* src)
+{
+    /* preemptive overflow correction:
+     * 1. correction is large enough:
+     *    lowLimit > (3<<29) ==> current > 3<<29 + 1<<windowLog
+     *    1<<windowLog <= newCurrent < 1<<chainLog + 1<<windowLog
+     *
+     *    current - newCurrent
+     *    > (3<<29 + 1<<windowLog) - (1<<windowLog + 1<<chainLog)
+     *    > (3<<29) - (1<<chainLog)
+     *    > (3<<29) - (1<<30)             (NOTE: chainLog <= 30)
+     *    > 1<<29
+     *
+     * 2. (ip+ZSTD_CHUNKSIZE_MAX - cctx->base) doesn't overflow:
+     *    After correction, current is less than (1<<chainLog + 1<<windowLog).
+     *    In 64-bit mode we are safe, because we have 64-bit ptrdiff_t.
+     *    In 32-bit mode we are safe, because (chainLog <= 29), so
+     *    ip+ZSTD_CHUNKSIZE_MAX - cctx->base < 1<<32.
+     * 3. (cctx->lowLimit + 1<<windowLog) < 1<<32:
+     *    windowLog <= 31 ==> 3<<29 + 1<<windowLog < 7<<29 < 1<<32.
+     */
+    U32 const cycleMask = (1U << cycleLog) - 1;
+    U32 const current = (U32)((BYTE const*)src - window->base);
+    U32 const newCurrent = (current & cycleMask) + maxDist;
+    U32 const correction = current - newCurrent;
+    assert((maxDist & cycleMask) == 0);
+    assert(current > newCurrent);
+    /* Loose bound, should be around 1<<29 (see above) */
+    assert(correction > 1<<28);
+
+    window->base += correction;
+    window->dictBase += correction;
+    window->lowLimit -= correction;
+    window->dictLimit -= correction;
+
+    DEBUGLOG(4, "Correction of 0x%x bytes to lowLimit=0x%x", correction,
+             window->lowLimit);
+    return correction;
+}
+
+/**
+ * ZSTD_window_enforceMaxDist():
+ * Updates lowLimit so that:
+ *    (srcEnd - base) - lowLimit == maxDist + loadedDictEnd
+ * This allows a simple check that index >= lowLimit to see if index is valid.
+ * This must be called before a block compression call, with srcEnd as the block
+ * source end.
+ * If loadedDictEndPtr is not NULL, we set it to zero once we update lowLimit.
+ * This is because dictionaries are allowed to be referenced as long as the last
+ * byte of the dictionary is in the window, but once they are out of range,
+ * they cannot be referenced. If loadedDictEndPtr is NULL, we use
+ * loadedDictEnd == 0.
+ */
+MEM_STATIC void ZSTD_window_enforceMaxDist(ZSTD_window_t* window,
+                                           void const* srcEnd, U32 maxDist,
+                                           U32* loadedDictEndPtr)
+{
+    U32 const current = (U32)((BYTE const*)srcEnd - window->base);
+    U32 loadedDictEnd = loadedDictEndPtr != NULL ? *loadedDictEndPtr : 0;
+    if (current > maxDist + loadedDictEnd) {
+        U32 const newLowLimit = current - maxDist;
+        if (window->lowLimit < newLowLimit) window->lowLimit = newLowLimit;
+        if (window->dictLimit < window->lowLimit) {
+            DEBUGLOG(5, "Update dictLimit from %u to %u", window->dictLimit,
+                     window->lowLimit);
+            window->dictLimit = window->lowLimit;
+        }
+        if (loadedDictEndPtr)
+            *loadedDictEndPtr = 0;
+    }
+}
+
+/**
+ * ZSTD_window_update():
+ * Updates the window by appending [src, src + srcSize) to the window.
+ * If it is not contiguous, the current prefix becomes the extDict, and we
+ * forget about the extDict. Handles overlap of the prefix and extDict.
+ * Returns non-zero if the segment is contiguous.
+ */
+MEM_STATIC U32 ZSTD_window_update(ZSTD_window_t* window,
+                                  void const* src, size_t srcSize)
+{
+    BYTE const* const ip = (BYTE const*)src;
+    U32 contiguous = 1;
+    /* Check if blocks follow each other */
+    if (src != window->nextSrc) {
+        /* not contiguous */
+        size_t const distanceFromBase = (size_t)(window->nextSrc - window->base);
+        DEBUGLOG(5, "Non contiguous blocks, new segment starts at %u",
+                 window->dictLimit);
+        window->lowLimit = window->dictLimit;
+        assert(distanceFromBase == (size_t)(U32)distanceFromBase);  /* should never overflow */
+        window->dictLimit = (U32)distanceFromBase;
+        window->dictBase = window->base;
+        window->base = ip - distanceFromBase;
+        // ms->nextToUpdate = window->dictLimit;
+        if (window->dictLimit - window->lowLimit < HASH_READ_SIZE) window->lowLimit = window->dictLimit;   /* too small extDict */
+        contiguous = 0;
+    }
+    window->nextSrc = ip + srcSize;
+    /* if input and dictionary overlap : reduce dictionary (area presumed modified by input) */
+    if ( (ip+srcSize > window->dictBase + window->lowLimit)
+       & (ip < window->dictBase + window->dictLimit)) {
+        ptrdiff_t const highInputIdx = (ip + srcSize) - window->dictBase;
+        U32 const lowLimitMax = (highInputIdx > (ptrdiff_t)window->dictLimit) ? window->dictLimit : (U32)highInputIdx;
+        window->lowLimit = lowLimitMax;
+    }
+    return contiguous;
+}
+
+#if defined (__cplusplus)
+}
+#endif
+
+
+/* ==============================================================
+ * Private declarations
+ * These prototypes shall only be called from within lib/compress
+ * ============================================================== */
+
+/* ZSTD_getCParamsFromCCtxParams() :
+ * cParams are built depending on compressionLevel, src size hints, 
+ * LDM and manually set compression parameters.
+ */
+ZSTD_compressionParameters ZSTD_getCParamsFromCCtxParams(
+        const ZSTD_CCtx_params* CCtxParams, U64 srcSizeHint, size_t dictSize);
+
+/*! ZSTD_initCStream_internal() :
+ *  Private use only. Init streaming operation.
+ *  expects params to be valid.
+ *  must receive dict, or cdict, or none, but not both.
+ *  @return : 0, or an error code */
+size_t ZSTD_initCStream_internal(ZSTD_CStream* zcs,
+                     const void* dict, size_t dictSize,
+                     const ZSTD_CDict* cdict,
+                     ZSTD_CCtx_params  params, unsigned long long pledgedSrcSize);
+
+/*! ZSTD_compressStream_generic() :
+ *  Private use only. To be called from zstdmt_compress.c in single-thread mode. */
+size_t ZSTD_compressStream_generic(ZSTD_CStream* zcs,
+                                   ZSTD_outBuffer* output,
+                                   ZSTD_inBuffer* input,
+                                   ZSTD_EndDirective const flushMode);
+
+/*! ZSTD_getCParamsFromCDict() :
+ *  as the name implies */
+ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict);
+
+/* ZSTD_compressBegin_advanced_internal() :
+ * Private use only. To be called from zstdmt_compress.c. */
+size_t ZSTD_compressBegin_advanced_internal(ZSTD_CCtx* cctx,
+                                    const void* dict, size_t dictSize,
+                                    ZSTD_dictContentType_e dictContentType,
+                                    const ZSTD_CDict* cdict,
+                                    ZSTD_CCtx_params params,
+                                    unsigned long long pledgedSrcSize);
+
+/* ZSTD_compress_advanced_internal() :
+ * Private use only. To be called from zstdmt_compress.c. */
+size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
+                                       void* dst, size_t dstCapacity,
+                                 const void* src, size_t srcSize,
+                                 const void* dict,size_t dictSize,
+                                 ZSTD_CCtx_params params);
+
+
+/* ZSTD_writeLastEmptyBlock() :
+ * output an empty Block with end-of-frame mark to complete a frame
+ * @return : size of data written into `dst` (== ZSTD_blockHeaderSize (defined in zstd_internal.h))
+ *           or an error code if `dstCapcity` is too small (<ZSTD_blockHeaderSize)
+ */
+size_t ZSTD_writeLastEmptyBlock(void* dst, size_t dstCapacity);
+
+
+/* ZSTD_referenceExternalSequences() :
+ * Must be called before starting a compression operation.
+ * seqs must parse a prefix of the source.
+ * This cannot be used when long range matching is enabled.
+ * Zstd will use these sequences, and pass the literals to a secondary block
+ * compressor.
+ * @return : An error code on failure.
+ * NOTE: seqs are not verified! Invalid sequences can cause out-of-bounds memory
+ * access and data corruption.
+ */
+size_t ZSTD_referenceExternalSequences(ZSTD_CCtx* cctx, rawSeq* seq, size_t nbSeq);
+
+
+#endif /* ZSTD_COMPRESS_H */

src/borg/algorithms/zstd/lib/compress/zstd_double_fast.c

@@ -8,47 +8,62 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
+#include "zstd_compress_internal.h"
 #include "zstd_double_fast.h"
 
 
-void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls)
+void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
+                              ZSTD_compressionParameters const* cParams,
+                              void const* end)
 {
-    U32* const hashLarge = cctx->hashTable;
-    U32  const hBitsL = cctx->appliedParams.cParams.hashLog;
-    U32* const hashSmall = cctx->chainTable;
-    U32  const hBitsS = cctx->appliedParams.cParams.chainLog;
-    const BYTE* const base = cctx->base;
-    const BYTE* ip = base + cctx->nextToUpdate;
+    U32* const hashLarge = ms->hashTable;
+    U32  const hBitsL = cParams->hashLog;
+    U32  const mls = cParams->searchLength;
+    U32* const hashSmall = ms->chainTable;
+    U32  const hBitsS = cParams->chainLog;
+    const BYTE* const base = ms->window.base;
+    const BYTE* ip = base + ms->nextToUpdate;
     const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
-    const size_t fastHashFillStep = 3;
-
-    while(ip <= iend) {
-        hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip - base);
-        hashLarge[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip - base);
-        ip += fastHashFillStep;
+    const U32 fastHashFillStep = 3;
+
+    /* Always insert every fastHashFillStep position into the hash tables.
+     * Insert the other positions into the large hash table if their entry
+     * is empty.
+     */
+    for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
+        U32 const current = (U32)(ip - base);
+        U32 i;
+        for (i = 0; i < fastHashFillStep; ++i) {
+            size_t const smHash = ZSTD_hashPtr(ip + i, hBitsS, mls);
+            size_t const lgHash = ZSTD_hashPtr(ip + i, hBitsL, 8);
+            if (i == 0)
+                hashSmall[smHash] = current + i;
+            if (i == 0 || hashLarge[lgHash] == 0)
+                hashLarge[lgHash] = current + i;
+        }
     }
 }
 
 
 FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
+size_t ZSTD_compressBlock_doubleFast_generic(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
+        U32 const mls /* template */)
 {
-    U32* const hashLong = cctx->hashTable;
-    const U32 hBitsL = cctx->appliedParams.cParams.hashLog;
-    U32* const hashSmall = cctx->chainTable;
-    const U32 hBitsS = cctx->appliedParams.cParams.chainLog;
-    seqStore_t* seqStorePtr = &(cctx->seqStore);
-    const BYTE* const base = cctx->base;
+    U32* const hashLong = ms->hashTable;
+    const U32 hBitsL = cParams->hashLog;
+    U32* const hashSmall = ms->chainTable;
+    const U32 hBitsS = cParams->chainLog;
+    const BYTE* const base = ms->window.base;
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
-    const U32 lowestIndex = cctx->dictLimit;
+    const U32 lowestIndex = ms->window.dictLimit;
     const BYTE* const lowest = base + lowestIndex;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - HASH_READ_SIZE;
-    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+    U32 offset_1=rep[0], offset_2=rep[1];
     U32 offsetSaved = 0;
 
     /* init */
@@ -75,7 +90,7 @@ size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
             /* favor repcode */
             mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
             ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
         } else {
             U32 offset;
             if ( (matchIndexL > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip)) ) {
@@ -98,14 +113,14 @@ size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
                     while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
                 }
             } else {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                ip += ((ip-anchor) >> kSearchStrength) + 1;
                 continue;
             }
 
             offset_2 = offset_1;
             offset_1 = offset;
 
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
         }
 
         /* match found */
@@ -128,61 +143,63 @@ size_t ZSTD_compressBlock_doubleFast_generic(ZSTD_CCtx* cctx,
                 { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; } /* swap offset_2 <=> offset_1 */
                 hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = (U32)(ip-base);
                 hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = (U32)(ip-base);
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH);
                 ip += rLength;
                 anchor = ip;
                 continue;   /* faster when present ... (?) */
     }   }   }
 
     /* save reps for next block */
-    seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
-    seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
+    rep[0] = offset_1 ? offset_1 : offsetSaved;
+    rep[1] = offset_2 ? offset_2 : offsetSaved;
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_doubleFast(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    const U32 mls = cParams->searchLength;
     switch(mls)
     {
     default: /* includes case 3 */
     case 4 :
-        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 4);
+        return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 4);
     case 5 :
-        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 5);
+        return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 5);
     case 6 :
-        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 6);
+        return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 6);
     case 7 :
-        return ZSTD_compressBlock_doubleFast_generic(ctx, src, srcSize, 7);
+        return ZSTD_compressBlock_doubleFast_generic(ms, seqStore, rep, cParams, src, srcSize, 7);
     }
 }
 
 
-static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
+static size_t ZSTD_compressBlock_doubleFast_extDict_generic(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
+        U32 const mls /* template */)
 {
-    U32* const hashLong = ctx->hashTable;
-    U32  const hBitsL = ctx->appliedParams.cParams.hashLog;
-    U32* const hashSmall = ctx->chainTable;
-    U32  const hBitsS = ctx->appliedParams.cParams.chainLog;
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const base = ctx->base;
-    const BYTE* const dictBase = ctx->dictBase;
+    U32* const hashLong = ms->hashTable;
+    U32  const hBitsL = cParams->hashLog;
+    U32* const hashSmall = ms->chainTable;
+    U32  const hBitsS = cParams->chainLog;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const dictBase = ms->window.dictBase;
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
-    const U32   lowestIndex = ctx->lowLimit;
+    const U32   lowestIndex = ms->window.lowLimit;
     const BYTE* const dictStart = dictBase + lowestIndex;
-    const U32   dictLimit = ctx->dictLimit;
+    const U32   dictLimit = ms->window.dictLimit;
     const BYTE* const lowPrefixPtr = base + dictLimit;
     const BYTE* const dictEnd = dictBase + dictLimit;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - 8;
-    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+    U32 offset_1=rep[0], offset_2=rep[1];
 
     /* Search Loop */
     while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
@@ -208,7 +225,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
             const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
             mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
             ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
         } else {
             if ((matchLongIndex > lowestIndex) && (MEM_read64(matchLong) == MEM_read64(ip))) {
                 const BYTE* matchEnd = matchLongIndex < dictLimit ? dictEnd : iend;
@@ -219,7 +236,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
                 while (((ip>anchor) & (matchLong>lowMatchPtr)) && (ip[-1] == matchLong[-1])) { ip--; matchLong--; mLength++; }   /* catch up */
                 offset_2 = offset_1;
                 offset_1 = offset;
-                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
 
             } else if ((matchIndex > lowestIndex) && (MEM_read32(match) == MEM_read32(ip))) {
                 size_t const h3 = ZSTD_hashPtr(ip+1, hBitsL, 8);
@@ -244,10 +261,10 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
                 }
                 offset_2 = offset_1;
                 offset_1 = offset;
-                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
 
             } else {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                ip += ((ip-anchor) >> kSearchStrength) + 1;
                 continue;
         }   }
 
@@ -271,7 +288,7 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
                     const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
                     size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
                     U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
+                    ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
                     hashSmall[ZSTD_hashPtr(ip, hBitsS, mls)] = current2;
                     hashLong[ZSTD_hashPtr(ip, hBitsL, 8)] = current2;
                     ip += repLength2;
@@ -282,27 +299,29 @@ static size_t ZSTD_compressBlock_doubleFast_extDict_generic(ZSTD_CCtx* ctx,
     }   }   }
 
     /* save reps for next block */
-    seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
+    rep[0] = offset_1;
+    rep[1] = offset_2;
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx,
-                         const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_doubleFast_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    U32 const mls = ctx->appliedParams.cParams.searchLength;
+    U32 const mls = cParams->searchLength;
     switch(mls)
     {
     default: /* includes case 3 */
     case 4 :
-        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 4);
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 4);
     case 5 :
-        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 5);
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 5);
     case 6 :
-        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 6);
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 6);
     case 7 :
-        return ZSTD_compressBlock_doubleFast_extDict_generic(ctx, src, srcSize, 7);
+        return ZSTD_compressBlock_doubleFast_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 7);
     }
 }

src/borg/algorithms/zstd/lib/compress/zstd_double_fast.h

@@ -11,15 +11,23 @@
 #ifndef ZSTD_DOUBLE_FAST_H
 #define ZSTD_DOUBLE_FAST_H
 
-#include "zstd_compress.h"
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
-void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls);
-size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+#include "mem.h"      /* U32 */
+#include "zstd_compress_internal.h"     /* ZSTD_CCtx, size_t */
+
+void ZSTD_fillDoubleHashTable(ZSTD_matchState_t* ms,
+                              ZSTD_compressionParameters const* cParams,
+                              void const* end);
+size_t ZSTD_compressBlock_doubleFast(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_doubleFast_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+
 
 #if defined (__cplusplus)
 }

src/borg/algorithms/zstd/lib/compress/zstd_fast.c

@@ -8,42 +8,52 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
+#include "zstd_compress_internal.h"
 #include "zstd_fast.h"
 
 
-void ZSTD_fillHashTable (ZSTD_CCtx* zc, const void* end, const U32 mls)
+void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
+                        ZSTD_compressionParameters const* cParams,
+                        void const* end)
 {
-    U32* const hashTable = zc->hashTable;
-    U32  const hBits = zc->appliedParams.cParams.hashLog;
-    const BYTE* const base = zc->base;
-    const BYTE* ip = base + zc->nextToUpdate;
+    U32* const hashTable = ms->hashTable;
+    U32  const hBits = cParams->hashLog;
+    U32  const mls = cParams->searchLength;
+    const BYTE* const base = ms->window.base;
+    const BYTE* ip = base + ms->nextToUpdate;
     const BYTE* const iend = ((const BYTE*)end) - HASH_READ_SIZE;
-    const size_t fastHashFillStep = 3;
+    const U32 fastHashFillStep = 3;
 
-    while(ip <= iend) {
-        hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip - base);
-        ip += fastHashFillStep;
+    /* Always insert every fastHashFillStep position into the hash table.
+     * Insert the other positions if their hash entry is empty.
+     */
+    for (; ip + fastHashFillStep - 1 <= iend; ip += fastHashFillStep) {
+        U32 const current = (U32)(ip - base);
+        U32 i;
+        for (i = 0; i < fastHashFillStep; ++i) {
+            size_t const hash = ZSTD_hashPtr(ip + i, hBits, mls);
+            if (i == 0 || hashTable[hash] == 0)
+                hashTable[hash] = current + i;
+        }
     }
 }
 
-
 FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
+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 hlog, U32 const stepSize, U32 const mls)
 {
-    U32* const hashTable = cctx->hashTable;
-    U32  const hBits = cctx->appliedParams.cParams.hashLog;
-    seqStore_t* seqStorePtr = &(cctx->seqStore);
-    const BYTE* const base = cctx->base;
+    U32* const hashTable = ms->hashTable;
+    const BYTE* const base = ms->window.base;
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
-    const U32   lowestIndex = cctx->dictLimit;
+    const U32   lowestIndex = ms->window.dictLimit;
     const BYTE* const lowest = base + lowestIndex;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - HASH_READ_SIZE;
-    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+    U32 offset_1=rep[0], offset_2=rep[1];
     U32 offsetSaved = 0;
 
     /* init */
@@ -56,7 +66,7 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
     /* Main Search Loop */
     while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
         size_t mLength;
-        size_t const h = ZSTD_hashPtr(ip, hBits, mls);
+        size_t const h = ZSTD_hashPtr(ip, hlog, mls);
         U32 const current = (U32)(ip-base);
         U32 const matchIndex = hashTable[h];
         const BYTE* match = base + matchIndex;
@@ -65,21 +75,21 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
         if ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1))) {
             mLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
             ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
         } else {
-            U32 offset;
-            if ( (matchIndex <= lowestIndex) || (MEM_read32(match) != MEM_read32(ip)) ) {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
+            if ( (matchIndex <= lowestIndex)
+              || (MEM_read32(match) != MEM_read32(ip)) ) {
+                assert(stepSize >= 1);
+                ip += ((ip-anchor) >> kSearchStrength) + stepSize;
                 continue;
             }
             mLength = ZSTD_count(ip+4, match+4, iend) + 4;
-            offset = (U32)(ip-match);
-            while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
-            offset_2 = offset_1;
-            offset_1 = offset;
-
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
-        }
+            {   U32 const offset = (U32)(ip-match);
+                while (((ip>anchor) & (match>lowest)) && (ip[-1] == match[-1])) { ip--; match--; mLength++; } /* catch up */
+                offset_2 = offset_1;
+                offset_1 = offset;
+                ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+        }   }
 
         /* match found */
         ip += mLength;
@@ -87,8 +97,8 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
 
         if (ip <= ilimit) {
             /* Fill Table */
-            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;  /* here because current+2 could be > iend-8 */
-            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
+            hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2;  /* here because current+2 could be > iend-8 */
+            hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
             /* check immediate repcode */
             while ( (ip <= ilimit)
                  && ( (offset_2>0)
@@ -96,65 +106,67 @@ size_t ZSTD_compressBlock_fast_generic(ZSTD_CCtx* cctx,
                 /* store sequence */
                 size_t const rLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
                 { U32 const tmpOff = offset_2; offset_2 = offset_1; offset_1 = tmpOff; }  /* swap offset_2 <=> offset_1 */
-                hashTable[ZSTD_hashPtr(ip, hBits, mls)] = (U32)(ip-base);
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+                hashTable[ZSTD_hashPtr(ip, hlog, mls)] = (U32)(ip-base);
+                ZSTD_storeSeq(seqStore, 0, anchor, 0, rLength-MINMATCH);
                 ip += rLength;
                 anchor = ip;
                 continue;   /* faster when present ... (?) */
     }   }   }
 
     /* save reps for next block */
-    seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : offsetSaved;
-    seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : offsetSaved;
+    rep[0] = offset_1 ? offset_1 : offsetSaved;
+    rep[1] = offset_2 ? offset_2 : offsetSaved;
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
-                       const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_fast(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    const U32 mls = ctx->appliedParams.cParams.searchLength;
+    U32 const hlog = cParams->hashLog;
+    U32 const mls = cParams->searchLength;
+    U32 const stepSize = cParams->targetLength;
     switch(mls)
     {
     default: /* includes case 3 */
     case 4 :
-        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 4);
+        return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 4);
     case 5 :
-        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 5);
+        return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 5);
     case 6 :
-        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 6);
+        return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 6);
     case 7 :
-        return ZSTD_compressBlock_fast_generic(ctx, src, srcSize, 7);
+        return ZSTD_compressBlock_fast_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 7);
     }
 }
 
 
-static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
-                                 const void* src, size_t srcSize,
-                                 const U32 mls)
+static size_t ZSTD_compressBlock_fast_extDict_generic(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize,
+        U32 const hlog, U32 const stepSize, U32 const mls)
 {
-    U32* hashTable = ctx->hashTable;
-    const U32 hBits = ctx->appliedParams.cParams.hashLog;
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    const BYTE* const base = ctx->base;
-    const BYTE* const dictBase = ctx->dictBase;
+    U32* hashTable = ms->hashTable;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const dictBase = ms->window.dictBase;
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
-    const U32   lowestIndex = ctx->lowLimit;
+    const U32   lowestIndex = ms->window.lowLimit;
     const BYTE* const dictStart = dictBase + lowestIndex;
-    const U32   dictLimit = ctx->dictLimit;
+    const U32   dictLimit = ms->window.dictLimit;
     const BYTE* const lowPrefixPtr = base + dictLimit;
     const BYTE* const dictEnd = dictBase + dictLimit;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - 8;
-    U32 offset_1=seqStorePtr->rep[0], offset_2=seqStorePtr->rep[1];
+    U32 offset_1=rep[0], offset_2=rep[1];
 
     /* Search Loop */
     while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
-        const size_t h = ZSTD_hashPtr(ip, hBits, mls);
+        const size_t h = ZSTD_hashPtr(ip, hlog, mls);
         const U32 matchIndex = hashTable[h];
         const BYTE* matchBase = matchIndex < dictLimit ? dictBase : base;
         const BYTE* match = matchBase + matchIndex;
@@ -170,11 +182,12 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
             const BYTE* repMatchEnd = repIndex < dictLimit ? dictEnd : iend;
             mLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, lowPrefixPtr) + 4;
             ip++;
-            ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, 0, mLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, ip-anchor, anchor, 0, mLength-MINMATCH);
         } else {
             if ( (matchIndex < lowestIndex) ||
                  (MEM_read32(match) != MEM_read32(ip)) ) {
-                ip += ((ip-anchor) >> g_searchStrength) + 1;
+                assert(stepSize >= 1);
+                ip += ((ip-anchor) >> kSearchStrength) + stepSize;
                 continue;
             }
             {   const BYTE* matchEnd = matchIndex < dictLimit ? dictEnd : iend;
@@ -185,7 +198,7 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
                 offset = current - matchIndex;
                 offset_2 = offset_1;
                 offset_1 = offset;
-                ZSTD_storeSeq(seqStorePtr, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, ip-anchor, anchor, offset + ZSTD_REP_MOVE, mLength-MINMATCH);
         }   }
 
         /* found a match : store it */
@@ -194,8 +207,8 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
 
         if (ip <= ilimit) {
             /* Fill Table */
-            hashTable[ZSTD_hashPtr(base+current+2, hBits, mls)] = current+2;
-            hashTable[ZSTD_hashPtr(ip-2, hBits, mls)] = (U32)(ip-2-base);
+            hashTable[ZSTD_hashPtr(base+current+2, hlog, mls)] = current+2;
+            hashTable[ZSTD_hashPtr(ip-2, hlog, mls)] = (U32)(ip-2-base);
             /* check immediate repcode */
             while (ip <= ilimit) {
                 U32 const current2 = (U32)(ip-base);
@@ -206,8 +219,8 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
                     const BYTE* const repEnd2 = repIndex2 < dictLimit ? dictEnd : iend;
                     size_t const repLength2 = ZSTD_count_2segments(ip+4, repMatch2+4, iend, repEnd2, lowPrefixPtr) + 4;
                     U32 tmpOffset = offset_2; offset_2 = offset_1; offset_1 = tmpOffset;   /* swap offset_2 <=> offset_1 */
-                    ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
-                    hashTable[ZSTD_hashPtr(ip, hBits, mls)] = current2;
+                    ZSTD_storeSeq(seqStore, 0, anchor, 0, repLength2-MINMATCH);
+                    hashTable[ZSTD_hashPtr(ip, hlog, mls)] = current2;
                     ip += repLength2;
                     anchor = ip;
                     continue;
@@ -216,27 +229,31 @@ static size_t ZSTD_compressBlock_fast_extDict_generic(ZSTD_CCtx* ctx,
     }   }   }
 
     /* save reps for next block */
-    seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
+    rep[0] = offset_1;
+    rep[1] = offset_2;
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
-                         const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_fast_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    U32 const mls = ctx->appliedParams.cParams.searchLength;
+    U32 const hlog = cParams->hashLog;
+    U32 const mls = cParams->searchLength;
+    U32 const stepSize = cParams->targetLength;
     switch(mls)
     {
     default: /* includes case 3 */
     case 4 :
-        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 4);
+        return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 4);
     case 5 :
-        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 5);
+        return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 5);
     case 6 :
-        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 6);
+        return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 6);
     case 7 :
-        return ZSTD_compressBlock_fast_extDict_generic(ctx, src, srcSize, 7);
+        return ZSTD_compressBlock_fast_extDict_generic(ms, seqStore, rep, src, srcSize, hlog, stepSize, 7);
     }
 }

src/borg/algorithms/zstd/lib/compress/zstd_fast.h

@@ -11,17 +11,22 @@
 #ifndef ZSTD_FAST_H
 #define ZSTD_FAST_H
 
-#include "zstd_compress.h"
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
-void ZSTD_fillHashTable(ZSTD_CCtx* zc, const void* end, const U32 mls);
-size_t ZSTD_compressBlock_fast(ZSTD_CCtx* ctx,
-                         const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_fast_extDict(ZSTD_CCtx* ctx,
-                         const void* src, size_t srcSize);
+#include "mem.h"      /* U32 */
+#include "zstd_compress_internal.h"
+
+void ZSTD_fillHashTable(ZSTD_matchState_t* ms,
+                        ZSTD_compressionParameters const* cParams,
+                        void const* end);
+size_t ZSTD_compressBlock_fast(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_fast_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
 
 #if defined (__cplusplus)
 }

src/borg/algorithms/zstd/lib/compress/zstd_lazy.c

@@ -8,78 +8,98 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
+#include "zstd_compress_internal.h"
 #include "zstd_lazy.h"
 
 
 /*-*************************************
 *  Binary Tree search
 ***************************************/
-/** ZSTD_insertBt1() : add one or multiple positions to tree.
-*   ip : assumed <= iend-8 .
-*   @return : nb of positions added */
-static U32 ZSTD_insertBt1(ZSTD_CCtx* zc, const BYTE* const ip, const U32 mls, const BYTE* const iend, U32 nbCompares,
-                          U32 extDict)
+
+void ZSTD_updateDUBT(
+                ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                const BYTE* ip, const BYTE* iend,
+                U32 mls)
 {
-    U32*   const hashTable = zc->hashTable;
-    U32    const hashLog = zc->appliedParams.cParams.hashLog;
-    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
-    U32*   const bt = zc->chainTable;
-    U32    const btLog  = zc->appliedParams.cParams.chainLog - 1;
+    U32* const hashTable = ms->hashTable;
+    U32  const hashLog = cParams->hashLog;
+
+    U32* const bt = ms->chainTable;
+    U32  const btLog  = cParams->chainLog - 1;
+    U32  const btMask = (1 << btLog) - 1;
+
+    const BYTE* const base = ms->window.base;
+    U32 const target = (U32)(ip - base);
+    U32 idx = ms->nextToUpdate;
+
+    if (idx != target)
+        DEBUGLOG(7, "ZSTD_updateDUBT, from %u to %u (dictLimit:%u)",
+                    idx, target, ms->window.dictLimit);
+    assert(ip + 8 <= iend);   /* condition for ZSTD_hashPtr */
+    (void)iend;
+
+    assert(idx >= ms->window.dictLimit);   /* condition for valid base+idx */
+    for ( ; idx < target ; idx++) {
+        size_t const h  = ZSTD_hashPtr(base + idx, hashLog, mls);   /* assumption : ip + 8 <= iend */
+        U32    const matchIndex = hashTable[h];
+
+        U32*   const nextCandidatePtr = bt + 2*(idx&btMask);
+        U32*   const sortMarkPtr  = nextCandidatePtr + 1;
+
+        DEBUGLOG(8, "ZSTD_updateDUBT: insert %u", idx);
+        hashTable[h] = idx;   /* Update Hash Table */
+        *nextCandidatePtr = matchIndex;   /* update BT like a chain */
+        *sortMarkPtr = ZSTD_DUBT_UNSORTED_MARK;
+    }
+    ms->nextToUpdate = target;
+}
+
+
+/** ZSTD_insertDUBT1() :
+ *  sort one already inserted but unsorted position
+ *  assumption : current >= btlow == (current - btmask)
+ *  doesn't fail */
+static void ZSTD_insertDUBT1(
+                 ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                 U32 current, const BYTE* inputEnd,
+                 U32 nbCompares, U32 btLow, int extDict)
+{
+    U32*   const bt = ms->chainTable;
+    U32    const btLog  = cParams->chainLog - 1;
     U32    const btMask = (1 << btLog) - 1;
-    U32 matchIndex = hashTable[h];
     size_t commonLengthSmaller=0, commonLengthLarger=0;
-    const BYTE* const base = zc->base;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const dictBase = ms->window.dictBase;
+    const U32 dictLimit = ms->window.dictLimit;
+    const BYTE* const ip = (current>=dictLimit) ? base + current : dictBase + current;
+    const BYTE* const iend = (current>=dictLimit) ? inputEnd : dictBase + dictLimit;
     const BYTE* const dictEnd = dictBase + dictLimit;
     const BYTE* const prefixStart = base + dictLimit;
     const BYTE* match;
-    const U32 current = (U32)(ip-base);
-    const U32 btLow = btMask >= current ? 0 : current - btMask;
     U32* smallerPtr = bt + 2*(current&btMask);
     U32* largerPtr  = smallerPtr + 1;
+    U32 matchIndex = *smallerPtr;
     U32 dummy32;   /* to be nullified at the end */
-    U32 const windowLow = zc->lowLimit;
-    U32 matchEndIdx = current+8;
-    size_t bestLength = 8;
-#ifdef ZSTD_C_PREDICT
-    U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
-    U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
-    predictedSmall += (predictedSmall>0);
-    predictedLarge += (predictedLarge>0);
-#endif /* ZSTD_C_PREDICT */
+    U32 const windowLow = ms->window.lowLimit;
 
-    assert(ip <= iend-8);   /* required for h calculation */
-    hashTable[h] = current;   /* Update Hash Table */
+    DEBUGLOG(8, "ZSTD_insertDUBT1(%u) (dictLimit=%u, lowLimit=%u)",
+                current, dictLimit, windowLow);
+    assert(current >= btLow);
+    assert(ip < iend);   /* condition for ZSTD_count */
 
     while (nbCompares-- && (matchIndex > windowLow)) {
         U32* const nextPtr = bt + 2*(matchIndex & btMask);
         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
-
-#ifdef ZSTD_C_PREDICT   /* note : can create issues when hlog small <= 11 */
-        const U32* predictPtr = bt + 2*((matchIndex-1) & btMask);   /* written this way, as bt is a roll buffer */
-        if (matchIndex == predictedSmall) {
-            /* no need to check length, result known */
-            *smallerPtr = matchIndex;
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-            predictedSmall = predictPtr[1] + (predictPtr[1]>0);
-            continue;
-        }
-        if (matchIndex == predictedLarge) {
-            *largerPtr = matchIndex;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-            predictedLarge = predictPtr[0] + (predictPtr[0]>0);
-            continue;
-        }
-#endif
-        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
-            match = base + matchIndex;
-            if (match[matchLength] == ip[matchLength])
-                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
+        assert(matchIndex < current);
+
+        if ( (!extDict)
+          || (matchIndex+matchLength >= dictLimit)  /* both in current segment*/
+          || (current < dictLimit) /* both in extDict */) {
+            const BYTE* const mBase = !extDict || ((matchIndex+matchLength) >= dictLimit) ? base : dictBase;
+            assert( (matchIndex+matchLength >= dictLimit)   /* might be wrong if extDict is incorrectly set to 0 */
+                 || (current < dictLimit) );
+            match = mBase + matchIndex;
+            matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
         } else {
             match = dictBase + matchIndex;
             matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
@@ -87,194 +107,226 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc, const BYTE* const ip, const U32 mls, co
                 match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
         }
 
-        if (matchLength > bestLength) {
-            bestLength = matchLength;
-            if (matchLength > matchEndIdx - matchIndex)
-                matchEndIdx = matchIndex + (U32)matchLength;
-        }
+        DEBUGLOG(8, "ZSTD_insertDUBT1: comparing %u with %u : found %u common bytes ",
+                    current, matchIndex, (U32)matchLength);
 
-        if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+        if (ip+matchLength == iend) {   /* equal : no way to know if inf or sup */
             break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
+        }
 
         if (match[matchLength] < ip[matchLength]) {  /* necessarily within buffer */
-            /* match+1 is smaller than current */
+            /* match is smaller than current */
             *smallerPtr = matchIndex;             /* update smaller idx */
             commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
             if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is smaller : next => %u",
+                        matchIndex, btLow, nextPtr[1]);
+            smallerPtr = nextPtr+1;               /* new "candidate" => larger than match, which was smaller than target */
+            matchIndex = nextPtr[1];              /* new matchIndex, larger than previous and closer to current */
         } else {
             /* match is larger than current */
             *largerPtr = matchIndex;
             commonLengthLarger = matchLength;
             if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
+            DEBUGLOG(8, "ZSTD_insertDUBT1: %u (>btLow=%u) is larger => %u",
+                        matchIndex, btLow, nextPtr[0]);
             largerPtr = nextPtr;
             matchIndex = nextPtr[0];
     }   }
 
     *smallerPtr = *largerPtr = 0;
-    if (bestLength > 384) return MIN(192, (U32)(bestLength - 384));   /* speed optimization */
-    if (matchEndIdx > current + 8) return matchEndIdx - (current + 8);
-    return 1;
 }
 
 
-static size_t ZSTD_insertBtAndFindBestMatch (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iend,
-                        size_t* offsetPtr,
-                        U32 nbCompares, const U32 mls,
-                        U32 extDict)
+static size_t ZSTD_DUBT_findBestMatch (
+                            ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                            const BYTE* const ip, const BYTE* const iend,
+                            size_t* offsetPtr,
+                            U32 const mls,
+                            U32 const extDict)
 {
-    U32*   const hashTable = zc->hashTable;
-    U32    const hashLog = zc->appliedParams.cParams.hashLog;
+    U32*   const hashTable = ms->hashTable;
+    U32    const hashLog = cParams->hashLog;
     size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
-    U32*   const bt = zc->chainTable;
-    U32    const btLog  = zc->appliedParams.cParams.chainLog - 1;
+    U32          matchIndex  = hashTable[h];
+
+    const BYTE* const base = ms->window.base;
+    U32    const current = (U32)(ip-base);
+    U32    const windowLow = ms->window.lowLimit;
+
+    U32*   const bt = ms->chainTable;
+    U32    const btLog  = cParams->chainLog - 1;
     U32    const btMask = (1 << btLog) - 1;
-    U32 matchIndex  = hashTable[h];
-    size_t commonLengthSmaller=0, commonLengthLarger=0;
-    const BYTE* const base = zc->base;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const U32 current = (U32)(ip-base);
-    const U32 btLow = btMask >= current ? 0 : current - btMask;
-    const U32 windowLow = zc->lowLimit;
-    U32* smallerPtr = bt + 2*(current&btMask);
-    U32* largerPtr  = bt + 2*(current&btMask) + 1;
-    U32 matchEndIdx = current+8;
-    U32 dummy32;   /* to be nullified at the end */
-    size_t bestLength = 0;
+    U32    const btLow = (btMask >= current) ? 0 : current - btMask;
+    U32    const unsortLimit = MAX(btLow, windowLow);
 
-    assert(ip <= iend-8);   /* required for h calculation */
-    hashTable[h] = current;   /* Update Hash Table */
+    U32*         nextCandidate = bt + 2*(matchIndex&btMask);
+    U32*         unsortedMark = bt + 2*(matchIndex&btMask) + 1;
+    U32          nbCompares = 1U << cParams->searchLog;
+    U32          nbCandidates = nbCompares;
+    U32          previousCandidate = 0;
 
-    while (nbCompares-- && (matchIndex > windowLow)) {
-        U32* const nextPtr = bt + 2*(matchIndex & btMask);
-        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
-        const BYTE* match;
+    DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%u) ", current);
+    assert(ip <= iend-8);   /* required for h calculation */
 
-        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
-            match = base + matchIndex;
-            if (match[matchLength] == ip[matchLength])
-                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iend) +1;
-        } else {
-            match = dictBase + matchIndex;
-            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
-            if (matchIndex+matchLength >= dictLimit)
-                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
-        }
+    /* reach end of unsorted candidates list */
+    while ( (matchIndex > unsortLimit)
+         && (*unsortedMark == ZSTD_DUBT_UNSORTED_MARK)
+         && (nbCandidates > 1) ) {
+        DEBUGLOG(8, "ZSTD_DUBT_findBestMatch: candidate %u is unsorted",
+                    matchIndex);
+        *unsortedMark = previousCandidate;
+        previousCandidate = matchIndex;
+        matchIndex = *nextCandidate;
+        nextCandidate = bt + 2*(matchIndex&btMask);
+        unsortedMark = bt + 2*(matchIndex&btMask) + 1;
+        nbCandidates --;
+    }
 
-        if (matchLength > bestLength) {
-            if (matchLength > matchEndIdx - matchIndex)
-                matchEndIdx = matchIndex + (U32)matchLength;
-            if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
-                bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
-            if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
-                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
-        }
+    if ( (matchIndex > unsortLimit)
+      && (*unsortedMark==ZSTD_DUBT_UNSORTED_MARK) ) {
+        DEBUGLOG(7, "ZSTD_DUBT_findBestMatch: nullify last unsorted candidate %u",
+                    matchIndex);
+        *nextCandidate = *unsortedMark = 0;   /* nullify next candidate if it's still unsorted (note : simplification, detrimental to compression ratio, beneficial for speed) */
+    }
 
-        if (match[matchLength] < ip[matchLength]) {
-            /* match is smaller than current */
-            *smallerPtr = matchIndex;             /* update smaller idx */
-            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-        } else {
-            /* match is larger than current */
-            *largerPtr = matchIndex;
-            commonLengthLarger = matchLength;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-    }   }
+    /* batch sort stacked candidates */
+    matchIndex = previousCandidate;
+    while (matchIndex) {  /* will end on matchIndex == 0 */
+        U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
+        U32 const nextCandidateIdx = *nextCandidateIdxPtr;
+        ZSTD_insertDUBT1(ms, cParams, matchIndex, iend,
+                         nbCandidates, unsortLimit, extDict);
+        matchIndex = nextCandidateIdx;
+        nbCandidates++;
+    }
 
-    *smallerPtr = *largerPtr = 0;
+    /* find longest match */
+    {   size_t commonLengthSmaller=0, commonLengthLarger=0;
+        const BYTE* const dictBase = ms->window.dictBase;
+        const U32 dictLimit = ms->window.dictLimit;
+        const BYTE* const dictEnd = dictBase + dictLimit;
+        const BYTE* const prefixStart = base + dictLimit;
+        U32* smallerPtr = bt + 2*(current&btMask);
+        U32* largerPtr  = bt + 2*(current&btMask) + 1;
+        U32 matchEndIdx = current+8+1;
+        U32 dummy32;   /* to be nullified at the end */
+        size_t bestLength = 0;
+
+        matchIndex  = hashTable[h];
+        hashTable[h] = current;   /* Update Hash Table */
+
+        while (nbCompares-- && (matchIndex > windowLow)) {
+            U32* const nextPtr = bt + 2*(matchIndex & btMask);
+            size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+            const BYTE* match;
+
+            if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+                match = base + matchIndex;
+                matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+            } else {
+                match = dictBase + matchIndex;
+                matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+                if (matchIndex+matchLength >= dictLimit)
+                    match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+            }
 
-    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
-    return bestLength;
-}
+            if (matchLength > bestLength) {
+                if (matchLength > matchEndIdx - matchIndex)
+                    matchEndIdx = matchIndex + (U32)matchLength;
+                if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
+                    bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
+                if (ip+matchLength == iend) {   /* equal : no way to know if inf or sup */
+                    break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+                }
+            }
 
+            if (match[matchLength] < ip[matchLength]) {
+                /* match is smaller than current */
+                *smallerPtr = matchIndex;             /* update smaller idx */
+                commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+                if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+                smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+                matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            } else {
+                /* match is larger than current */
+                *largerPtr = matchIndex;
+                commonLengthLarger = matchLength;
+                if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+                largerPtr = nextPtr;
+                matchIndex = nextPtr[0];
+        }   }
 
-void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
-{
-    const BYTE* const base = zc->base;
-    const U32 target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
+        *smallerPtr = *largerPtr = 0;
 
-    while(idx < target)
-        idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 0);
+        assert(matchEndIdx > current+8); /* ensure nextToUpdate is increased */
+        ms->nextToUpdate = matchEndIdx - 8;   /* skip repetitive patterns */
+        if (bestLength >= MINMATCH) {
+            U32 const mIndex = current - ((U32)*offsetPtr - ZSTD_REP_MOVE); (void)mIndex;
+            DEBUGLOG(8, "ZSTD_DUBT_findBestMatch(%u) : found match of length %u and offsetCode %u (pos %u)",
+                        current, (U32)bestLength, (U32)*offsetPtr, mIndex);
+        }
+        return bestLength;
+    }
 }
 
+
 /** ZSTD_BtFindBestMatch() : Tree updater, providing best match */
 static size_t ZSTD_BtFindBestMatch (
-                        ZSTD_CCtx* zc,
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* const ip, const BYTE* const iLimit,
                         size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 mls)
+                        const U32 mls /* template */)
 {
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 0);
+    DEBUGLOG(7, "ZSTD_BtFindBestMatch");
+    if (ip < ms->window.base + ms->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateDUBT(ms, cParams, ip, iLimit, mls);
+    return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, 0);
 }
 
 
 static size_t ZSTD_BtFindBestMatch_selectMLS (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+                        size_t* offsetPtr)
 {
-    switch(matchLengthSearch)
+    switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
-    case 5 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4);
+    case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5);
     case 7 :
-    case 6 : return ZSTD_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6);
     }
 }
 
 
-void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
-{
-    const BYTE* const base = zc->base;
-    const U32 target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
-
-    while (idx < target) idx += ZSTD_insertBt1(zc, base+idx, mls, iend, nbCompares, 1);
-}
-
-
 /** Tree updater, providing best match */
 static size_t ZSTD_BtFindBestMatch_extDict (
-                        ZSTD_CCtx* zc,
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* const ip, const BYTE* const iLimit,
                         size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 mls)
+                        const U32 mls)
 {
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls, 1);
+    DEBUGLOG(7, "ZSTD_BtFindBestMatch_extDict");
+    if (ip < ms->window.base + ms->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateDUBT(ms, cParams, ip, iLimit, mls);
+    return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, 1);
 }
 
 
 static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+                        size_t* offsetPtr)
 {
-    switch(matchLengthSearch)
+    switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
-    case 5 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 4 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 4);
+    case 5 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 5);
     case 7 :
-    case 6 : return ZSTD_BtFindBestMatch_extDict(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    case 6 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 6);
     }
 }
 
@@ -287,15 +339,17 @@ static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
 
 /* Update chains up to ip (excluded)
    Assumption : always within prefix (i.e. not within extDict) */
-U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
+static U32 ZSTD_insertAndFindFirstIndex_internal(
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip, U32 const mls)
 {
-    U32* const hashTable  = zc->hashTable;
-    const U32 hashLog = zc->appliedParams.cParams.hashLog;
-    U32* const chainTable = zc->chainTable;
-    const U32 chainMask = (1 << zc->appliedParams.cParams.chainLog) - 1;
-    const BYTE* const base = zc->base;
+    U32* const hashTable  = ms->hashTable;
+    const U32 hashLog = cParams->hashLog;
+    U32* const chainTable = ms->chainTable;
+    const U32 chainMask = (1 << cParams->chainLog) - 1;
+    const BYTE* const base = ms->window.base;
     const U32 target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
+    U32 idx = ms->nextToUpdate;
 
     while(idx < target) { /* catch up */
         size_t const h = ZSTD_hashPtr(base+idx, hashLog, mls);
@@ -304,45 +358,52 @@ U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls)
         idx++;
     }
 
-    zc->nextToUpdate = target;
+    ms->nextToUpdate = target;
     return hashTable[ZSTD_hashPtr(ip, hashLog, mls)];
 }
 
+U32 ZSTD_insertAndFindFirstIndex(
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip)
+{
+    return ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, cParams->searchLength);
+}
+
 
 /* inlining is important to hardwire a hot branch (template emulation) */
 FORCE_INLINE_TEMPLATE
 size_t ZSTD_HcFindBestMatch_generic (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* const ip, const BYTE* const iLimit,
                         size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 mls, const U32 extDict)
+                        const U32 mls, const U32 extDict)
 {
-    U32* const chainTable = zc->chainTable;
-    const U32 chainSize = (1 << zc->appliedParams.cParams.chainLog);
+    U32* const chainTable = ms->chainTable;
+    const U32 chainSize = (1 << cParams->chainLog);
     const U32 chainMask = chainSize-1;
-    const BYTE* const base = zc->base;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const dictBase = ms->window.dictBase;
+    const U32 dictLimit = ms->window.dictLimit;
     const BYTE* const prefixStart = base + dictLimit;
     const BYTE* const dictEnd = dictBase + dictLimit;
-    const U32 lowLimit = zc->lowLimit;
+    const U32 lowLimit = ms->window.lowLimit;
     const U32 current = (U32)(ip-base);
     const U32 minChain = current > chainSize ? current - chainSize : 0;
-    int nbAttempts=maxNbAttempts;
+    U32 nbAttempts = 1U << cParams->searchLog;
     size_t ml=4-1;
 
     /* HC4 match finder */
-    U32 matchIndex = ZSTD_insertAndFindFirstIndex (zc, ip, mls);
+    U32 matchIndex = ZSTD_insertAndFindFirstIndex_internal(ms, cParams, ip, mls);
 
     for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
-        const BYTE* match;
         size_t currentMl=0;
         if ((!extDict) || matchIndex >= dictLimit) {
-            match = base + matchIndex;
+            const BYTE* const match = base + matchIndex;
             if (match[ml] == ip[ml])   /* potentially better */
                 currentMl = ZSTD_count(ip, match, iLimit);
         } else {
-            match = dictBase + matchIndex;
+            const BYTE* const match = dictBase + matchIndex;
+            assert(match+4 <= dictEnd);
             if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
                 currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dictEnd, prefixStart) + 4;
         }
@@ -363,35 +424,33 @@ size_t ZSTD_HcFindBestMatch_generic (
 
 
 FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
-                        ZSTD_CCtx* zc,
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+                        size_t* offsetPtr)
 {
-    switch(matchLengthSearch)
+    switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 0);
-    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 0);
+    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 0);
+    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 0);
     case 7 :
-    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 0);
+    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 0);
     }
 }
 
 
 FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
-                        ZSTD_CCtx* zc,
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+                        size_t* const offsetPtr)
 {
-    switch(matchLengthSearch)
+    switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4, 1);
-    case 5 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5, 1);
+    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 1);
+    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 1);
     case 7 :
-    case 6 : return ZSTD_HcFindBestMatch_generic(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6, 1);
+    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 1);
     }
 }
 
@@ -400,30 +459,29 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
 *  Common parser - lazy strategy
 *********************************/
 FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
-                                       const void* src, size_t srcSize,
-                                       const U32 searchMethod, const U32 depth)
+size_t ZSTD_compressBlock_lazy_generic(
+                        ZSTD_matchState_t* ms, seqStore_t* seqStore,
+                        U32 rep[ZSTD_REP_NUM],
+                        ZSTD_compressionParameters const* cParams,
+                        const void* src, size_t srcSize,
+                        const U32 searchMethod, const U32 depth)
 {
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base + ctx->dictLimit;
+    const BYTE* const base = ms->window.base + ms->window.dictLimit;
 
-    U32 const maxSearches = 1 << ctx->appliedParams.cParams.searchLog;
-    U32 const mls = ctx->appliedParams.cParams.searchLength;
-
-    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
-                        size_t* offsetPtr,
-                        U32 maxNbAttempts, U32 matchLengthSearch);
+    typedef size_t (*searchMax_f)(
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
     searchMax_f const searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS;
-    U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1], savedOffset=0;
+    U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
 
     /* init */
     ip += (ip==base);
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ms->nextToUpdate3 = ms->nextToUpdate;
     {   U32 const maxRep = (U32)(ip-base);
         if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
         if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
@@ -444,13 +502,13 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
 
         /* first search (depth 0) */
         {   size_t offsetFound = 99999999;
-            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
+            size_t const ml2 = searchMax(ms, cParams, ip, iend, &offsetFound);
             if (ml2 > matchLength)
                 matchLength = ml2, start = ip, offset=offsetFound;
         }
 
         if (matchLength < 4) {
-            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
+            ip += ((ip-anchor) >> kSearchStrength) + 1;   /* jump faster over incompressible sections */
             continue;
         }
 
@@ -466,7 +524,7 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
                     matchLength = mlRep, offset = 0, start = ip;
             }
             {   size_t offset2=99999999;
-                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
                 int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
                 int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
                 if ((ml2 >= 4) && (gain2 > gain1)) {
@@ -485,7 +543,7 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
                         matchLength = ml2, offset = 0, start = ip;
                 }
                 {   size_t offset2=99999999;
-                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                    size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
                     int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
                     int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
                     if ((ml2 >= 4) && (gain2 > gain1)) {
@@ -502,93 +560,98 @@ size_t ZSTD_compressBlock_lazy_generic(ZSTD_CCtx* ctx,
          */
         /* catch up */
         if (offset) {
-            while ( (start > anchor)
-                 && (start > base+offset-ZSTD_REP_MOVE)
-                 && (start[-1] == (start-offset+ZSTD_REP_MOVE)[-1]) )  /* only search for offset within prefix */
+            while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > base))
+                 && (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) )  /* only search for offset within prefix */
                 { start--; matchLength++; }
             offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
         }
         /* store sequence */
 _storeSequence:
         {   size_t const litLength = start - anchor;
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
             anchor = ip = start + matchLength;
         }
 
         /* check immediate repcode */
-        while ( (ip <= ilimit)
-             && ((offset_2>0)
-             & (MEM_read32(ip) == MEM_read32(ip - offset_2)) )) {
+        while ( ((ip <= ilimit) & (offset_2>0))
+             && (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
             /* store sequence */
             matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
             offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
-            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
             ip += matchLength;
             anchor = ip;
             continue;   /* faster when present ... (?) */
     }   }
 
     /* Save reps for next block */
-    seqStorePtr->repToConfirm[0] = offset_1 ? offset_1 : savedOffset;
-    seqStorePtr->repToConfirm[1] = offset_2 ? offset_2 : savedOffset;
+    rep[0] = offset_1 ? offset_1 : savedOffset;
+    rep[1] = offset_2 ? offset_2 : savedOffset;
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_btlazy2(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 1, 2);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2);
 }
 
-size_t ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_lazy2(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 2);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2);
 }
 
-size_t ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_lazy(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 1);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1);
 }
 
-size_t ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_greedy(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_lazy_generic(ctx, src, srcSize, 0, 0);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0);
 }
 
 
 FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
-                                     const void* src, size_t srcSize,
-                                     const U32 searchMethod, const U32 depth)
+size_t ZSTD_compressBlock_lazy_extDict_generic(
+                        ZSTD_matchState_t* ms, seqStore_t* seqStore,
+                        U32 rep[ZSTD_REP_NUM],
+                        ZSTD_compressionParameters const* cParams,
+                        const void* src, size_t srcSize,
+                        const U32 searchMethod, const U32 depth)
 {
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base;
-    const U32 dictLimit = ctx->dictLimit;
-    const U32 lowestIndex = ctx->lowLimit;
+    const BYTE* const base = ms->window.base;
+    const U32 dictLimit = ms->window.dictLimit;
+    const U32 lowestIndex = ms->window.lowLimit;
     const BYTE* const prefixStart = base + dictLimit;
-    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* const dictBase = ms->window.dictBase;
     const BYTE* const dictEnd  = dictBase + dictLimit;
-    const BYTE* const dictStart  = dictBase + ctx->lowLimit;
+    const BYTE* const dictStart  = dictBase + lowestIndex;
 
-    const U32 maxSearches = 1 << ctx->appliedParams.cParams.searchLog;
-    const U32 mls = ctx->appliedParams.cParams.searchLength;
-
-    typedef size_t (*searchMax_f)(ZSTD_CCtx* zc, const BYTE* ip, const BYTE* iLimit,
-                        size_t* offsetPtr,
-                        U32 maxNbAttempts, U32 matchLengthSearch);
+    typedef size_t (*searchMax_f)(
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip, const BYTE* iLimit, size_t* offsetPtr);
     searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_selectMLS_extDict : ZSTD_HcFindBestMatch_extDict_selectMLS;
 
-    U32 offset_1 = seqStorePtr->rep[0], offset_2 = seqStorePtr->rep[1];
+    U32 offset_1 = rep[0], offset_2 = rep[1];
 
     /* init */
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ms->nextToUpdate3 = ms->nextToUpdate;
     ip += (ip == prefixStart);
 
     /* Match Loop */
@@ -612,13 +675,13 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
 
         /* first search (depth 0) */
         {   size_t offsetFound = 99999999;
-            size_t const ml2 = searchMax(ctx, ip, iend, &offsetFound, maxSearches, mls);
+            size_t const ml2 = searchMax(ms, cParams, ip, iend, &offsetFound);
             if (ml2 > matchLength)
                 matchLength = ml2, start = ip, offset=offsetFound;
         }
 
          if (matchLength < 4) {
-            ip += ((ip-anchor) >> g_searchStrength) + 1;   /* jump faster over incompressible sections */
+            ip += ((ip-anchor) >> kSearchStrength) + 1;   /* jump faster over incompressible sections */
             continue;
         }
 
@@ -645,7 +708,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
 
             /* search match, depth 1 */
             {   size_t offset2=99999999;
-                size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
                 int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
                 int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 4);
                 if ((ml2 >= 4) && (gain2 > gain1)) {
@@ -675,7 +738,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
 
                 /* search match, depth 2 */
                 {   size_t offset2=99999999;
-                    size_t const ml2 = searchMax(ctx, ip, iend, &offset2, maxSearches, mls);
+                    size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
                     int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
                     int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 7);
                     if ((ml2 >= 4) && (gain2 > gain1)) {
@@ -697,7 +760,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(ZSTD_CCtx* ctx,
         /* store sequence */
 _storeSequence:
         {   size_t const litLength = start - anchor;
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, (U32)offset, matchLength-MINMATCH);
+            ZSTD_storeSeq(seqStore, litLength, anchor, (U32)offset, matchLength-MINMATCH);
             anchor = ip = start + matchLength;
         }
 
@@ -712,7 +775,7 @@ _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(seqStorePtr, 0, anchor, 0, matchLength-MINMATCH);
+                ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
                 ip += matchLength;
                 anchor = ip;
                 continue;   /* faster when present ... (?) */
@@ -721,29 +784,41 @@ _storeSequence:
     }   }
 
     /* Save reps for next block */
-    seqStorePtr->repToConfirm[0] = offset_1; seqStorePtr->repToConfirm[1] = offset_2;
+    rep[0] = offset_1;
+    rep[1] = offset_2;
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_greedy_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 0);
+    return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0);
 }
 
-size_t ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_lazy_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
+
 {
-    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 1);
+    return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1);
 }
 
-size_t ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_lazy2_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
+
 {
-    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 0, 2);
+    return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2);
 }
 
-size_t ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_btlazy2_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
+
 {
-    return ZSTD_compressBlock_lazy_extDict_generic(ctx, src, srcSize, 1, 2);
+    return ZSTD_compressBlock_lazy_extDict_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2);
 }

src/borg/algorithms/zstd/lib/compress/zstd_lazy.h

@@ -11,25 +11,43 @@
 #ifndef ZSTD_LAZY_H
 #define ZSTD_LAZY_H
 
-#include "zstd_compress.h"
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
-U32 ZSTD_insertAndFindFirstIndex (ZSTD_CCtx* zc, const BYTE* ip, U32 mls);
-void ZSTD_updateTree(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls);
-void ZSTD_updateTree_extDict(ZSTD_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls);
-
-size_t ZSTD_compressBlock_btlazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_lazy2(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_lazy(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_greedy(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-
-size_t ZSTD_compressBlock_greedy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_lazy_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_lazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_btlazy2_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+#include "zstd_compress_internal.h"
+
+U32 ZSTD_insertAndFindFirstIndex(
+        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+        const BYTE* ip);
+
+void ZSTD_preserveUnsortedMark (U32* const table, U32 const size, U32 const reducerValue);  /*! used in ZSTD_reduceIndex(). pre-emptively increase value of ZSTD_DUBT_UNSORTED_MARK */
+
+size_t ZSTD_compressBlock_btlazy2(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+
+size_t ZSTD_compressBlock_greedy_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btlazy2_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
 
 #if defined (__cplusplus)
 }

src/borg/algorithms/zstd/lib/compress/zstd_ldm.c

@@ -17,36 +17,45 @@
 #define LDM_HASH_RLOG 7
 #define LDM_HASH_CHAR_OFFSET 10
 
-size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm)
+void ZSTD_ldm_adjustParameters(ldmParams_t* params,
+                               ZSTD_compressionParameters const* cParams)
 {
+    U32 const windowLog = cParams->windowLog;
     ZSTD_STATIC_ASSERT(LDM_BUCKET_SIZE_LOG <= ZSTD_LDM_BUCKETSIZELOG_MAX);
-    params->enableLdm = enableLdm>0;
-    params->hashLog = 0;
-    params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
-    params->minMatchLength = LDM_MIN_MATCH_LENGTH;
-    params->hashEveryLog = ZSTD_LDM_HASHEVERYLOG_NOTSET;
-    return 0;
-}
-
-void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog)
-{
+    DEBUGLOG(4, "ZSTD_ldm_adjustParameters");
+    if (!params->bucketSizeLog) params->bucketSizeLog = LDM_BUCKET_SIZE_LOG;
+    if (!params->minMatchLength) params->minMatchLength = LDM_MIN_MATCH_LENGTH;
+    if (cParams->strategy >= ZSTD_btopt) {
+      /* Get out of the way of the optimal parser */
+      U32 const minMatch = MAX(cParams->targetLength, params->minMatchLength);
+      assert(minMatch >= ZSTD_LDM_MINMATCH_MIN);
+      assert(minMatch <= ZSTD_LDM_MINMATCH_MAX);
+      params->minMatchLength = minMatch;
+    }
     if (params->hashLog == 0) {
         params->hashLog = MAX(ZSTD_HASHLOG_MIN, windowLog - LDM_HASH_RLOG);
         assert(params->hashLog <= ZSTD_HASHLOG_MAX);
     }
-    if (params->hashEveryLog == ZSTD_LDM_HASHEVERYLOG_NOTSET) {
+    if (params->hashEveryLog == 0) {
         params->hashEveryLog =
                 windowLog < params->hashLog ? 0 : windowLog - params->hashLog;
     }
     params->bucketSizeLog = MIN(params->bucketSizeLog, params->hashLog);
 }
 
-size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog) {
-    size_t const ldmHSize = ((size_t)1) << hashLog;
-    size_t const ldmBucketSizeLog = MIN(bucketSizeLog, hashLog);
+size_t 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) << (hashLog - ldmBucketSizeLog);
-    return ldmBucketSize + (ldmHSize * (sizeof(ldmEntry_t)));
+        ((size_t)1) << (params.hashLog - ldmBucketSizeLog);
+    size_t const totalSize = ldmBucketSize + ldmHSize * sizeof(ldmEntry_t);
+    return params.enableLdm ? totalSize : 0;
+}
+
+size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize)
+{
+    return params.enableLdm ? (maxChunkSize / params.minMatchLength) : 0;
 }
 
 /** ZSTD_ldm_getSmallHash() :
@@ -167,6 +176,7 @@ static U64 ZSTD_ldm_ipow(U64 base, U64 exp)
 }
 
 U64 ZSTD_ldm_getHashPower(U32 minMatchLength) {
+    DEBUGLOG(4, "ZSTD_ldm_getHashPower: mml=%u", minMatchLength);
     assert(minMatchLength >= ZSTD_LDM_MINMATCH_MIN);
     return ZSTD_ldm_ipow(prime8bytes, minMatchLength - 1);
 }
@@ -205,21 +215,22 @@ static size_t ZSTD_ldm_countBackwardsMatch(
  *
  *  The tables for the other strategies are filled within their
  *  block compressors. */
-static size_t ZSTD_ldm_fillFastTables(ZSTD_CCtx* zc, const void* end)
+static size_t ZSTD_ldm_fillFastTables(ZSTD_matchState_t* ms,
+                                      ZSTD_compressionParameters const* cParams,
+                                      void const* end)
 {
     const BYTE* const iend = (const BYTE*)end;
-    const U32 mls = zc->appliedParams.cParams.searchLength;
 
-    switch(zc->appliedParams.cParams.strategy)
+    switch(cParams->strategy)
     {
     case ZSTD_fast:
-        ZSTD_fillHashTable(zc, iend, mls);
-        zc->nextToUpdate = (U32)(iend - zc->base);
+        ZSTD_fillHashTable(ms, cParams, iend);
+        ms->nextToUpdate = (U32)(iend - ms->window.base);
         break;
 
     case ZSTD_dfast:
-        ZSTD_fillDoubleHashTable(zc, iend, mls);
-        zc->nextToUpdate = (U32)(iend - zc->base);
+        ZSTD_fillDoubleHashTable(ms, cParams, iend);
+        ms->nextToUpdate = (U32)(iend - ms->window.base);
         break;
 
     case ZSTD_greedy:
@@ -268,69 +279,62 @@ static U64 ZSTD_ldm_fillLdmHashTable(ldmState_t* state,
  *  Sets cctx->nextToUpdate to a position corresponding closer to anchor
  *  if it is far way
  *  (after a long match, only update tables a limited amount). */
-static void ZSTD_ldm_limitTableUpdate(ZSTD_CCtx* cctx, const BYTE* anchor)
+static void ZSTD_ldm_limitTableUpdate(ZSTD_matchState_t* ms, const BYTE* anchor)
 {
-    U32 const current = (U32)(anchor - cctx->base);
-    if (current > cctx->nextToUpdate + 1024) {
-        cctx->nextToUpdate =
-            current - MIN(512, current - cctx->nextToUpdate - 1024);
+    U32 const current = (U32)(anchor - ms->window.base);
+    if (current > ms->nextToUpdate + 1024) {
+        ms->nextToUpdate =
+            current - MIN(512, current - ms->nextToUpdate - 1024);
     }
 }
 
-typedef size_t (*ZSTD_blockCompressor) (ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-/* defined in zstd_compress.c */
-ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, int extDict);
-
-FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
-                                      const void* src, size_t srcSize)
+static size_t ZSTD_ldm_generateSequences_internal(
+        ldmState_t* ldmState, rawSeqStore_t* rawSeqStore,
+        ldmParams_t const* params, void const* src, size_t srcSize)
 {
-    ldmState_t* const ldmState = &(cctx->ldmState);
-    const ldmParams_t ldmParams = cctx->appliedParams.ldmParams;
-    const U64 hashPower = ldmState->hashPower;
-    const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
-    const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
-    const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
-    seqStore_t* const seqStorePtr = &(cctx->seqStore);
-    const BYTE* const base = cctx->base;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32   lowestIndex = cctx->dictLimit;
-    const BYTE* const lowest = base + lowestIndex;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
-
-    const ZSTD_blockCompressor blockCompressor =
-        ZSTD_selectBlockCompressor(cctx->appliedParams.cParams.strategy, 0);
-    U32* const repToConfirm = seqStorePtr->repToConfirm;
-    U32 savedRep[ZSTD_REP_NUM];
+    /* LDM parameters */
+    int const extDict = ZSTD_window_hasExtDict(ldmState->window);
+    U32 const minMatchLength = params->minMatchLength;
+    U64 const hashPower = ldmState->hashPower;
+    U32 const hBits = params->hashLog - params->bucketSizeLog;
+    U32 const ldmBucketSize = 1U << params->bucketSizeLog;
+    U32 const hashEveryLog = params->hashEveryLog;
+    U32 const ldmTagMask = (1U << params->hashEveryLog) - 1;
+    /* Prefix and extDict parameters */
+    U32 const dictLimit = ldmState->window.dictLimit;
+    U32 const lowestIndex = extDict ? ldmState->window.lowLimit : dictLimit;
+    BYTE const* const base = ldmState->window.base;
+    BYTE const* const dictBase = extDict ? ldmState->window.dictBase : NULL;
+    BYTE const* const dictStart = extDict ? dictBase + lowestIndex : NULL;
+    BYTE const* const dictEnd = extDict ? dictBase + dictLimit : NULL;
+    BYTE const* const lowPrefixPtr = base + dictLimit;
+    /* Input bounds */
+    BYTE const* const istart = (BYTE const*)src;
+    BYTE const* const iend = istart + srcSize;
+    BYTE const* const ilimit = iend - MAX(minMatchLength, HASH_READ_SIZE);
+    /* Input positions */
+    BYTE const* anchor = istart;
+    BYTE const* ip = istart;
+    /* Rolling hash */
+    BYTE const* lastHashed = NULL;
     U64 rollingHash = 0;
-    const BYTE* lastHashed = NULL;
-    size_t i, lastLiterals;
 
-    /* Save seqStorePtr->rep and copy repToConfirm */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
-
-    /* Main Search Loop */
-    while (ip < ilimit) {   /* < instead of <=, because repcode check at (ip+1) */
+    while (ip <= ilimit) {
         size_t mLength;
         U32 const current = (U32)(ip - base);
         size_t forwardMatchLength = 0, backwardMatchLength = 0;
         ldmEntry_t* bestEntry = NULL;
         if (ip != istart) {
             rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
-                                              lastHashed[ldmParams.minMatchLength],
+                                              lastHashed[minMatchLength],
                                               hashPower);
         } else {
-            rollingHash = ZSTD_ldm_getRollingHash(ip, ldmParams.minMatchLength);
+            rollingHash = ZSTD_ldm_getRollingHash(ip, minMatchLength);
         }
         lastHashed = ip;
 
         /* Do not insert and do not look for a match */
-        if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
-                ldmTagMask) {
+        if (ZSTD_ldm_getTag(rollingHash, hBits, hashEveryLog) != ldmTagMask) {
            ip++;
            continue;
         }
@@ -340,27 +344,49 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
             ldmEntry_t* const bucket =
                 ZSTD_ldm_getBucket(ldmState,
                                    ZSTD_ldm_getSmallHash(rollingHash, hBits),
-                                   ldmParams);
+                                   *params);
             ldmEntry_t* cur;
             size_t bestMatchLength = 0;
             U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
 
             for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
-                const BYTE* const pMatch = cur->offset + base;
                 size_t curForwardMatchLength, curBackwardMatchLength,
                        curTotalMatchLength;
                 if (cur->checksum != checksum || cur->offset <= lowestIndex) {
                     continue;
                 }
-
-                curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
-                if (curForwardMatchLength < ldmParams.minMatchLength) {
-                    continue;
+                if (extDict) {
+                    BYTE const* const curMatchBase =
+                        cur->offset < dictLimit ? dictBase : base;
+                    BYTE const* const pMatch = curMatchBase + cur->offset;
+                    BYTE const* const matchEnd =
+                        cur->offset < dictLimit ? dictEnd : iend;
+                    BYTE const* const lowMatchPtr =
+                        cur->offset < dictLimit ? dictStart : lowPrefixPtr;
+
+                    curForwardMatchLength = ZSTD_count_2segments(
+                                                ip, pMatch, iend,
+                                                matchEnd, lowPrefixPtr);
+                    if (curForwardMatchLength < minMatchLength) {
+                        continue;
+                    }
+                    curBackwardMatchLength =
+                        ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
+                                                     lowMatchPtr);
+                    curTotalMatchLength = curForwardMatchLength +
+                                          curBackwardMatchLength;
+                } else { /* !extDict */
+                    BYTE const* const pMatch = base + cur->offset;
+                    curForwardMatchLength = ZSTD_count(ip, pMatch, iend);
+                    if (curForwardMatchLength < minMatchLength) {
+                        continue;
+                    }
+                    curBackwardMatchLength =
+                        ZSTD_ldm_countBackwardsMatch(ip, anchor, pMatch,
+                                                     lowPrefixPtr);
+                    curTotalMatchLength = curForwardMatchLength +
+                                          curBackwardMatchLength;
                 }
-                curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
-                                             ip, anchor, pMatch, lowest);
-                curTotalMatchLength = curForwardMatchLength +
-                                      curBackwardMatchLength;
 
                 if (curTotalMatchLength > bestMatchLength) {
                     bestMatchLength = curTotalMatchLength;
@@ -375,7 +401,7 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
         if (bestEntry == NULL) {
             ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash,
                                              hBits, current,
-                                             ldmParams);
+                                             *params);
             ip++;
             continue;
         }
@@ -384,324 +410,244 @@ size_t ZSTD_compressBlock_ldm_generic(ZSTD_CCtx* cctx,
         mLength = forwardMatchLength + backwardMatchLength;
         ip -= backwardMatchLength;
 
-        /* Call the block compressor on the remaining literals */
         {
+            /* Store the sequence:
+             * ip = current - backwardMatchLength
+             * The match is at (bestEntry->offset - backwardMatchLength)
+             */
             U32 const matchIndex = bestEntry->offset;
-            const BYTE* const match = base + matchIndex - backwardMatchLength;
-            U32 const offset = (U32)(ip - match);
-
-            /* Overwrite rep codes */
-            for (i = 0; i < ZSTD_REP_NUM; i++)
-                seqStorePtr->rep[i] = repToConfirm[i];
-
-            /* Fill tables for block compressor */
-            ZSTD_ldm_limitTableUpdate(cctx, anchor);
-            ZSTD_ldm_fillFastTables(cctx, anchor);
-
-            /* Call block compressor and get remaining literals */
-            lastLiterals = blockCompressor(cctx, anchor, ip - anchor);
-            cctx->nextToUpdate = (U32)(ip - base);
-
-            /* Update repToConfirm with the new offset */
-            for (i = ZSTD_REP_NUM - 1; i > 0; i--)
-                repToConfirm[i] = repToConfirm[i-1];
-            repToConfirm[0] = offset;
-
-            /* Store the sequence with the leftover literals */
-            ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
-                          offset + ZSTD_REP_MOVE, mLength - MINMATCH);
+            U32 const offset = current - matchIndex;
+            rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
+
+            /* Out of sequence storage */
+            if (rawSeqStore->size == rawSeqStore->capacity)
+                return ERROR(dstSize_tooSmall);
+            seq->litLength = (U32)(ip - anchor);
+            seq->matchLength = (U32)mLength;
+            seq->offset = offset;
+            rawSeqStore->size++;
         }
 
         /* Insert the current entry into the hash table */
         ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
                                          (U32)(lastHashed - base),
-                                         ldmParams);
+                                         *params);
 
         assert(ip + backwardMatchLength == lastHashed);
 
         /* Fill the hash table from lastHashed+1 to ip+mLength*/
         /* Heuristic: don't need to fill the entire table at end of block */
-        if (ip + mLength < ilimit) {
+        if (ip + mLength <= ilimit) {
             rollingHash = ZSTD_ldm_fillLdmHashTable(
                               ldmState, rollingHash, lastHashed,
-                              ip + mLength, base, hBits, ldmParams);
+                              ip + mLength, base, hBits, *params);
             lastHashed = ip + mLength - 1;
         }
         ip += mLength;
         anchor = ip;
-        /* Check immediate repcode */
-        while ( (ip < ilimit)
-             && ( (repToConfirm[1] > 0) && (repToConfirm[1] <= (U32)(ip-lowest))
-             && (MEM_read32(ip) == MEM_read32(ip - repToConfirm[1])) )) {
-
-            size_t const rLength = ZSTD_count(ip+4, ip+4-repToConfirm[1],
-                                              iend) + 4;
-            /* Swap repToConfirm[1] <=> repToConfirm[0] */
-            {
-                U32 const tmpOff = repToConfirm[1];
-                repToConfirm[1] = repToConfirm[0];
-                repToConfirm[0] = tmpOff;
-            }
-
-            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
-
-            /* Fill the  hash table from lastHashed+1 to ip+rLength*/
-            if (ip + rLength < ilimit) {
-                rollingHash = ZSTD_ldm_fillLdmHashTable(
-                                ldmState, rollingHash, lastHashed,
-                                ip + rLength, base, hBits, ldmParams);
-                lastHashed = ip + rLength - 1;
-            }
-            ip += rLength;
-            anchor = ip;
-        }
     }
-
-    /* Overwrite rep */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        seqStorePtr->rep[i] = repToConfirm[i];
-
-    ZSTD_ldm_limitTableUpdate(cctx, anchor);
-    ZSTD_ldm_fillFastTables(cctx, anchor);
-
-    lastLiterals = blockCompressor(cctx, anchor, iend - anchor);
-    cctx->nextToUpdate = (U32)(iend - base);
-
-    /* Restore seqStorePtr->rep */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        seqStorePtr->rep[i] = savedRep[i];
-
-    /* Return the last literals size */
-    return lastLiterals;
+    return iend - anchor;
 }
 
-size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* ctx,
-                              const void* src, size_t srcSize)
+/*! ZSTD_ldm_reduceTable() :
+ *  reduce table indexes by `reducerValue` */
+static void ZSTD_ldm_reduceTable(ldmEntry_t* const table, U32 const size,
+                                 U32 const reducerValue)
 {
-    return ZSTD_compressBlock_ldm_generic(ctx, src, srcSize);
+    U32 u;
+    for (u = 0; u < size; u++) {
+        if (table[u].offset < reducerValue) table[u].offset = 0;
+        else table[u].offset -= reducerValue;
+    }
 }
 
-static size_t ZSTD_compressBlock_ldm_extDict_generic(
-                                 ZSTD_CCtx* ctx,
-                                 const void* src, size_t srcSize)
+size_t ZSTD_ldm_generateSequences(
+        ldmState_t* ldmState, rawSeqStore_t* sequences,
+        ldmParams_t const* params, void const* src, size_t srcSize)
 {
-    ldmState_t* const ldmState = &(ctx->ldmState);
-    const ldmParams_t ldmParams = ctx->appliedParams.ldmParams;
-    const U64 hashPower = ldmState->hashPower;
-    const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
-    const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
-    const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
-    seqStore_t* const seqStorePtr = &(ctx->seqStore);
-    const BYTE* const base = ctx->base;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32   lowestIndex = ctx->lowLimit;
-    const BYTE* const dictStart = dictBase + lowestIndex;
-    const U32   dictLimit = ctx->dictLimit;
-    const BYTE* const lowPrefixPtr = base + dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
-
-    const ZSTD_blockCompressor blockCompressor =
-        ZSTD_selectBlockCompressor(ctx->appliedParams.cParams.strategy, 1);
-    U32* const repToConfirm = seqStorePtr->repToConfirm;
-    U32 savedRep[ZSTD_REP_NUM];
-    U64 rollingHash = 0;
-    const BYTE* lastHashed = NULL;
-    size_t i, lastLiterals;
-
-    /* Save seqStorePtr->rep and copy repToConfirm */
-    for (i = 0; i < ZSTD_REP_NUM; i++) {
-        savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
-    }
-
-    /* Search Loop */
-    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
-        size_t mLength;
-        const U32 current = (U32)(ip-base);
-        size_t forwardMatchLength = 0, backwardMatchLength = 0;
-        ldmEntry_t* bestEntry = NULL;
-        if (ip != istart) {
-          rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
-                                       lastHashed[ldmParams.minMatchLength],
-                                       hashPower);
+    U32 const maxDist = 1U << params->windowLog;
+    BYTE const* const istart = (BYTE const*)src;
+    BYTE const* const iend = istart + srcSize;
+    size_t const kMaxChunkSize = 1 << 20;
+    size_t const nbChunks = (srcSize / kMaxChunkSize) + ((srcSize % kMaxChunkSize) != 0);
+    size_t chunk;
+    size_t leftoverSize = 0;
+
+    assert(ZSTD_CHUNKSIZE_MAX >= kMaxChunkSize);
+    /* Check that ZSTD_window_update() has been called for this chunk prior
+     * to passing it to this function.
+     */
+    assert(ldmState->window.nextSrc >= (BYTE const*)src + srcSize);
+    /* The input could be very large (in zstdmt), so it must be broken up into
+     * chunks to enforce the maximmum distance and handle overflow correction.
+     */
+    assert(sequences->pos <= sequences->size);
+    assert(sequences->size <= sequences->capacity);
+    for (chunk = 0; chunk < nbChunks && sequences->size < sequences->capacity; ++chunk) {
+        BYTE const* const chunkStart = istart + chunk * kMaxChunkSize;
+        size_t const remaining = (size_t)(iend - chunkStart);
+        BYTE const *const chunkEnd =
+            (remaining < kMaxChunkSize) ? iend : chunkStart + kMaxChunkSize;
+        size_t const chunkSize = chunkEnd - chunkStart;
+        size_t newLeftoverSize;
+        size_t const prevSize = sequences->size;
+
+        assert(chunkStart < iend);
+        /* 1. Perform overflow correction if necessary. */
+        if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) {
+            U32 const ldmHSize = 1U << params->hashLog;
+            U32 const correction = ZSTD_window_correctOverflow(
+                &ldmState->window, /* cycleLog */ 0, maxDist, src);
+            ZSTD_ldm_reduceTable(ldmState->hashTable, ldmHSize, correction);
+        }
+        /* 2. We enforce the maximum offset allowed.
+         *
+         * kMaxChunkSize should be small enough that we don't lose too much of
+         * the window through early invalidation.
+         * TODO: * Test the chunk size.
+         *       * Try invalidation after the sequence generation and test the
+         *         the offset against maxDist directly.
+         */
+        ZSTD_window_enforceMaxDist(&ldmState->window, chunkEnd, maxDist, NULL);
+        /* 3. Generate the sequences for the chunk, and get newLeftoverSize. */
+        newLeftoverSize = ZSTD_ldm_generateSequences_internal(
+            ldmState, sequences, params, chunkStart, chunkSize);
+        if (ZSTD_isError(newLeftoverSize))
+            return newLeftoverSize;
+        /* 4. We add the leftover literals from previous iterations to the first
+         *    newly generated sequence, or add the `newLeftoverSize` if none are
+         *    generated.
+         */
+        /* Prepend the leftover literals from the last call */
+        if (prevSize < sequences->size) {
+            sequences->seq[prevSize].litLength += (U32)leftoverSize;
+            leftoverSize = newLeftoverSize;
         } else {
-            rollingHash = ZSTD_ldm_getRollingHash(ip, ldmParams.minMatchLength);
+            assert(newLeftoverSize == chunkSize);
+            leftoverSize += chunkSize;
         }
-        lastHashed = ip;
+    }
+    return 0;
+}
 
-        if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
-                ldmTagMask) {
-            /* Don't insert and don't look for a match */
-           ip++;
-           continue;
+void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize, U32 const minMatch) {
+    while (srcSize > 0 && rawSeqStore->pos < rawSeqStore->size) {
+        rawSeq* seq = rawSeqStore->seq + rawSeqStore->pos;
+        if (srcSize <= seq->litLength) {
+            /* Skip past srcSize literals */
+            seq->litLength -= (U32)srcSize;
+            return;
         }
-
-        /* Get the best entry and compute the match lengths */
-        {
-            ldmEntry_t* const bucket =
-                ZSTD_ldm_getBucket(ldmState,
-                                   ZSTD_ldm_getSmallHash(rollingHash, hBits),
-                                   ldmParams);
-            ldmEntry_t* cur;
-            size_t bestMatchLength = 0;
-            U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
-
-            for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
-                const BYTE* const curMatchBase =
-                    cur->offset < dictLimit ? dictBase : base;
-                const BYTE* const pMatch = curMatchBase + cur->offset;
-                const BYTE* const matchEnd =
-                    cur->offset < dictLimit ? dictEnd : iend;
-                const BYTE* const lowMatchPtr =
-                    cur->offset < dictLimit ? dictStart : lowPrefixPtr;
-                size_t curForwardMatchLength, curBackwardMatchLength,
-                       curTotalMatchLength;
-
-                if (cur->checksum != checksum || cur->offset <= lowestIndex) {
-                    continue;
-                }
-
-                curForwardMatchLength = ZSTD_count_2segments(
-                                            ip, pMatch, iend,
-                                            matchEnd, lowPrefixPtr);
-                if (curForwardMatchLength < ldmParams.minMatchLength) {
-                    continue;
-                }
-                curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
-                                             ip, anchor, pMatch, lowMatchPtr);
-                curTotalMatchLength = curForwardMatchLength +
-                                      curBackwardMatchLength;
-
-                if (curTotalMatchLength > bestMatchLength) {
-                    bestMatchLength = curTotalMatchLength;
-                    forwardMatchLength = curForwardMatchLength;
-                    backwardMatchLength = curBackwardMatchLength;
-                    bestEntry = cur;
+        srcSize -= seq->litLength;
+        seq->litLength = 0;
+        if (srcSize < seq->matchLength) {
+            /* Skip past the first srcSize of the match */
+            seq->matchLength -= (U32)srcSize;
+            if (seq->matchLength < minMatch) {
+                /* The match is too short, omit it */
+                if (rawSeqStore->pos + 1 < rawSeqStore->size) {
+                    seq[1].litLength += seq[0].matchLength;
                 }
+                rawSeqStore->pos++;
             }
+            return;
         }
+        srcSize -= seq->matchLength;
+        seq->matchLength = 0;
+        rawSeqStore->pos++;
+    }
+}
 
-        /* No match found -- continue searching */
-        if (bestEntry == NULL) {
-            ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
-                                             (U32)(lastHashed - base),
-                                             ldmParams);
-            ip++;
-            continue;
+/**
+ * If the sequence length is longer than remaining then the sequence is split
+ * between this block and the next.
+ *
+ * Returns the current sequence to handle, or if the rest of the block should
+ * be literals, it returns a sequence with offset == 0.
+ */
+static rawSeq maybeSplitSequence(rawSeqStore_t* rawSeqStore,
+                                 U32 const remaining, U32 const minMatch)
+{
+    rawSeq sequence = rawSeqStore->seq[rawSeqStore->pos];
+    assert(sequence.offset > 0);
+    /* Likely: No partial sequence */
+    if (remaining >= sequence.litLength + sequence.matchLength) {
+        rawSeqStore->pos++;
+        return sequence;
+    }
+    /* Cut the sequence short (offset == 0 ==> rest is literals). */
+    if (remaining <= sequence.litLength) {
+        sequence.offset = 0;
+    } else if (remaining < sequence.litLength + sequence.matchLength) {
+        sequence.matchLength = remaining - sequence.litLength;
+        if (sequence.matchLength < minMatch) {
+            sequence.offset = 0;
         }
+    }
+    /* Skip past `remaining` bytes for the future sequences. */
+    ZSTD_ldm_skipSequences(rawSeqStore, remaining, minMatch);
+    return sequence;
+}
 
-        /* Match found */
-        mLength = forwardMatchLength + backwardMatchLength;
-        ip -= backwardMatchLength;
-
-        /* Call the block compressor on the remaining literals */
-        {
-            /* ip = current - backwardMatchLength
-             * The match is at (bestEntry->offset - backwardMatchLength) */
-            U32 const matchIndex = bestEntry->offset;
-            U32 const offset = current - matchIndex;
-
-            /* Overwrite rep codes */
-            for (i = 0; i < ZSTD_REP_NUM; i++)
-                seqStorePtr->rep[i] = repToConfirm[i];
-
-            /* Fill the hash table for the block compressor */
-            ZSTD_ldm_limitTableUpdate(ctx, anchor);
-            ZSTD_ldm_fillFastTables(ctx, anchor);
+size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
+    ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+    ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize,
+    int const extDict)
+{
+    unsigned const minMatch = cParams->searchLength;
+    ZSTD_blockCompressor const blockCompressor =
+        ZSTD_selectBlockCompressor(cParams->strategy, extDict);
+    BYTE const* const base = ms->window.base;
+    /* Input bounds */
+    BYTE const* const istart = (BYTE const*)src;
+    BYTE const* const iend = istart + srcSize;
+    /* Input positions */
+    BYTE const* ip = istart;
+
+    assert(rawSeqStore->pos <= rawSeqStore->size);
+    assert(rawSeqStore->size <= rawSeqStore->capacity);
+    /* Loop through each sequence and apply the block compressor to the lits */
+    while (rawSeqStore->pos < rawSeqStore->size && ip < iend) {
+        /* maybeSplitSequence updates rawSeqStore->pos */
+        rawSeq const sequence = maybeSplitSequence(rawSeqStore,
+                                                   (U32)(iend - ip), minMatch);
+        int i;
+        /* End signal */
+        if (sequence.offset == 0)
+            break;
 
-            /* Call block compressor and get remaining literals  */
-            lastLiterals = blockCompressor(ctx, anchor, ip - anchor);
-            ctx->nextToUpdate = (U32)(ip - base);
+        assert(sequence.offset <= (1U << cParams->windowLog));
+        assert(ip + sequence.litLength + sequence.matchLength <= iend);
 
-            /* Update repToConfirm with the new offset */
+        /* Fill tables for block compressor */
+        ZSTD_ldm_limitTableUpdate(ms, ip);
+        ZSTD_ldm_fillFastTables(ms, cParams, ip);
+        /* Run the block compressor */
+        {
+            size_t const newLitLength =
+                blockCompressor(ms, seqStore, rep, cParams, ip,
+                                sequence.litLength);
+            ip += sequence.litLength;
+            ms->nextToUpdate = (U32)(ip - base);
+            /* Update the repcodes */
             for (i = ZSTD_REP_NUM - 1; i > 0; i--)
-                repToConfirm[i] = repToConfirm[i-1];
-            repToConfirm[0] = offset;
-
-            /* Store the sequence with the leftover literals */
-            ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
-                          offset + ZSTD_REP_MOVE, mLength - MINMATCH);
-        }
-
-        /* Insert the current entry into the hash table */
-        ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
-                                         (U32)(lastHashed - base),
-                                         ldmParams);
-
-        /* Fill the hash table from lastHashed+1 to ip+mLength */
-        assert(ip + backwardMatchLength == lastHashed);
-        if (ip + mLength < ilimit) {
-            rollingHash = ZSTD_ldm_fillLdmHashTable(
-                              ldmState, rollingHash, lastHashed,
-                              ip + mLength, base, hBits,
-                              ldmParams);
-            lastHashed = ip + mLength - 1;
-        }
-        ip += mLength;
-        anchor = ip;
-
-        /* check immediate repcode */
-        while (ip < ilimit) {
-            U32 const current2 = (U32)(ip-base);
-            U32 const repIndex2 = current2 - repToConfirm[1];
-            const BYTE* repMatch2 = repIndex2 < dictLimit ?
-                                    dictBase + repIndex2 : base + repIndex2;
-            if ( (((U32)((dictLimit-1) - repIndex2) >= 3) &
-                        (repIndex2 > lowestIndex))  /* intentional overflow */
-               && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
-                const BYTE* const repEnd2 = repIndex2 < dictLimit ?
-                                            dictEnd : iend;
-                size_t const repLength2 =
-                        ZSTD_count_2segments(ip+4, repMatch2+4, iend,
-                                             repEnd2, lowPrefixPtr) + 4;
-
-                U32 tmpOffset = repToConfirm[1];
-                repToConfirm[1] = repToConfirm[0];
-                repToConfirm[0] = tmpOffset;
-
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
-
-                /* Fill the  hash table from lastHashed+1 to ip+repLength2*/
-                if (ip + repLength2 < ilimit) {
-                    rollingHash = ZSTD_ldm_fillLdmHashTable(
-                                      ldmState, rollingHash, lastHashed,
-                                      ip + repLength2, base, hBits,
-                                      ldmParams);
-                    lastHashed = ip + repLength2 - 1;
-                }
-                ip += repLength2;
-                anchor = ip;
-                continue;
-            }
-            break;
+                rep[i] = rep[i-1];
+            rep[0] = sequence.offset;
+            /* Store the sequence */
+            ZSTD_storeSeq(seqStore, newLitLength, ip - newLitLength,
+                          sequence.offset + ZSTD_REP_MOVE,
+                          sequence.matchLength - MINMATCH);
+            ip += sequence.matchLength;
         }
     }
-
-    /* Overwrite rep */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        seqStorePtr->rep[i] = repToConfirm[i];
-
-    ZSTD_ldm_limitTableUpdate(ctx, anchor);
-    ZSTD_ldm_fillFastTables(ctx, anchor);
-
-    /* Call the block compressor one last time on the last literals */
-    lastLiterals = blockCompressor(ctx, anchor, iend - anchor);
-    ctx->nextToUpdate = (U32)(iend - base);
-
-    /* Restore seqStorePtr->rep */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        seqStorePtr->rep[i] = savedRep[i];
-
-    /* Return the last literals size */
-    return lastLiterals;
-}
-
-size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
-                                      const void* src, size_t srcSize)
-{
-    return ZSTD_compressBlock_ldm_extDict_generic(ctx, src, srcSize);
+    /* Fill the tables for the block compressor */
+    ZSTD_ldm_limitTableUpdate(ms, ip);
+    ZSTD_ldm_fillFastTables(ms, cParams, ip);
+    /* Compress the last literals */
+    {
+        size_t const lastLiterals = blockCompressor(ms, seqStore, rep, cParams,
+                                                    ip, iend - ip);
+        ms->nextToUpdate = (U32)(iend - base);
+        return lastLiterals;
+    }
 }

src/borg/algorithms/zstd/lib/compress/zstd_ldm.h

@@ -10,43 +10,83 @@
 #ifndef ZSTD_LDM_H
 #define ZSTD_LDM_H
 
-#include "zstd_compress.h"
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
+#include "zstd_compress_internal.h"   /* ldmParams_t, U32 */
+#include "zstd.h"   /* ZSTD_CCtx, size_t */
+
 /*-*************************************
 *  Long distance matching
 ***************************************/
 
 #define ZSTD_LDM_DEFAULT_WINDOW_LOG ZSTD_WINDOWLOG_DEFAULTMAX
-#define ZSTD_LDM_HASHEVERYLOG_NOTSET 9999
 
-/** ZSTD_compressBlock_ldm_generic() :
+/**
+ * ZSTD_ldm_generateSequences():
  *
- *  This is a block compressor intended for long distance matching.
+ * Generates the sequences using the long distance match finder.
+ * Generates long range matching sequences in `sequences`, which parse a prefix
+ * of the source. `sequences` must be large enough to store every sequence,
+ * which can be checked with `ZSTD_ldm_getMaxNbSeq()`.
+ * @returns 0 or an error code.
  *
- *  The function searches for matches of length at least
- *  ldmParams.minMatchLength using a hash table in cctx->ldmState.
- *  Matches can be at a distance of up to cParams.windowLog.
+ * NOTE: The user must have called ZSTD_window_update() for all of the input
+ * they have, even if they pass it to ZSTD_ldm_generateSequences() in chunks.
+ * NOTE: This function returns an error if it runs out of space to store
+ *       sequences.
+ */
+size_t ZSTD_ldm_generateSequences(
+            ldmState_t* ldms, rawSeqStore_t* sequences,
+            ldmParams_t const* params, void const* src, size_t srcSize);
+
+/**
+ * ZSTD_ldm_blockCompress():
+ *
+ * Compresses a block using the predefined sequences, along with a secondary
+ * block compressor. The literals section of every sequence is passed to the
+ * secondary block compressor, and those sequences are interspersed with the
+ * predefined sequences. Returns the length of the last literals.
+ * Updates `rawSeqStore.pos` to indicate how many sequences have been consumed.
+ * `rawSeqStore.seq` may also be updated to split the last sequence between two
+ * blocks.
+ * @return The length of the last literals.
+ *
+ * NOTE: The source must be at most the maximum block size, but the predefined
+ * sequences can be any size, and may be longer than the block. In the case that
+ * they are longer than the block, the last sequences may need to be split into
+ * two. We handle that case correctly, and update `rawSeqStore` appropriately.
+ * NOTE: This function does not return any errors.
+ */
+size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
+            ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+            ZSTD_compressionParameters const* cParams,
+            void const* src, size_t srcSize,
+            int const extDict);
+
+/**
+ * ZSTD_ldm_skipSequences():
  *
- *  Upon finding a match, the unmatched literals are compressed using a
- *  ZSTD_blockCompressor (depending on the strategy in the compression
- *  parameters), which stores the matched sequences. The "long distance"
- *  match is then stored with the remaining literals from the
- *  ZSTD_blockCompressor. */
-size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* cctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
-                                      const void* src, size_t srcSize);
+ * Skip past `srcSize` bytes worth of sequences in `rawSeqStore`.
+ * Avoids emitting matches less than `minMatch` bytes.
+ * Must be called for data with is not passed to ZSTD_ldm_blockCompress().
+ */
+void ZSTD_ldm_skipSequences(rawSeqStore_t* rawSeqStore, size_t srcSize,
+    U32 const minMatch);
 
-/** ZSTD_ldm_initializeParameters() :
- *  Initialize the long distance matching parameters to their default values. */
-size_t ZSTD_ldm_initializeParameters(ldmParams_t* params, U32 enableLdm);
 
 /** ZSTD_ldm_getTableSize() :
- *  Estimate the space needed for long distance matching tables. */
-size_t ZSTD_ldm_getTableSize(U32 hashLog, U32 bucketSizeLog);
+ *  Estimate the space needed for long distance matching tables or 0 if LDM is
+ *  disabled.
+ */
+size_t ZSTD_ldm_getTableSize(ldmParams_t params);
+
+/** ZSTD_ldm_getSeqSpace() :
+ *  Return an upper bound on the number of sequences that can be produced by
+ *  the long distance matcher, or 0 if LDM is disabled.
+ */
+size_t ZSTD_ldm_getMaxNbSeq(ldmParams_t params, size_t maxChunkSize);
 
 /** ZSTD_ldm_getTableSize() :
  *  Return prime8bytes^(minMatchLength-1) */
@@ -57,8 +97,12 @@ U64 ZSTD_ldm_getHashPower(U32 minMatchLength);
  *  windowLog and params->hashLog.
  *
  *  Ensures that params->bucketSizeLog is <= params->hashLog (setting it to
- *  params->hashLog if it is not). */
-void ZSTD_ldm_adjustParameters(ldmParams_t* params, U32 windowLog);
+ *  params->hashLog if it is not).
+ *
+ *  Ensures that the minMatchLength >= targetLength during optimal parsing.
+ */
+void ZSTD_ldm_adjustParameters(ldmParams_t* params,
+                               ZSTD_compressionParameters const* cParams);
 
 #if defined (__cplusplus)
 }

src/borg/algorithms/zstd/lib/compress/zstd_opt.c

@@ -8,36 +8,34 @@
  * You may select, at your option, one of the above-listed licenses.
  */
 
+#include "zstd_compress_internal.h"
 #include "zstd_opt.h"
-#include "zstd_lazy.h"
 
 
-#define ZSTD_LITFREQ_ADD    2
-#define ZSTD_FREQ_DIV       4
-#define ZSTD_MAX_PRICE      (1<<30)
+#define ZSTD_LITFREQ_ADD    2   /* scaling factor for litFreq, so that frequencies adapt faster to new stats. Also used for matchSum (?) */
+#define ZSTD_FREQ_DIV       4   /* log factor when using previous stats to init next stats */
+#define ZSTD_MAX_PRICE     (1<<30)
+
 
 /*-*************************************
 *  Price functions for optimal parser
 ***************************************/
 static void ZSTD_setLog2Prices(optState_t* optPtr)
 {
-    optPtr->log2matchLengthSum = ZSTD_highbit32(optPtr->matchLengthSum+1);
-    optPtr->log2litLengthSum = ZSTD_highbit32(optPtr->litLengthSum+1);
     optPtr->log2litSum = ZSTD_highbit32(optPtr->litSum+1);
+    optPtr->log2litLengthSum = ZSTD_highbit32(optPtr->litLengthSum+1);
+    optPtr->log2matchLengthSum = ZSTD_highbit32(optPtr->matchLengthSum+1);
     optPtr->log2offCodeSum = ZSTD_highbit32(optPtr->offCodeSum+1);
-    optPtr->factor = 1 + ((optPtr->litSum>>5) / optPtr->litLengthSum) + ((optPtr->litSum<<1) / (optPtr->litSum + optPtr->matchSum));
 }
 
 
-static void ZSTD_rescaleFreqs(optState_t* optPtr, const BYTE* src, size_t srcSize)
+static void ZSTD_rescaleFreqs(optState_t* const optPtr,
+                              const BYTE* const src, size_t const srcSize)
 {
-    unsigned u;
-
-    optPtr->cachedLiterals = NULL;
-    optPtr->cachedPrice = optPtr->cachedLitLength = 0;
     optPtr->staticPrices = 0;
 
-    if (optPtr->litLengthSum == 0) {
+    if (optPtr->litLengthSum == 0) {  /* first init */
+        unsigned u;
         if (srcSize <= 1024) optPtr->staticPrices = 1;
 
         assert(optPtr->litFreq!=NULL);
@@ -45,44 +43,41 @@ static void ZSTD_rescaleFreqs(optState_t* optPtr, const BYTE* src, size_t srcSiz
             optPtr->litFreq[u] = 0;
         for (u=0; u<srcSize; u++)
             optPtr->litFreq[src[u]]++;
-
         optPtr->litSum = 0;
-        optPtr->litLengthSum = MaxLL+1;
-        optPtr->matchLengthSum = MaxML+1;
-        optPtr->offCodeSum = (MaxOff+1);
-        optPtr->matchSum = (ZSTD_LITFREQ_ADD<<Litbits);
-
         for (u=0; u<=MaxLit; u++) {
-            optPtr->litFreq[u] = 1 + (optPtr->litFreq[u]>>ZSTD_FREQ_DIV);
+            optPtr->litFreq[u] = 1 + (optPtr->litFreq[u] >> ZSTD_FREQ_DIV);
             optPtr->litSum += optPtr->litFreq[u];
         }
+
         for (u=0; u<=MaxLL; u++)
             optPtr->litLengthFreq[u] = 1;
+        optPtr->litLengthSum = MaxLL+1;
         for (u=0; u<=MaxML; u++)
             optPtr->matchLengthFreq[u] = 1;
+        optPtr->matchLengthSum = MaxML+1;
         for (u=0; u<=MaxOff; u++)
             optPtr->offCodeFreq[u] = 1;
+        optPtr->offCodeSum = (MaxOff+1);
+
     } else {
-        optPtr->matchLengthSum = 0;
-        optPtr->litLengthSum = 0;
-        optPtr->offCodeSum = 0;
-        optPtr->matchSum = 0;
-        optPtr->litSum = 0;
+        unsigned u;
 
+        optPtr->litSum = 0;
         for (u=0; u<=MaxLit; u++) {
-            optPtr->litFreq[u] = 1 + (optPtr->litFreq[u]>>(ZSTD_FREQ_DIV+1));
+            optPtr->litFreq[u] = 1 + (optPtr->litFreq[u] >> (ZSTD_FREQ_DIV+1));
             optPtr->litSum += optPtr->litFreq[u];
         }
+        optPtr->litLengthSum = 0;
         for (u=0; u<=MaxLL; u++) {
             optPtr->litLengthFreq[u] = 1 + (optPtr->litLengthFreq[u]>>(ZSTD_FREQ_DIV+1));
             optPtr->litLengthSum += optPtr->litLengthFreq[u];
         }
+        optPtr->matchLengthSum = 0;
         for (u=0; u<=MaxML; u++) {
             optPtr->matchLengthFreq[u] = 1 + (optPtr->matchLengthFreq[u]>>ZSTD_FREQ_DIV);
             optPtr->matchLengthSum += optPtr->matchLengthFreq[u];
-            optPtr->matchSum += optPtr->matchLengthFreq[u] * (u + 3);
         }
-        optPtr->matchSum *= ZSTD_LITFREQ_ADD;
+        optPtr->offCodeSum = 0;
         for (u=0; u<=MaxOff; u++) {
             optPtr->offCodeFreq[u] = 1 + (optPtr->offCodeFreq[u]>>ZSTD_FREQ_DIV);
             optPtr->offCodeSum += optPtr->offCodeFreq[u];
@@ -93,114 +88,146 @@ static void ZSTD_rescaleFreqs(optState_t* optPtr, const BYTE* src, size_t srcSiz
 }
 
 
-static U32 ZSTD_getLiteralPrice(optState_t* optPtr, U32 litLength, const BYTE* literals)
+/* ZSTD_rawLiteralsCost() :
+ * cost of literals (only) in given segment (which length can be null)
+ * does not include cost of literalLength symbol */
+static U32 ZSTD_rawLiteralsCost(const BYTE* const literals, U32 const litLength,
+                                const optState_t* const optPtr)
 {
-    U32 price, u;
-
-    if (optPtr->staticPrices)
-        return ZSTD_highbit32((U32)litLength+1) + (litLength*6);
-
-    if (litLength == 0)
-        return optPtr->log2litLengthSum - ZSTD_highbit32(optPtr->litLengthFreq[0]+1);
+    if (optPtr->staticPrices) return (litLength*6);  /* 6 bit per literal - no statistic used */
+    if (litLength == 0) return 0;
 
     /* literals */
-    if (optPtr->cachedLiterals == literals) {
-        U32 const additional = litLength - optPtr->cachedLitLength;
-        const BYTE* literals2 = optPtr->cachedLiterals + optPtr->cachedLitLength;
-        price = optPtr->cachedPrice + additional * optPtr->log2litSum;
-        for (u=0; u < additional; u++)
-            price -= ZSTD_highbit32(optPtr->litFreq[literals2[u]]+1);
-        optPtr->cachedPrice = price;
-        optPtr->cachedLitLength = litLength;
-    } else {
-        price = litLength * optPtr->log2litSum;
+    {   U32 u;
+        U32 cost = litLength * optPtr->log2litSum;
         for (u=0; u < litLength; u++)
-            price -= ZSTD_highbit32(optPtr->litFreq[literals[u]]+1);
-
-        if (litLength >= 12) {
-            optPtr->cachedLiterals = literals;
-            optPtr->cachedPrice = price;
-            optPtr->cachedLitLength = litLength;
-        }
+            cost -= ZSTD_highbit32(optPtr->litFreq[literals[u]]+1);
+        return cost;
     }
+}
+
+/* ZSTD_litLengthPrice() :
+ * cost of literalLength symbol */
+static U32 ZSTD_litLengthPrice(U32 const litLength, const optState_t* const optPtr)
+{
+    if (optPtr->staticPrices) return ZSTD_highbit32((U32)litLength+1);
 
     /* literal Length */
-    {   const BYTE LL_deltaCode = 19;
-        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
-        price += LL_bits[llCode] + optPtr->log2litLengthSum - ZSTD_highbit32(optPtr->litLengthFreq[llCode]+1);
+    {   U32 const llCode = ZSTD_LLcode(litLength);
+        U32 const price = LL_bits[llCode] + optPtr->log2litLengthSum - ZSTD_highbit32(optPtr->litLengthFreq[llCode]+1);
+        return price;
     }
+}
 
-    return price;
+/* ZSTD_litLengthPrice() :
+ * cost of the literal part of a sequence,
+ * including literals themselves, and literalLength symbol */
+static U32 ZSTD_fullLiteralsCost(const BYTE* const literals, U32 const litLength,
+                                 const optState_t* const optPtr)
+{
+    return ZSTD_rawLiteralsCost(literals, litLength, optPtr)
+         + ZSTD_litLengthPrice(litLength, optPtr);
 }
 
+/* ZSTD_litLengthContribution() :
+ * @return ( cost(litlength) - cost(0) )
+ * this value can then be added to rawLiteralsCost()
+ * to provide a cost which is directly comparable to a match ending at same position */
+static int ZSTD_litLengthContribution(U32 const litLength, const optState_t* const optPtr)
+{
+    if (optPtr->staticPrices) return ZSTD_highbit32(litLength+1);
 
-FORCE_INLINE_TEMPLATE U32 ZSTD_getPrice(optState_t* optPtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength, const int ultra)
+    /* literal Length */
+    {   U32 const llCode = ZSTD_LLcode(litLength);
+        int const contribution = LL_bits[llCode]
+                        + ZSTD_highbit32(optPtr->litLengthFreq[0]+1)
+                        - ZSTD_highbit32(optPtr->litLengthFreq[llCode]+1);
+#if 1
+        return contribution;
+#else
+        return MAX(0, contribution); /* sometimes better, sometimes not ... */
+#endif
+    }
+}
+
+/* ZSTD_literalsContribution() :
+ * creates a fake cost for the literals part of a sequence
+ * which can be compared to the ending cost of a match
+ * should a new match start at this position */
+static int ZSTD_literalsContribution(const BYTE* const literals, U32 const litLength,
+                                     const optState_t* const optPtr)
+{
+    int const contribution = ZSTD_rawLiteralsCost(literals, litLength, optPtr)
+                           + ZSTD_litLengthContribution(litLength, optPtr);
+    return contribution;
+}
+
+/* ZSTD_getMatchPrice() :
+ * Provides the cost of the match part (offset + matchLength) of a sequence
+ * Must be combined with ZSTD_fullLiteralsCost() to get the full cost of a sequence.
+ * optLevel: when <2, favors small offset for decompression speed (improved cache efficiency) */
+FORCE_INLINE_TEMPLATE U32 ZSTD_getMatchPrice(
+                                    U32 const offset, U32 const matchLength,
+                                    const optState_t* const optPtr,
+                                    int const optLevel)
 {
-    /* offset */
     U32 price;
-    BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
+    U32 const offCode = ZSTD_highbit32(offset+1);
+    U32 const mlBase = matchLength - MINMATCH;
+    assert(matchLength >= MINMATCH);
 
-    if (optPtr->staticPrices)
-        return ZSTD_getLiteralPrice(optPtr, litLength, literals) + ZSTD_highbit32((U32)matchLength+1) + 16 + offCode;
+    if (optPtr->staticPrices)  /* fixed scheme, do not use statistics */
+        return ZSTD_highbit32((U32)mlBase+1) + 16 + offCode;
 
     price = offCode + optPtr->log2offCodeSum - ZSTD_highbit32(optPtr->offCodeFreq[offCode]+1);
-    if (!ultra && offCode >= 20) price += (offCode-19)*2;
+    if ((optLevel<2) /*static*/ && offCode >= 20) price += (offCode-19)*2; /* handicap for long distance offsets, favor decompression speed */
 
     /* match Length */
-    {   const BYTE ML_deltaCode = 36;
-        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
+    {   U32 const mlCode = ZSTD_MLcode(mlBase);
         price += ML_bits[mlCode] + optPtr->log2matchLengthSum - ZSTD_highbit32(optPtr->matchLengthFreq[mlCode]+1);
     }
 
-    return price + ZSTD_getLiteralPrice(optPtr, litLength, literals) + optPtr->factor;
+    DEBUGLOG(8, "ZSTD_getMatchPrice(ml:%u) = %u", matchLength, price);
+    return price;
 }
 
-
-static void ZSTD_updatePrice(optState_t* optPtr, U32 litLength, const BYTE* literals, U32 offset, U32 matchLength)
+static void ZSTD_updateStats(optState_t* const optPtr,
+                             U32 litLength, const BYTE* literals,
+                             U32 offsetCode, U32 matchLength)
 {
-    U32 u;
-
     /* literals */
-    optPtr->litSum += litLength*ZSTD_LITFREQ_ADD;
-    for (u=0; u < litLength; u++)
-        optPtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
+    {   U32 u;
+        for (u=0; u < litLength; u++)
+            optPtr->litFreq[literals[u]] += ZSTD_LITFREQ_ADD;
+        optPtr->litSum += litLength*ZSTD_LITFREQ_ADD;
+    }
 
     /* literal Length */
-    {   const BYTE LL_deltaCode = 19;
-        const BYTE llCode = (litLength>63) ? (BYTE)ZSTD_highbit32(litLength) + LL_deltaCode : LL_Code[litLength];
+    {   U32 const llCode = ZSTD_LLcode(litLength);
         optPtr->litLengthFreq[llCode]++;
         optPtr->litLengthSum++;
     }
 
-    /* match offset */
-    {   BYTE const offCode = (BYTE)ZSTD_highbit32(offset+1);
-        optPtr->offCodeSum++;
+    /* match offset code (0-2=>repCode; 3+=>offset+2) */
+    {   U32 const offCode = ZSTD_highbit32(offsetCode+1);
+        assert(offCode <= MaxOff);
         optPtr->offCodeFreq[offCode]++;
+        optPtr->offCodeSum++;
     }
 
     /* match Length */
-    {   const BYTE ML_deltaCode = 36;
-        const BYTE mlCode = (matchLength>127) ? (BYTE)ZSTD_highbit32(matchLength) + ML_deltaCode : ML_Code[matchLength];
+    {   U32 const mlBase = matchLength - MINMATCH;
+        U32 const mlCode = ZSTD_MLcode(mlBase);
         optPtr->matchLengthFreq[mlCode]++;
         optPtr->matchLengthSum++;
     }
-
-    ZSTD_setLog2Prices(optPtr);
 }
 
 
-#define SET_PRICE(pos, mlen_, offset_, litlen_, price_)   \
-    {                                                 \
-        while (last_pos < pos)  { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; } \
-        opt[pos].mlen = mlen_;                         \
-        opt[pos].off = offset_;                        \
-        opt[pos].litlen = litlen_;                     \
-        opt[pos].price = price_;                       \
-    }
-
-
-/* function safe only for comparisons */
-static U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
+/* ZSTD_readMINMATCH() :
+ * function safe only for comparisons
+ * assumption : memPtr must be at least 4 bytes before end of buffer */
+MEM_STATIC U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
 {
     switch (length)
     {
@@ -216,15 +243,15 @@ static U32 ZSTD_readMINMATCH(const void* memPtr, U32 length)
 
 /* Update hashTable3 up to ip (excluded)
    Assumption : always within prefix (i.e. not within extDict) */
-static
-U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* zc, const BYTE* ip)
+static U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_matchState_t* ms, const BYTE* const ip)
 {
-    U32* const hashTable3  = zc->hashTable3;
-    U32 const hashLog3  = zc->hashLog3;
-    const BYTE* const base = zc->base;
-    U32 idx = zc->nextToUpdate3;
-    const U32 target = zc->nextToUpdate3 = (U32)(ip - base);
-    const size_t hash3 = ZSTD_hash3Ptr(ip, hashLog3);
+    U32* const hashTable3 = ms->hashTable3;
+    U32 const hashLog3 = ms->hashLog3;
+    const BYTE* const base = ms->window.base;
+    U32 idx = ms->nextToUpdate3;
+    U32 const target = ms->nextToUpdate3 = (U32)(ip - base);
+    size_t const hash3 = ZSTD_hash3Ptr(ip, hashLog3);
+    assert(hashLog3 > 0);
 
     while(idx < target) {
         hashTable3[ZSTD_hash3Ptr(base+idx, hashLog3)] = idx;
@@ -238,102 +265,284 @@ U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* zc, const BYTE* ip)
 /*-*************************************
 *  Binary Tree search
 ***************************************/
-static U32 ZSTD_insertBtAndGetAllMatches (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        U32 nbCompares, const U32 mls,
-                        U32 extDict, ZSTD_match_t* matches, const U32 minMatchLen)
+/** ZSTD_insertBt1() : add one or multiple positions to tree.
+ *  ip : assumed <= iend-8 .
+ * @return : nb of positions added */
+static U32 ZSTD_insertBt1(
+                ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                const BYTE* const ip, const BYTE* const iend,
+                U32 const mls, U32 const extDict)
 {
-    const BYTE* const base = zc->base;
+    U32*   const hashTable = ms->hashTable;
+    U32    const hashLog = cParams->hashLog;
+    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
+    U32*   const bt = ms->chainTable;
+    U32    const btLog  = cParams->chainLog - 1;
+    U32    const btMask = (1 << btLog) - 1;
+    U32 matchIndex = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const dictBase = ms->window.dictBase;
+    const U32 dictLimit = ms->window.dictLimit;
+    const BYTE* const dictEnd = dictBase + dictLimit;
+    const BYTE* const prefixStart = base + dictLimit;
+    const BYTE* match;
     const U32 current = (U32)(ip-base);
-    const U32 hashLog = zc->appliedParams.cParams.hashLog;
-    const size_t h  = ZSTD_hashPtr(ip, hashLog, mls);
-    U32* const hashTable = zc->hashTable;
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = smallerPtr + 1;
+    U32 dummy32;   /* to be nullified at the end */
+    U32 const windowLow = ms->window.lowLimit;
+    U32 matchEndIdx = current+8+1;
+    size_t bestLength = 8;
+    U32 nbCompares = 1U << cParams->searchLog;
+#ifdef ZSTD_C_PREDICT
+    U32 predictedSmall = *(bt + 2*((current-1)&btMask) + 0);
+    U32 predictedLarge = *(bt + 2*((current-1)&btMask) + 1);
+    predictedSmall += (predictedSmall>0);
+    predictedLarge += (predictedLarge>0);
+#endif /* ZSTD_C_PREDICT */
+
+    DEBUGLOG(8, "ZSTD_insertBt1 (%u)", current);
+
+    assert(ip <= iend-8);   /* required for h calculation */
+    hashTable[h] = current;   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow)) {
+        U32* const nextPtr = bt + 2*(matchIndex & btMask);
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+        assert(matchIndex < current);
+
+#ifdef ZSTD_C_PREDICT   /* note : can create issues when hlog small <= 11 */
+        const U32* predictPtr = bt + 2*((matchIndex-1) & btMask);   /* written this way, as bt is a roll buffer */
+        if (matchIndex == predictedSmall) {
+            /* no need to check length, result known */
+            *smallerPtr = matchIndex;
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            predictedSmall = predictPtr[1] + (predictPtr[1]>0);
+            continue;
+        }
+        if (matchIndex == predictedLarge) {
+            *largerPtr = matchIndex;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+            predictedLarge = predictPtr[0] + (predictPtr[0]>0);
+            continue;
+        }
+#endif
+
+        if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            assert(matchIndex+matchLength >= dictLimit);   /* might be wrong if extDict is incorrectly set to 0 */
+            match = base + matchIndex;
+            matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+        } else {
+            match = dictBase + matchIndex;
+            matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iend, dictEnd, prefixStart);
+            if (matchIndex+matchLength >= dictLimit)
+                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+        }
+
+        if (matchLength > bestLength) {
+            bestLength = matchLength;
+            if (matchLength > matchEndIdx - matchIndex)
+                matchEndIdx = matchIndex + (U32)matchLength;
+        }
+
+        if (ip+matchLength == iend) {   /* equal : no way to know if inf or sup */
+            break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
+        }
+
+        if (match[matchLength] < ip[matchLength]) {  /* necessarily within buffer */
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
+            smallerPtr = nextPtr+1;               /* new "candidate" => larger than match, which was smaller than target */
+            matchIndex = nextPtr[1];              /* new matchIndex, larger than previous and closer to current */
+        } else {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop searching */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+    }   }
+
+    *smallerPtr = *largerPtr = 0;
+    if (bestLength > 384) return MIN(192, (U32)(bestLength - 384));   /* speed optimization */
+    assert(matchEndIdx > current + 8);
+    return matchEndIdx - (current + 8);
+}
+
+FORCE_INLINE_TEMPLATE
+void ZSTD_updateTree_internal(
+                ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                const BYTE* const ip, const BYTE* const iend,
+                const U32 mls, const U32 extDict)
+{
+    const BYTE* const base = ms->window.base;
+    U32 const target = (U32)(ip - base);
+    U32 idx = ms->nextToUpdate;
+    DEBUGLOG(7, "ZSTD_updateTree_internal, from %u to %u  (extDict:%u)",
+                idx, target, extDict);
+
+    while(idx < target)
+        idx += ZSTD_insertBt1(ms, cParams, base+idx, iend, mls, extDict);
+    ms->nextToUpdate = target;
+}
+
+void ZSTD_updateTree(
+                ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                const BYTE* ip, const BYTE* iend)
+{
+    ZSTD_updateTree_internal(ms, cParams, ip, iend, cParams->searchLength, 0 /*extDict*/);
+}
+
+FORCE_INLINE_TEMPLATE
+U32 ZSTD_insertBtAndGetAllMatches (
+                    ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                    const BYTE* const ip, const BYTE* const iLimit, int const extDict,
+                    U32 rep[ZSTD_REP_NUM], U32 const ll0,
+                    ZSTD_match_t* matches, const U32 lengthToBeat, U32 const mls /* template */)
+{
+    U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
+    const BYTE* const base = ms->window.base;
+    U32 const current = (U32)(ip-base);
+    U32 const hashLog = cParams->hashLog;
+    U32 const minMatch = (mls==3) ? 3 : 4;
+    U32* const hashTable = ms->hashTable;
+    size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
     U32 matchIndex  = hashTable[h];
-    U32* const bt   = zc->chainTable;
-    const U32 btLog = zc->appliedParams.cParams.chainLog - 1;
-    const U32 btMask= (1U << btLog) - 1;
+    U32* const bt   = ms->chainTable;
+    U32 const btLog = cParams->chainLog - 1;
+    U32 const btMask= (1U << btLog) - 1;
     size_t commonLengthSmaller=0, commonLengthLarger=0;
-    const BYTE* const dictBase = zc->dictBase;
-    const U32 dictLimit = zc->dictLimit;
+    const BYTE* const dictBase = ms->window.dictBase;
+    U32 const dictLimit = ms->window.dictLimit;
     const BYTE* const dictEnd = dictBase + dictLimit;
     const BYTE* const prefixStart = base + dictLimit;
-    const U32 btLow = btMask >= current ? 0 : current - btMask;
-    const U32 windowLow = zc->lowLimit;
+    U32 const btLow = btMask >= current ? 0 : current - btMask;
+    U32 const windowLow = ms->window.lowLimit;
     U32* smallerPtr = bt + 2*(current&btMask);
     U32* largerPtr  = bt + 2*(current&btMask) + 1;
-    U32 matchEndIdx = current+8;
+    U32 matchEndIdx = current+8+1;   /* farthest referenced position of any match => detects repetitive patterns */
     U32 dummy32;   /* to be nullified at the end */
     U32 mnum = 0;
-
-    const U32 minMatch = (mls == 3) ? 3 : 4;
-    size_t bestLength = minMatchLen-1;
-
-    if (minMatch == 3) { /* HC3 match finder */
-        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3 (zc, ip);
-        if (matchIndex3>windowLow && (current - matchIndex3 < (1<<18))) {
-            const BYTE* match;
-            size_t currentMl=0;
-            if ((!extDict) || matchIndex3 >= dictLimit) {
-                match = base + matchIndex3;
-                if (match[bestLength] == ip[bestLength]) currentMl = ZSTD_count(ip, match, iLimit);
+    U32 nbCompares = 1U << cParams->searchLog;
+
+    size_t bestLength = lengthToBeat-1;
+    DEBUGLOG(7, "ZSTD_insertBtAndGetAllMatches");
+
+    /* check repCode */
+    {   U32 const lastR = ZSTD_REP_NUM + ll0;
+        U32 repCode;
+        for (repCode = ll0; repCode < lastR; repCode++) {
+            U32 const repOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+            U32 const repIndex = current - repOffset;
+            U32 repLen = 0;
+            assert(current >= dictLimit);
+            if (repOffset-1 /* intentional overflow, discards 0 and -1 */ < current-dictLimit) {  /* equivalent to `current > repIndex >= dictLimit` */
+                if (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repOffset, minMatch)) {
+                    repLen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repOffset, iLimit) + minMatch;
+                }
+            } else {  /* repIndex < dictLimit || repIndex >= current */
+                const BYTE* const repMatch = dictBase + repIndex;
+                assert(current >= windowLow);
+                if ( extDict /* this case only valid in extDict mode */
+                  && ( ((repOffset-1) /*intentional overflow*/ < current - windowLow)  /* equivalent to `current > repIndex >= windowLow` */
+                     & (((U32)((dictLimit-1) - repIndex) >= 3) ) /* intentional overflow : do not test positions overlapping 2 memory segments */)
+                  && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
+                    repLen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iLimit, dictEnd, prefixStart) + minMatch;
+            }   }
+            /* save longer solution */
+            if (repLen > bestLength) {
+                DEBUGLOG(8, "found rep-match %u of length %u",
+                            repCode - ll0, (U32)repLen);
+                bestLength = repLen;
+                matches[mnum].off = repCode - ll0;
+                matches[mnum].len = (U32)repLen;
+                mnum++;
+                if ( (repLen > sufficient_len)
+                   | (ip+repLen == iLimit) ) {  /* best possible */
+                    return mnum;
+    }   }   }   }
+
+    /* HC3 match finder */
+    if ((mls == 3) /*static*/ && (bestLength < mls)) {
+        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3(ms, ip);
+        if ((matchIndex3 > windowLow)
+          & (current - matchIndex3 < (1<<18)) /*heuristic : longer distance likely too expensive*/ ) {
+            size_t mlen;
+            if ((!extDict) /*static*/ || (matchIndex3 >= dictLimit)) {
+                const BYTE* const match = base + matchIndex3;
+                mlen = ZSTD_count(ip, match, iLimit);
             } else {
-                match = dictBase + matchIndex3;
-                if (ZSTD_readMINMATCH(match, MINMATCH) == ZSTD_readMINMATCH(ip, MINMATCH))    /* assumption : matchIndex3 <= dictLimit-4 (by table construction) */
-                    currentMl = ZSTD_count_2segments(ip+MINMATCH, match+MINMATCH, iLimit, dictEnd, prefixStart) + MINMATCH;
+                const BYTE* const match = dictBase + matchIndex3;
+                mlen = ZSTD_count_2segments(ip, match, iLimit, dictEnd, prefixStart);
             }
 
             /* save best solution */
-            if (currentMl > bestLength) {
-                bestLength = currentMl;
-                matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex3;
-                matches[mnum].len = (U32)currentMl;
-                mnum++;
-                if (currentMl > ZSTD_OPT_NUM) goto update;
-                if (ip+currentMl == iLimit) goto update; /* best possible, and avoid read overflow*/
-            }
-        }
-    }
+            if (mlen >= mls /* == 3 > bestLength */) {
+                DEBUGLOG(8, "found small match with hlog3, of length %u",
+                            (U32)mlen);
+                bestLength = mlen;
+                assert(current > matchIndex3);
+                assert(mnum==0);  /* no prior solution */
+                matches[0].off = (current - matchIndex3) + ZSTD_REP_MOVE;
+                matches[0].len = (U32)mlen;
+                mnum = 1;
+                if ( (mlen > sufficient_len) |
+                     (ip+mlen == iLimit) ) {  /* best possible length */
+                    ms->nextToUpdate = current+1;  /* skip insertion */
+                    return 1;
+    }   }   }   }
 
     hashTable[h] = current;   /* Update Hash Table */
 
     while (nbCompares-- && (matchIndex > windowLow)) {
-        U32* nextPtr = bt + 2*(matchIndex & btMask);
+        U32* const nextPtr = bt + 2*(matchIndex & btMask);
         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
         const BYTE* match;
+        assert(current > matchIndex);
 
         if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            assert(matchIndex+matchLength >= dictLimit);  /* ensure the condition is correct when !extDict */
             match = base + matchIndex;
-            if (match[matchLength] == ip[matchLength]) {
-                matchLength += ZSTD_count(ip+matchLength+1, match+matchLength+1, iLimit) +1;
-            }
+            matchLength += ZSTD_count(ip+matchLength, match+matchLength, iLimit);
         } else {
             match = dictBase + matchIndex;
             matchLength += ZSTD_count_2segments(ip+matchLength, match+matchLength, iLimit, dictEnd, prefixStart);
             if (matchIndex+matchLength >= dictLimit)
-                match = base + matchIndex;   /* to prepare for next usage of match[matchLength] */
+                match = base + matchIndex;   /* prepare for match[matchLength] */
         }
 
         if (matchLength > bestLength) {
-            if (matchLength > matchEndIdx - matchIndex) matchEndIdx = matchIndex + (U32)matchLength;
+            DEBUGLOG(8, "found match of length %u at distance %u",
+                    (U32)matchLength, current - matchIndex);
+            assert(matchEndIdx > matchIndex);
+            if (matchLength > matchEndIdx - matchIndex)
+                matchEndIdx = matchIndex + (U32)matchLength;
             bestLength = matchLength;
-            matches[mnum].off = ZSTD_REP_MOVE_OPT + current - matchIndex;
+            matches[mnum].off = (current - matchIndex) + ZSTD_REP_MOVE;
             matches[mnum].len = (U32)matchLength;
             mnum++;
             if (matchLength > ZSTD_OPT_NUM) break;
-            if (ip+matchLength == iLimit)   /* equal : no way to know if inf or sup */
-                break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+            if (ip+matchLength == iLimit) {  /* equal : no way to know if inf or sup */
+                break;   /* drop, to preserve bt consistency (miss a little bit of compression) */
+            }
         }
 
         if (match[matchLength] < ip[matchLength]) {
-            /* match is smaller than current */
+            /* match smaller than current */
             *smallerPtr = matchIndex;             /* update smaller idx */
             commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
             if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            smallerPtr = nextPtr+1;               /* new candidate => larger than match, which was smaller than current */
+            matchIndex = nextPtr[1];              /* new matchIndex, larger than previous, closer to current */
         } else {
-            /* match is larger than current */
             *largerPtr = matchIndex;
             commonLengthLarger = matchLength;
             if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
@@ -343,65 +552,30 @@ static U32 ZSTD_insertBtAndGetAllMatches (
 
     *smallerPtr = *largerPtr = 0;
 
-update:
-    zc->nextToUpdate = (matchEndIdx > current + 8) ? matchEndIdx - 8 : current+1;
+    assert(matchEndIdx > current+8);
+    ms->nextToUpdate = matchEndIdx - 8;  /* skip repetitive patterns */
     return mnum;
 }
 
 
-/** Tree updater, providing best match */
-static U32 ZSTD_BtGetAllMatches (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 0, matches, minMatchLen);
-}
-
-
-static U32 ZSTD_BtGetAllMatches_selectMLS (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* ip, const BYTE* const iHighLimit,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
+FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip, const BYTE* const iHighLimit, int const extDict,
+                        U32 rep[ZSTD_REP_NUM], U32 const ll0,
+                        ZSTD_match_t* matches, U32 const lengthToBeat)
 {
+    U32 const matchLengthSearch = cParams->searchLength;
+    DEBUGLOG(7, "ZSTD_BtGetAllMatches");
+    if (ip < ms->window.base + ms->nextToUpdate) return 0;   /* skipped area */
+    ZSTD_updateTree_internal(ms, cParams, ip, iHighLimit, matchLengthSearch, extDict);
     switch(matchLengthSearch)
     {
-    case 3 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
+    case 3 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 3);
     default :
-    case 4 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
-    case 5 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
+    case 4 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 4);
+    case 5 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 5);
     case 7 :
-    case 6 : return ZSTD_BtGetAllMatches(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
-    }
-}
-
-/** Tree updater, providing best match */
-static U32 ZSTD_BtGetAllMatches_extDict (
-                        ZSTD_CCtx* zc,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        const U32 maxNbAttempts, const U32 mls, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    if (ip < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    ZSTD_updateTree_extDict(zc, ip, iLimit, maxNbAttempts, mls);
-    return ZSTD_insertBtAndGetAllMatches(zc, ip, iLimit, maxNbAttempts, mls, 1, matches, minMatchLen);
-}
-
-
-static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
-                        const BYTE* ip, const BYTE* const iHighLimit,
-                        const U32 maxNbAttempts, const U32 matchLengthSearch, ZSTD_match_t* matches, const U32 minMatchLen)
-{
-    switch(matchLengthSearch)
-    {
-    case 3 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 3, matches, minMatchLen);
-    default :
-    case 4 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 4, matches, minMatchLen);
-    case 5 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 5, matches, minMatchLen);
-    case 7 :
-    case 6 : return ZSTD_BtGetAllMatches_extDict(zc, ip, iHighLimit, maxNbAttempts, 6, matches, minMatchLen);
+    case 6 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 6);
     }
 }
 
@@ -409,549 +583,341 @@ static U32 ZSTD_BtGetAllMatches_selectMLS_extDict (
 /*-*******************************
 *  Optimal parser
 *********************************/
-FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
-                                      const void* src, size_t srcSize, const int ultra)
-{
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    optState_t* optStatePtr = &(ctx->optState);
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base;
-    const BYTE* const prefixStart = base + ctx->dictLimit;
-
-    const U32 maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
-    const U32 sufficient_len = ctx->appliedParams.cParams.targetLength;
-    const U32 mls = ctx->appliedParams.cParams.searchLength;
-    const U32 minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
-
-    ZSTD_optimal_t* opt = optStatePtr->priceTable;
-    ZSTD_match_t* matches = optStatePtr->matchTable;
-    const BYTE* inr;
-    U32 offset, rep[ZSTD_REP_NUM];
-
-    /* init */
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
-    ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize);
-    ip += (ip==prefixStart);
-    { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=seqStorePtr->rep[i]; }
-
-    /* Match Loop */
-    while (ip < ilimit) {
-        U32 cur, match_num, last_pos, litlen, price;
-        U32 u, mlen, best_mlen, best_off, litLength;
-        memset(opt, 0, sizeof(ZSTD_optimal_t));
-        last_pos = 0;
-        litlen = (U32)(ip - anchor);
-
-        /* check repCode */
-        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
-            for (i=(ip == anchor); i<last_i; i++) {
-                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
-                if ( (repCur > 0) && (repCur < (S32)(ip-prefixStart))
-                    && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(ip - repCur, minMatch))) {
-                    mlen = (U32)ZSTD_count(ip+minMatch, ip+minMatch-repCur, iend) + minMatch;
-                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
-                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
-                        goto _storeSequence;
-                    }
-                    best_off = i - (ip == anchor);
-                    do {
-                        price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                        if (mlen > last_pos || price < opt[mlen].price)
-                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
-                        mlen--;
-                    } while (mlen >= minMatch);
-        }   }   }
-
-        match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, ip, iend, maxSearches, mls, matches, minMatch);
-
-        if (!last_pos && !match_num) { ip++; continue; }
-
-        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-            best_mlen = matches[match_num-1].len;
-            best_off = matches[match_num-1].off;
-            cur = 0;
-            last_pos = 1;
-            goto _storeSequence;
-        }
-
-        /* set prices using matches at position = 0 */
-        best_mlen = (last_pos) ? last_pos : minMatch;
-        for (u = 0; u < match_num; u++) {
-            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-            best_mlen = matches[u].len;
-            while (mlen <= best_mlen) {
-                price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                if (mlen > last_pos || price < opt[mlen].price)
-                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);   /* note : macro modifies last_pos */
-                mlen++;
-        }   }
-
-        if (last_pos < minMatch) { ip++; continue; }
-
-        /* initialize opt[0] */
-        { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
-        opt[0].mlen = 1;
-        opt[0].litlen = litlen;
-
-         /* check further positions */
-        for (cur = 1; cur <= last_pos; cur++) {
-           inr = ip + cur;
-
-           if (opt[cur-1].mlen == 1) {
-                litlen = opt[cur-1].litlen + 1;
-                if (cur > litlen) {
-                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-litlen);
-                } else
-                    price = ZSTD_getLiteralPrice(optStatePtr, litlen, anchor);
-           } else {
-                litlen = 1;
-                price = opt[cur - 1].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-1);
-           }
-
-           if (cur > last_pos || price <= opt[cur].price)
-                SET_PRICE(cur, 1, 0, litlen, price);
-
-           if (cur == last_pos) break;
-
-           if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
-               continue;
-
-           mlen = opt[cur].mlen;
-           if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
-                opt[cur].rep[2] = opt[cur-mlen].rep[1];
-                opt[cur].rep[1] = opt[cur-mlen].rep[0];
-                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
-           } else {
-                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
-                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
-                /* If opt[cur].off == ZSTD_REP_MOVE_OPT, then mlen != 1.
-                 * offset ZSTD_REP_MOVE_OPT is used for the special case
-                 * litLength == 0, where offset 0 means something special.
-                 * mlen == 1 means the previous byte was stored as a literal,
-                 * so they are mutually exclusive.
-                 */
-                assert(!(opt[cur].off == ZSTD_REP_MOVE_OPT && mlen == 1));
-                opt[cur].rep[0] = (opt[cur].off == ZSTD_REP_MOVE_OPT) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
-           }
-
-            best_mlen = minMatch;
-            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
-                for (i=(opt[cur].mlen != 1); i<last_i; i++) {  /* check rep */
-                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
-                    if ( (repCur > 0) && (repCur < (S32)(inr-prefixStart))
-                       && (ZSTD_readMINMATCH(inr, minMatch) == ZSTD_readMINMATCH(inr - repCur, minMatch))) {
-                       mlen = (U32)ZSTD_count(inr+minMatch, inr+minMatch - repCur, iend) + minMatch;
-
-                       if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
-                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
-                            goto _storeSequence;
-                       }
-
-                       best_off = i - (opt[cur].mlen != 1);
-                       if (mlen > best_mlen) best_mlen = mlen;
-
-                       do {
-                           if (opt[cur].mlen == 1) {
-                                litlen = opt[cur].litlen;
-                                if (cur > litlen) {
-                                    price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
-                                } else
-                                    price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                            } else {
-                                litlen = 0;
-                                price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
-                            }
-
-                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
-                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
-                            mlen--;
-                        } while (mlen >= minMatch);
-            }   }   }
-
-            match_num = ZSTD_BtGetAllMatches_selectMLS(ctx, inr, iend, maxSearches, mls, matches, best_mlen);
-
-            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-                best_mlen = matches[match_num-1].len;
-                best_off = matches[match_num-1].off;
-                last_pos = cur + 1;
-                goto _storeSequence;
-            }
-
-            /* set prices using matches at position = cur */
-            for (u = 0; u < match_num; u++) {
-                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-                best_mlen = matches[u].len;
-
-                while (mlen <= best_mlen) {
-                    if (opt[cur].mlen == 1) {
-                        litlen = opt[cur].litlen;
-                        if (cur > litlen)
-                            price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
-                        else
-                            price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                    } else {
-                        litlen = 0;
-                        price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
-                    }
-
-                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
-                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
-
-                    mlen++;
-        }   }   }
-
-        best_mlen = opt[last_pos].mlen;
-        best_off = opt[last_pos].off;
-        cur = last_pos - best_mlen;
+typedef struct repcodes_s {
+    U32 rep[3];
+} repcodes_t;
 
-        /* store sequence */
-_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
-        opt[0].mlen = 1;
-
-        while (1) {
-            mlen = opt[cur].mlen;
-            offset = opt[cur].off;
-            opt[cur].mlen = best_mlen;
-            opt[cur].off = best_off;
-            best_mlen = mlen;
-            best_off = offset;
-            if (mlen > cur) break;
-            cur -= mlen;
+repcodes_t ZSTD_updateRep(U32 const rep[3], U32 const offset, U32 const ll0)
+{
+    repcodes_t newReps;
+    if (offset >= ZSTD_REP_NUM) {  /* full offset */
+        newReps.rep[2] = rep[1];
+        newReps.rep[1] = rep[0];
+        newReps.rep[0] = offset - ZSTD_REP_MOVE;
+    } else {   /* repcode */
+        U32 const repCode = offset + ll0;
+        if (repCode > 0) {  /* note : if repCode==0, no change */
+            U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+            newReps.rep[2] = (repCode >= 2) ? rep[1] : rep[2];
+            newReps.rep[1] = rep[0];
+            newReps.rep[0] = currentOffset;
+        } else {   /* repCode == 0 */
+            memcpy(&newReps, rep, sizeof(newReps));
         }
+    }
+    return newReps;
+}
 
-        for (u = 0; u <= last_pos;) {
-            u += opt[u].mlen;
-        }
 
-        for (cur=0; cur < last_pos; ) {
-            mlen = opt[cur].mlen;
-            if (mlen == 1) { ip++; cur++; continue; }
-            offset = opt[cur].off;
-            cur += mlen;
-            litLength = (U32)(ip - anchor);
-
-            if (offset > ZSTD_REP_MOVE_OPT) {
-                rep[2] = rep[1];
-                rep[1] = rep[0];
-                rep[0] = offset - ZSTD_REP_MOVE_OPT;
-                offset--;
-            } else {
-                if (offset != 0) {
-                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
-                    if (offset != 1) rep[2] = rep[1];
-                    rep[1] = rep[0];
-                    rep[0] = best_off;
-                }
-                if (litLength==0) offset--;
-            }
+typedef struct {
+    const BYTE* anchor;
+    U32 litlen;
+    U32 rawLitCost;
+} cachedLiteralPrice_t;
 
-            ZSTD_updatePrice(optStatePtr, litLength, anchor, offset, mlen-MINMATCH);
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
-            anchor = ip = ip + mlen;
-    }    }   /* for (cur=0; cur < last_pos; ) */
-
-    /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->repToConfirm[i] = rep[i]; }
+static U32 ZSTD_rawLiteralsCost_cached(
+                            cachedLiteralPrice_t* const cachedLitPrice,
+                            const BYTE* const anchor, U32 const litlen,
+                            const optState_t* const optStatePtr)
+{
+    U32 startCost;
+    U32 remainingLength;
+    const BYTE* startPosition;
+
+    if (anchor == cachedLitPrice->anchor) {
+        startCost = cachedLitPrice->rawLitCost;
+        startPosition = anchor + cachedLitPrice->litlen;
+        assert(litlen >= cachedLitPrice->litlen);
+        remainingLength = litlen - cachedLitPrice->litlen;
+    } else {
+        startCost = 0;
+        startPosition = anchor;
+        remainingLength = litlen;
+    }
 
-    /* Return the last literals size */
-    return iend - anchor;
+    {   U32 const rawLitCost = startCost + ZSTD_rawLiteralsCost(startPosition, remainingLength, optStatePtr);
+        cachedLitPrice->anchor = anchor;
+        cachedLitPrice->litlen = litlen;
+        cachedLitPrice->rawLitCost = rawLitCost;
+        return rawLitCost;
+    }
 }
 
-
-size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+static U32 ZSTD_fullLiteralsCost_cached(
+                            cachedLiteralPrice_t* const cachedLitPrice,
+                            const BYTE* const anchor, U32 const litlen,
+                            const optState_t* const optStatePtr)
 {
-    return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0);
+    return ZSTD_rawLiteralsCost_cached(cachedLitPrice, anchor, litlen, optStatePtr)
+         + ZSTD_litLengthPrice(litlen, optStatePtr);
 }
 
-size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+static int ZSTD_literalsContribution_cached(
+                            cachedLiteralPrice_t* const cachedLitPrice,
+                            const BYTE* const anchor, U32 const litlen,
+                            const optState_t* const optStatePtr)
 {
-    return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 1);
+    int const contribution = ZSTD_rawLiteralsCost_cached(cachedLitPrice, anchor, litlen, optStatePtr)
+                           + ZSTD_litLengthContribution(litlen, optStatePtr);
+    return contribution;
 }
 
-
 FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_opt_extDict_generic(ZSTD_CCtx* ctx,
-                                     const void* src, size_t srcSize, const int ultra)
+size_t ZSTD_compressBlock_opt_generic(ZSTD_matchState_t* ms,seqStore_t* seqStore,
+                                      U32 rep[ZSTD_REP_NUM],
+                                      ZSTD_compressionParameters const* cParams,
+                                      const void* src, size_t srcSize,
+                                      const int optLevel, const int extDict)
 {
-    seqStore_t* seqStorePtr = &(ctx->seqStore);
-    optState_t* optStatePtr = &(ctx->optState);
+    optState_t* const optStatePtr = &ms->opt;
     const BYTE* const istart = (const BYTE*)src;
     const BYTE* ip = istart;
     const BYTE* anchor = istart;
     const BYTE* const iend = istart + srcSize;
     const BYTE* const ilimit = iend - 8;
-    const BYTE* const base = ctx->base;
-    const U32 lowestIndex = ctx->lowLimit;
-    const U32 dictLimit = ctx->dictLimit;
-    const BYTE* const prefixStart = base + dictLimit;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const dictEnd  = dictBase + dictLimit;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const prefixStart = base + ms->window.dictLimit;
 
-    const U32 maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
-    const U32 sufficient_len = ctx->appliedParams.cParams.targetLength;
-    const U32 mls = ctx->appliedParams.cParams.searchLength;
-    const U32 minMatch = (ctx->appliedParams.cParams.searchLength == 3) ? 3 : 4;
+    U32 const sufficient_len = MIN(cParams->targetLength, ZSTD_OPT_NUM -1);
+    U32 const minMatch = (cParams->searchLength == 3) ? 3 : 4;
 
-    ZSTD_optimal_t* opt = optStatePtr->priceTable;
-    ZSTD_match_t* matches = optStatePtr->matchTable;
-    const BYTE* inr;
+    ZSTD_optimal_t* const opt = optStatePtr->priceTable;
+    ZSTD_match_t* const matches = optStatePtr->matchTable;
+    cachedLiteralPrice_t cachedLitPrice;
 
     /* init */
-    U32 offset, rep[ZSTD_REP_NUM];
-    { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=seqStorePtr->rep[i]; }
-
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    DEBUGLOG(5, "ZSTD_compressBlock_opt_generic");
+    ms->nextToUpdate3 = ms->nextToUpdate;
     ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize);
     ip += (ip==prefixStart);
+    memset(&cachedLitPrice, 0, sizeof(cachedLitPrice));
 
     /* Match Loop */
     while (ip < ilimit) {
-        U32 cur, match_num, last_pos, litlen, price;
-        U32 u, mlen, best_mlen, best_off, litLength;
-        U32 current = (U32)(ip-base);
-        memset(opt, 0, sizeof(ZSTD_optimal_t));
-        last_pos = 0;
-        opt[0].litlen = (U32)(ip - anchor);
-
-        /* check repCode */
-        {   U32 i, last_i = ZSTD_REP_CHECK + (ip==anchor);
-            for (i = (ip==anchor); i<last_i; i++) {
-                const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : rep[i];
-                const U32 repIndex = (U32)(current - repCur);
-                const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-                const BYTE* const repMatch = repBase + repIndex;
-                if ( (repCur > 0 && repCur <= (S32)current)
-                   && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
-                   && (ZSTD_readMINMATCH(ip, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
-                    /* repcode detected we should take it */
-                    const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                    mlen = (U32)ZSTD_count_2segments(ip+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
-
-                    if (mlen > sufficient_len || mlen >= ZSTD_OPT_NUM) {
-                        best_mlen = mlen; best_off = i; cur = 0; last_pos = 1;
-                        goto _storeSequence;
-                    }
-
-                    best_off = i - (ip==anchor);
-                    litlen = opt[0].litlen;
-                    do {
-                        price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                        if (mlen > last_pos || price < opt[mlen].price)
-                            SET_PRICE(mlen, mlen, i, litlen, price);   /* note : macro modifies last_pos */
-                        mlen--;
-                    } while (mlen >= minMatch);
-        }   }   }
-
-        match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, ip, iend, maxSearches, mls, matches, minMatch);  /* first search (depth 0) */
-
-        if (!last_pos && !match_num) { ip++; continue; }
-
-        { U32 i; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
-        opt[0].mlen = 1;
-
-        if (match_num && (matches[match_num-1].len > sufficient_len || matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-            best_mlen = matches[match_num-1].len;
-            best_off = matches[match_num-1].off;
-            cur = 0;
-            last_pos = 1;
-            goto _storeSequence;
-        }
-
-        best_mlen = (last_pos) ? last_pos : minMatch;
-
-        /* set prices using matches at position = 0 */
-        for (u = 0; u < match_num; u++) {
-            mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-            best_mlen = matches[u].len;
-            litlen = opt[0].litlen;
-            while (mlen <= best_mlen) {
-                price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                if (mlen > last_pos || price < opt[mlen].price)
-                    SET_PRICE(mlen, mlen, matches[u].off, litlen, price);
-                mlen++;
-        }   }
-
-        if (last_pos < minMatch) {
-            ip++; continue;
+        U32 cur, last_pos = 0;
+        U32 best_mlen, best_off;
+
+        /* find first match */
+        {   U32 const litlen = (U32)(ip - anchor);
+            U32 const ll0 = !litlen;
+            U32 const nbMatches = ZSTD_BtGetAllMatches(ms, cParams, ip, iend, extDict, rep, ll0, matches, minMatch);
+            if (!nbMatches) { ip++; continue; }
+
+            /* initialize opt[0] */
+            { U32 i ; for (i=0; i<ZSTD_REP_NUM; i++) opt[0].rep[i] = rep[i]; }
+            opt[0].mlen = 1;
+            opt[0].litlen = litlen;
+
+            /* large match -> immediate encoding */
+            {   U32 const maxML = matches[nbMatches-1].len;
+                DEBUGLOG(7, "found %u matches of maxLength=%u and offset=%u at cPos=%u => start new serie",
+                            nbMatches, maxML, matches[nbMatches-1].off, (U32)(ip-prefixStart));
+
+                if (maxML > sufficient_len) {
+                    best_mlen = maxML;
+                    best_off = matches[nbMatches-1].off;
+                    DEBUGLOG(7, "large match (%u>%u), immediate encoding",
+                                best_mlen, sufficient_len);
+                    cur = 0;
+                    last_pos = 1;
+                    goto _shortestPath;
+            }   }
+
+            /* set prices for first matches starting position == 0 */
+            {   U32 const literalsPrice = ZSTD_fullLiteralsCost_cached(&cachedLitPrice, anchor, litlen, optStatePtr);
+                U32 pos;
+                U32 matchNb;
+                for (pos = 0; pos < minMatch; pos++) {
+                    opt[pos].mlen = 1;
+                    opt[pos].price = ZSTD_MAX_PRICE;
+                }
+                for (matchNb = 0; matchNb < nbMatches; matchNb++) {
+                    U32 const offset = matches[matchNb].off;
+                    U32 const end = matches[matchNb].len;
+                    repcodes_t const repHistory = ZSTD_updateRep(rep, offset, ll0);
+                    for ( ; pos <= end ; pos++ ) {
+                        U32 const matchPrice = literalsPrice + ZSTD_getMatchPrice(offset, pos, optStatePtr, optLevel);
+                        DEBUGLOG(7, "rPos:%u => set initial price : %u",
+                                    pos, matchPrice);
+                        opt[pos].mlen = pos;
+                        opt[pos].off = offset;
+                        opt[pos].litlen = litlen;
+                        opt[pos].price = matchPrice;
+                        memcpy(opt[pos].rep, &repHistory, sizeof(repHistory));
+                }   }
+                last_pos = pos-1;
+            }
         }
 
         /* check further positions */
         for (cur = 1; cur <= last_pos; cur++) {
-            inr = ip + cur;
+            const BYTE* const inr = ip + cur;
+            assert(cur < ZSTD_OPT_NUM);
 
-            if (opt[cur-1].mlen == 1) {
-                litlen = opt[cur-1].litlen + 1;
+            /* Fix current position with one literal if cheaper */
+            {   U32 const litlen = (opt[cur-1].mlen == 1) ? opt[cur-1].litlen + 1 : 1;
+                int price;  /* note : contribution can be negative */
                 if (cur > litlen) {
-                    price = opt[cur - litlen].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-litlen);
-                } else
-                    price = ZSTD_getLiteralPrice(optStatePtr, litlen, anchor);
-            } else {
-                litlen = 1;
-                price = opt[cur - 1].price + ZSTD_getLiteralPrice(optStatePtr, litlen, inr-1);
-            }
-
-            if (cur > last_pos || price <= opt[cur].price)
-                SET_PRICE(cur, 1, 0, litlen, price);
+                    price = opt[cur - litlen].price + ZSTD_literalsContribution(inr-litlen, litlen, optStatePtr);
+                } else {
+                    price = ZSTD_literalsContribution_cached(&cachedLitPrice, anchor, litlen, optStatePtr);
+                }
+                assert(price < 1000000000); /* overflow check */
+                if (price <= opt[cur].price) {
+                    DEBUGLOG(7, "rPos:%u : better price (%u<%u) using literal",
+                                cur, price, opt[cur].price);
+                    opt[cur].mlen = 1;
+                    opt[cur].off = 0;
+                    opt[cur].litlen = litlen;
+                    opt[cur].price = price;
+                    memcpy(opt[cur].rep, opt[cur-1].rep, sizeof(opt[cur].rep));
+            }   }
+
+            /* last match must start at a minimum distance of 8 from oend */
+            if (inr > ilimit) continue;
 
             if (cur == last_pos) break;
 
-            if (inr > ilimit)  /* last match must start at a minimum distance of 8 from oend */
-                continue;
-
-            mlen = opt[cur].mlen;
-            if (opt[cur].off > ZSTD_REP_MOVE_OPT) {
-                opt[cur].rep[2] = opt[cur-mlen].rep[1];
-                opt[cur].rep[1] = opt[cur-mlen].rep[0];
-                opt[cur].rep[0] = opt[cur].off - ZSTD_REP_MOVE_OPT;
-            } else {
-                opt[cur].rep[2] = (opt[cur].off > 1) ? opt[cur-mlen].rep[1] : opt[cur-mlen].rep[2];
-                opt[cur].rep[1] = (opt[cur].off > 0) ? opt[cur-mlen].rep[0] : opt[cur-mlen].rep[1];
-                assert(!(opt[cur].off == ZSTD_REP_MOVE_OPT && mlen == 1));
-                opt[cur].rep[0] = (opt[cur].off == ZSTD_REP_MOVE_OPT) ? (opt[cur-mlen].rep[0] - 1) : (opt[cur-mlen].rep[opt[cur].off]);
-            }
+             if ( (optLevel==0) /*static*/
+               && (opt[cur+1].price <= opt[cur].price) )
+                continue;  /* skip unpromising positions; about ~+6% speed, -0.01 ratio */
+
+            {   U32 const ll0 = (opt[cur].mlen != 1);
+                U32 const litlen = (opt[cur].mlen == 1) ? opt[cur].litlen : 0;
+                U32 const previousPrice = (cur > litlen) ? opt[cur-litlen].price : 0;
+                U32 const basePrice = previousPrice + ZSTD_fullLiteralsCost(inr-litlen, litlen, optStatePtr);
+                U32 const nbMatches = ZSTD_BtGetAllMatches(ms, cParams, inr, iend, extDict, opt[cur].rep, ll0, matches, minMatch);
+                U32 matchNb;
+                if (!nbMatches) continue;
+
+                {   U32 const maxML = matches[nbMatches-1].len;
+                    DEBUGLOG(7, "rPos:%u, found %u matches, of maxLength=%u",
+                                cur, nbMatches, maxML);
+
+                    if ( (maxML > sufficient_len)
+                       | (cur + maxML >= ZSTD_OPT_NUM) ) {
+                        best_mlen = maxML;
+                        best_off = matches[nbMatches-1].off;
+                        last_pos = cur + 1;
+                        goto _shortestPath;
+                    }
+                }
 
-            best_mlen = minMatch;
-            {   U32 i, last_i = ZSTD_REP_CHECK + (mlen != 1);
-                for (i = (mlen != 1); i<last_i; i++) {
-                    const S32 repCur = (i==ZSTD_REP_MOVE_OPT) ? (opt[cur].rep[0] - 1) : opt[cur].rep[i];
-                    const U32 repIndex = (U32)(current+cur - repCur);
-                    const BYTE* const repBase = repIndex < dictLimit ? dictBase : base;
-                    const BYTE* const repMatch = repBase + repIndex;
-                    if ( (repCur > 0 && repCur <= (S32)(current+cur))
-                      && (((U32)((dictLimit-1) - repIndex) >= 3) & (repIndex>lowestIndex))  /* intentional overflow */
-                      && (ZSTD_readMINMATCH(inr, minMatch) == ZSTD_readMINMATCH(repMatch, minMatch)) ) {
-                        /* repcode detected */
-                        const BYTE* const repEnd = repIndex < dictLimit ? dictEnd : iend;
-                        mlen = (U32)ZSTD_count_2segments(inr+minMatch, repMatch+minMatch, iend, repEnd, prefixStart) + minMatch;
-
-                        if (mlen > sufficient_len || cur + mlen >= ZSTD_OPT_NUM) {
-                            best_mlen = mlen; best_off = i; last_pos = cur + 1;
-                            goto _storeSequence;
+                /* set prices using matches found at position == cur */
+                for (matchNb = 0; matchNb < nbMatches; matchNb++) {
+                    U32 const offset = matches[matchNb].off;
+                    repcodes_t const repHistory = ZSTD_updateRep(opt[cur].rep, offset, ll0);
+                    U32 const lastML = matches[matchNb].len;
+                    U32 const startML = (matchNb>0) ? matches[matchNb-1].len+1 : minMatch;
+                    U32 mlen;
+
+                    DEBUGLOG(7, "testing match %u => offCode=%u, mlen=%u, llen=%u",
+                                matchNb, matches[matchNb].off, lastML, litlen);
+
+                    for (mlen = lastML; mlen >= startML; mlen--) {
+                        U32 const pos = cur + mlen;
+                        int const price = basePrice + ZSTD_getMatchPrice(offset, mlen, optStatePtr, optLevel);
+
+                        if ((pos > last_pos) || (price < opt[pos].price)) {
+                            DEBUGLOG(7, "rPos:%u => new better price (%u<%u)",
+                                        pos, price, opt[pos].price);
+                            while (last_pos < pos) { opt[last_pos+1].price = ZSTD_MAX_PRICE; last_pos++; }
+                            opt[pos].mlen = mlen;
+                            opt[pos].off = offset;
+                            opt[pos].litlen = litlen;
+                            opt[pos].price = price;
+                            memcpy(opt[pos].rep, &repHistory, sizeof(repHistory));
+                        } else {
+                            if (optLevel==0) break;  /* gets ~+10% speed for about -0.01 ratio loss */
                         }
-
-                        best_off = i - (opt[cur].mlen != 1);
-                        if (mlen > best_mlen) best_mlen = mlen;
-
-                        do {
-                            if (opt[cur].mlen == 1) {
-                                litlen = opt[cur].litlen;
-                                if (cur > litlen) {
-                                    price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, inr-litlen, best_off, mlen - MINMATCH, ultra);
-                                } else
-                                    price = ZSTD_getPrice(optStatePtr, litlen, anchor, best_off, mlen - MINMATCH, ultra);
-                            } else {
-                                litlen = 0;
-                                price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, best_off, mlen - MINMATCH, ultra);
-                            }
-
-                            if (cur + mlen > last_pos || price <= opt[cur + mlen].price)
-                                SET_PRICE(cur + mlen, mlen, i, litlen, price);
-                            mlen--;
-                        } while (mlen >= minMatch);
             }   }   }
-
-            match_num = ZSTD_BtGetAllMatches_selectMLS_extDict(ctx, inr, iend, maxSearches, mls, matches, minMatch);
-
-            if (match_num > 0 && (matches[match_num-1].len > sufficient_len || cur + matches[match_num-1].len >= ZSTD_OPT_NUM)) {
-                best_mlen = matches[match_num-1].len;
-                best_off = matches[match_num-1].off;
-                last_pos = cur + 1;
-                goto _storeSequence;
-            }
-
-            /* set prices using matches at position = cur */
-            for (u = 0; u < match_num; u++) {
-                mlen = (u>0) ? matches[u-1].len+1 : best_mlen;
-                best_mlen = matches[u].len;
-
-                while (mlen <= best_mlen) {
-                    if (opt[cur].mlen == 1) {
-                        litlen = opt[cur].litlen;
-                        if (cur > litlen)
-                            price = opt[cur - litlen].price + ZSTD_getPrice(optStatePtr, litlen, ip+cur-litlen, matches[u].off-1, mlen - MINMATCH, ultra);
-                        else
-                            price = ZSTD_getPrice(optStatePtr, litlen, anchor, matches[u].off-1, mlen - MINMATCH, ultra);
-                    } else {
-                        litlen = 0;
-                        price = opt[cur].price + ZSTD_getPrice(optStatePtr, 0, NULL, matches[u].off-1, mlen - MINMATCH, ultra);
-                    }
-
-                    if (cur + mlen > last_pos || (price < opt[cur + mlen].price))
-                        SET_PRICE(cur + mlen, mlen, matches[u].off, litlen, price);
-
-                    mlen++;
-        }   }   }   /* for (cur = 1; cur <= last_pos; cur++) */
+        }  /* for (cur = 1; cur <= last_pos; cur++) */
 
         best_mlen = opt[last_pos].mlen;
         best_off = opt[last_pos].off;
         cur = last_pos - best_mlen;
 
-        /* store sequence */
-_storeSequence:   /* cur, last_pos, best_mlen, best_off have to be set */
-        opt[0].mlen = 1;
-
-        while (1) {
-            mlen = opt[cur].mlen;
-            offset = opt[cur].off;
-            opt[cur].mlen = best_mlen;
-            opt[cur].off = best_off;
-            best_mlen = mlen;
-            best_off = offset;
-            if (mlen > cur) break;
-            cur -= mlen;
-        }
-
-        for (u = 0; u <= last_pos; ) {
-            u += opt[u].mlen;
-        }
+_shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
+        assert(opt[0].mlen == 1);
+
+        /* reverse traversal */
+        DEBUGLOG(7, "start reverse traversal (last_pos:%u, cur:%u)",
+                    last_pos, cur);
+        {   U32 selectedMatchLength = best_mlen;
+            U32 selectedOffset = best_off;
+            U32 pos = cur;
+            while (1) {
+                U32 const mlen = opt[pos].mlen;
+                U32 const off = opt[pos].off;
+                opt[pos].mlen = selectedMatchLength;
+                opt[pos].off = selectedOffset;
+                selectedMatchLength = mlen;
+                selectedOffset = off;
+                if (mlen > pos) break;
+                pos -= mlen;
+        }   }
 
-        for (cur=0; cur < last_pos; ) {
-            mlen = opt[cur].mlen;
-            if (mlen == 1) { ip++; cur++; continue; }
-            offset = opt[cur].off;
-            cur += mlen;
-            litLength = (U32)(ip - anchor);
-
-            if (offset > ZSTD_REP_MOVE_OPT) {
-                rep[2] = rep[1];
-                rep[1] = rep[0];
-                rep[0] = offset - ZSTD_REP_MOVE_OPT;
-                offset--;
-            } else {
-                if (offset != 0) {
-                    best_off = (offset==ZSTD_REP_MOVE_OPT) ? (rep[0] - 1) : (rep[offset]);
-                    if (offset != 1) rep[2] = rep[1];
+        /* save sequences */
+        {   U32 pos;
+            for (pos=0; pos < last_pos; ) {
+                U32 const llen = (U32)(ip - anchor);
+                U32 const mlen = opt[pos].mlen;
+                U32 const offset = opt[pos].off;
+                if (mlen == 1) { ip++; pos++; continue; }  /* literal position => move on */
+                pos += mlen; ip += mlen;
+
+                /* repcodes update : like ZSTD_updateRep(), but update in place */
+                if (offset >= ZSTD_REP_NUM) {  /* full offset */
+                    rep[2] = rep[1];
                     rep[1] = rep[0];
-                    rep[0] = best_off;
+                    rep[0] = offset - ZSTD_REP_MOVE;
+                } else {   /* repcode */
+                    U32 const repCode = offset + (llen==0);
+                    if (repCode) {  /* note : if repCode==0, no change */
+                        U32 const currentOffset = (repCode==ZSTD_REP_NUM) ? (rep[0] - 1) : rep[repCode];
+                        if (repCode >= 2) rep[2] = rep[1];
+                        rep[1] = rep[0];
+                        rep[0] = currentOffset;
+                    }
                 }
 
-                if (litLength==0) offset--;
-            }
-
-            ZSTD_updatePrice(optStatePtr, litLength, anchor, offset, mlen-MINMATCH);
-            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, mlen-MINMATCH);
-            anchor = ip = ip + mlen;
-    }    }   /* for (cur=0; cur < last_pos; ) */
-
-    /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->repToConfirm[i] = rep[i]; }
+                ZSTD_updateStats(optStatePtr, llen, anchor, offset, mlen);
+                ZSTD_storeSeq(seqStore, llen, anchor, offset, mlen-MINMATCH);
+                anchor = ip;
+        }   }
+        ZSTD_setLog2Prices(optStatePtr);
+    }   /* while (ip < ilimit) */
 
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_btopt(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
+{
+    DEBUGLOG(5, "ZSTD_compressBlock_btopt");
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 0 /*optLevel*/, 0 /*extDict*/);
+}
+
+size_t ZSTD_compressBlock_btultra(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
+{
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 2 /*optLevel*/, 0 /*extDict*/);
+}
+
+size_t ZSTD_compressBlock_btopt_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 0);
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 0 /*optLevel*/, 1 /*extDict*/);
 }
 
-size_t ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+size_t ZSTD_compressBlock_btultra_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize)
 {
-    return ZSTD_compressBlock_opt_extDict_generic(ctx, src, srcSize, 1);
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 2 /*optLevel*/, 1 /*extDict*/);
 }

src/borg/algorithms/zstd/lib/compress/zstd_opt.h

@@ -11,17 +11,29 @@
 #ifndef ZSTD_OPT_H
 #define ZSTD_OPT_H
 
-#include "zstd_compress.h"
-
 #if defined (__cplusplus)
 extern "C" {
 #endif
 
-size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+#include "zstd_compress_internal.h"
+
+void ZSTD_updateTree(
+        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+        const BYTE* ip, const BYTE* iend);  /* used in ZSTD_loadDictionaryContent() */
+
+size_t ZSTD_compressBlock_btopt(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
 
-size_t ZSTD_compressBlock_btopt_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_btultra_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
+size_t ZSTD_compressBlock_btopt_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_btultra_extDict(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        ZSTD_compressionParameters const* cParams, void const* src, size_t srcSize);
 
 #if defined (__cplusplus)
 }

src/borg/algorithms/zstd/lib/compress/zstdmt_compress.h

@@ -30,15 +30,15 @@
 
 /* ===   Memory management   === */
 typedef struct ZSTDMT_CCtx_s ZSTDMT_CCtx;
-ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbThreads);
-ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbThreads,
+ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx(unsigned nbWorkers);
+ZSTDLIB_API ZSTDMT_CCtx* ZSTDMT_createCCtx_advanced(unsigned nbWorkers,
                                                     ZSTD_customMem cMem);
 ZSTDLIB_API size_t ZSTDMT_freeCCtx(ZSTDMT_CCtx* mtctx);
 
 ZSTDLIB_API size_t ZSTDMT_sizeof_CCtx(ZSTDMT_CCtx* mtctx);
 
 
-/* ===   Simple buffer-to-butter one-pass function   === */
+/* ===   Simple one-pass compression function   === */
 
 ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
                                        void* dst, size_t dstCapacity,
@@ -50,7 +50,7 @@ ZSTDLIB_API size_t ZSTDMT_compressCCtx(ZSTDMT_CCtx* mtctx,
 /* ===   Streaming functions   === */
 
 ZSTDLIB_API size_t ZSTDMT_initCStream(ZSTDMT_CCtx* mtctx, int compressionLevel);
-ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize);    /**< pledgedSrcSize is optional and can be zero == unknown */
+ZSTDLIB_API size_t ZSTDMT_resetCStream(ZSTDMT_CCtx* mtctx, unsigned long long pledgedSrcSize);  /**< if srcSize is not known at reset time, use ZSTD_CONTENTSIZE_UNKNOWN. Note: for compatibility with older programs, 0 means the same as ZSTD_CONTENTSIZE_UNKNOWN, but it will change in the future to mean "empty" */
 
 ZSTDLIB_API size_t ZSTDMT_compressStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output, ZSTD_inBuffer* input);
 
@@ -60,15 +60,15 @@ ZSTDLIB_API size_t ZSTDMT_endStream(ZSTDMT_CCtx* mtctx, ZSTD_outBuffer* output);
 
 /* ===   Advanced functions and parameters  === */
 
-#ifndef ZSTDMT_SECTION_SIZE_MIN
-#  define ZSTDMT_SECTION_SIZE_MIN (1U << 20)   /* 1 MB - Minimum size of each compression job */
+#ifndef ZSTDMT_JOBSIZE_MIN
+#  define ZSTDMT_JOBSIZE_MIN (1U << 20)   /* 1 MB - Minimum size of each compression job */
 #endif
 
 ZSTDLIB_API size_t ZSTDMT_compress_advanced(ZSTDMT_CCtx* mtctx,
                                            void* dst, size_t dstCapacity,
                                      const void* src, size_t srcSize,
                                      const ZSTD_CDict* cdict,
-                                           ZSTD_parameters const params,
+                                           ZSTD_parameters params,
                                            unsigned overlapLog);
 
 ZSTDLIB_API size_t ZSTDMT_initCStream_advanced(ZSTDMT_CCtx* mtctx,
@@ -84,35 +84,59 @@ ZSTDLIB_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
 /* ZSTDMT_parameter :
  * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
 typedef enum {
-    ZSTDMT_p_sectionSize,        /* size of input "section". Each section is compressed in parallel. 0 means default, which is dynamically determined within compression functions */
-    ZSTDMT_p_overlapSectionLog   /* Log of overlapped section; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window */
+    ZSTDMT_p_jobSize,           /* Each job is compressed in parallel. By default, this value is dynamically determined depending on compression parameters. Can be set explicitly here. */
+    ZSTDMT_p_overlapSectionLog  /* Each job may reload a part of previous job to enhance compressionr ratio; 0 == no overlap, 6(default) == use 1/8th of window, >=9 == use full window. This is a "sticky" parameter : its value will be re-used on next compression job */
 } ZSTDMT_parameter;
 
 /* ZSTDMT_setMTCtxParameter() :
  * allow setting individual parameters, one at a time, among a list of enums defined in ZSTDMT_parameter.
- * The function must be called typically after ZSTD_createCCtx().
+ * The function must be called typically after ZSTD_createCCtx() but __before ZSTDMT_init*() !__
  * Parameters not explicitly reset by ZSTDMT_init*() remain the same in consecutive compression sessions.
  * @return : 0, or an error code (which can be tested using ZSTD_isError()) */
 ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter parameter, unsigned value);
 
 
 /*! ZSTDMT_compressStream_generic() :
- *  Combines ZSTDMT_compressStream() with ZSTDMT_flushStream() or ZSTDMT_endStream()
+ *  Combines ZSTDMT_compressStream() with optional ZSTDMT_flushStream() or ZSTDMT_endStream()
  *  depending on flush directive.
  * @return : minimum amount of data still to be flushed
  *           0 if fully flushed
- *           or an error code */
+ *           or an error code
+ *  note : needs to be init using any ZSTD_initCStream*() variant */
 ZSTDLIB_API size_t ZSTDMT_compressStream_generic(ZSTDMT_CCtx* mtctx,
                                                 ZSTD_outBuffer* output,
                                                 ZSTD_inBuffer* input,
                                                 ZSTD_EndDirective endOp);
 
 
-/* ===   Private definitions; never ever use directly  === */
+/* ========================================================
+ * ===  Private interface, for use by ZSTD_compress.c   ===
+ * ===  Not exposed in libzstd. Never invoke directly   ===
+ * ======================================================== */
 
 size_t ZSTDMT_CCtxParam_setMTCtxParameter(ZSTD_CCtx_params* params, ZSTDMT_parameter parameter, unsigned value);
 
-size_t ZSTDMT_initializeCCtxParameters(ZSTD_CCtx_params* params, unsigned nbThreads);
+/* ZSTDMT_CCtxParam_setNbWorkers()
+ * Set nbWorkers, and clamp it.
+ * Also reset jobSize and overlapLog */
+size_t ZSTDMT_CCtxParam_setNbWorkers(ZSTD_CCtx_params* params, unsigned nbWorkers);
+
+/*! ZSTDMT_updateCParams_whileCompressing() :
+ *  Updates only a selected set of compression parameters, to remain compatible with current frame.
+ *  New parameters will be applied to next compression job. */
+void ZSTDMT_updateCParams_whileCompressing(ZSTDMT_CCtx* mtctx, const ZSTD_CCtx_params* cctxParams);
+
+/* ZSTDMT_getNbWorkers():
+ * @return nb threads currently active in mtctx.
+ * mtctx must be valid */
+unsigned ZSTDMT_getNbWorkers(const ZSTDMT_CCtx* mtctx);
+
+/* ZSTDMT_getFrameProgression():
+ * tells how much data has been consumed (input) and produced (output) for current frame.
+ * able to count progression inside worker threads.
+ */
+ZSTD_frameProgression ZSTDMT_getFrameProgression(ZSTDMT_CCtx* mtctx);
+
 
 /*! ZSTDMT_initCStream_internal() :
  *  Private use only. Init streaming operation.
@@ -120,7 +144,7 @@ size_t ZSTDMT_initializeCCtxParameters(ZSTD_CCtx_params* params, unsigned nbThre
  *  must receive dict, or cdict, or none, but not both.
  *  @return : 0, or an error code */
 size_t ZSTDMT_initCStream_internal(ZSTDMT_CCtx* zcs,
-                    const void* dict, size_t dictSize, ZSTD_dictMode_e dictMode,
+                    const void* dict, size_t dictSize, ZSTD_dictContentType_e dictContentType,
                     const ZSTD_CDict* cdict,
                     ZSTD_CCtx_params params, unsigned long long pledgedSrcSize);
 

src/borg/algorithms/zstd/lib/decompress/huf_decompress.c

@@ -49,18 +49,19 @@
 ****************************************************************/
 #define HUF_isError ERR_isError
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+#define CHECK_F(f) { size_t const err_ = (f); if (HUF_isError(err_)) return err_; }
 
 
 /* **************************************************************
 *  Byte alignment for workSpace management
 ****************************************************************/
-#define HUF_ALIGN(x, a) HUF_ALIGN_MASK((x), (a) - 1)
+#define HUF_ALIGN(x, a)         HUF_ALIGN_MASK((x), (a) - 1)
 #define HUF_ALIGN_MASK(x, mask) (((x) + (mask)) & ~(mask))
 
+
 /*-***************************/
 /*  generic DTableDesc       */
 /*-***************************/
-
 typedef struct { BYTE maxTableLog; BYTE tableType; BYTE tableLog; BYTE reserved; } DTableDesc;
 
 static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
@@ -74,7 +75,6 @@ static DTableDesc HUF_getDTableDesc(const HUF_DTable* table)
 /*-***************************/
 /*  single-symbol decoding   */
 /*-***************************/
-
 typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX2;   /* single-symbol decoding */
 
 size_t HUF_readDTableX2_wksp(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize)
@@ -94,10 +94,7 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable, const void* src, size_t srcSize
     huffWeight = (BYTE *)((U32 *)workSpace + spaceUsed32);
     spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
 
-    if ((spaceUsed32 << 2) > wkspSize)
-        return ERROR(tableLog_tooLarge);
-    workSpace = (U32 *)workSpace + spaceUsed32;
-    wkspSize -= (spaceUsed32 << 2);
+    if ((spaceUsed32 << 2) > wkspSize) return ERROR(tableLog_tooLarge);
 
     HUF_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
     /* memset(huffWeight, 0, sizeof(huffWeight)); */   /* is not necessary, even though some analyzer complain ... */
@@ -144,8 +141,10 @@ size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)
                                  workSpace, sizeof(workSpace));
 }
 
+typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX4;  /* double-symbols decoding */
 
-static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
+FORCE_INLINE_TEMPLATE BYTE
+HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
 {
     size_t const val = BIT_lookBitsFast(Dstream, dtLog); /* note : dtLog >= 1 */
     BYTE const c = dt[val].byte;
@@ -156,7 +155,7 @@ static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, con
 #define HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr) \
     *ptr++ = HUF_decodeSymbolX2(DStreamPtr, dt, dtLog)
 
-#define HUF_DECODE_SYMBOLX2_1(ptr, DStreamPtr) \
+#define HUF_DECODE_SYMBOLX2_1(ptr, DStreamPtr)  \
     if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
         HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
 
@@ -164,30 +163,33 @@ static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, con
     if (MEM_64bits()) \
         HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
 
-HINT_INLINE size_t HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
+HINT_INLINE size_t
+HUF_decodeStreamX2(BYTE* p, BIT_DStream_t* const bitDPtr, BYTE* const pEnd, const HUF_DEltX2* const dt, const U32 dtLog)
 {
     BYTE* const pStart = p;
 
     /* up to 4 symbols at a time */
-    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p <= pEnd-4)) {
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-3)) {
         HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
         HUF_DECODE_SYMBOLX2_1(p, bitDPtr);
         HUF_DECODE_SYMBOLX2_2(p, bitDPtr);
         HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
     }
 
-    /* closer to the end */
-    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) && (p < pEnd))
-        HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
+    /* [0-3] symbols remaining */
+    if (MEM_32bits())
+        while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd))
+            HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
 
-    /* no more data to retrieve from bitstream, hence no need to reload */
+    /* no more data to retrieve from bitstream, no need to reload */
     while (p < pEnd)
         HUF_DECODE_SYMBOLX2_0(p, bitDPtr);
 
     return pEnd-pStart;
 }
 
-static size_t HUF_decompress1X2_usingDTable_internal(
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress1X2_usingDTable_internal_body(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
     const HUF_DTable* DTable)
@@ -200,58 +202,17 @@ static size_t HUF_decompress1X2_usingDTable_internal(
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
     U32 const dtLog = dtd.tableLog;
 
-    { size_t const errorCode = BIT_initDStream(&bitD, cSrc, cSrcSize);
-      if (HUF_isError(errorCode)) return errorCode; }
+    CHECK_F( BIT_initDStream(&bitD, cSrc, cSrcSize) );
 
     HUF_decodeStreamX2(op, &bitD, oend, dt, dtLog);
 
-    /* check */
     if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
 
     return dstSize;
 }
 
-size_t HUF_decompress1X2_usingDTable(
-          void* dst,  size_t dstSize,
-    const void* cSrc, size_t cSrcSize,
-    const HUF_DTable* DTable)
-{
-    DTableDesc dtd = HUF_getDTableDesc(DTable);
-    if (dtd.tableType != 0) return ERROR(GENERIC);
-    return HUF_decompress1X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
-}
-
-size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
-                                   const void* cSrc, size_t cSrcSize,
-                                   void* workSpace, size_t wkspSize)
-{
-    const BYTE* ip = (const BYTE*) cSrc;
-
-    size_t const hSize = HUF_readDTableX2_wksp(DCtx, cSrc, cSrcSize, workSpace, wkspSize);
-    if (HUF_isError(hSize)) return hSize;
-    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
-    ip += hSize; cSrcSize -= hSize;
-
-    return HUF_decompress1X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
-}
-
-
-size_t HUF_decompress1X2_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
-                              const void* cSrc, size_t cSrcSize)
-{
-    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
-    return HUF_decompress1X2_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
-                                       workSpace, sizeof(workSpace));
-}
-
-size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
-{
-    HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
-    return HUF_decompress1X2_DCtx (DTable, dst, dstSize, cSrc, cSrcSize);
-}
-
-
-static size_t HUF_decompress4X2_usingDTable_internal(
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress4X2_usingDTable_internal_body(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
     const HUF_DTable* DTable)
@@ -286,23 +247,19 @@ static size_t HUF_decompress4X2_usingDTable_internal(
         BYTE* op2 = opStart2;
         BYTE* op3 = opStart3;
         BYTE* op4 = opStart4;
-        U32 endSignal;
+        U32 endSignal = BIT_DStream_unfinished;
         DTableDesc const dtd = HUF_getDTableDesc(DTable);
         U32 const dtLog = dtd.tableLog;
 
         if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
-        { size_t const errorCode = BIT_initDStream(&bitD1, istart1, length1);
-          if (HUF_isError(errorCode)) return errorCode; }
-        { size_t const errorCode = BIT_initDStream(&bitD2, istart2, length2);
-          if (HUF_isError(errorCode)) return errorCode; }
-        { size_t const errorCode = BIT_initDStream(&bitD3, istart3, length3);
-          if (HUF_isError(errorCode)) return errorCode; }
-        { size_t const errorCode = BIT_initDStream(&bitD4, istart4, length4);
-          if (HUF_isError(errorCode)) return errorCode; }
+        CHECK_F( BIT_initDStream(&bitD1, istart1, length1) );
+        CHECK_F( BIT_initDStream(&bitD2, istart2, length2) );
+        CHECK_F( BIT_initDStream(&bitD3, istart3, length3) );
+        CHECK_F( BIT_initDStream(&bitD4, istart4, length4) );
 
-        /* 16-32 symbols per loop (4-8 symbols per stream) */
+        /* up to 16 symbols per loop (4 symbols per stream) in 64-bit mode */
         endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
-        for ( ; (endSignal==BIT_DStream_unfinished) && (op4<(oend-7)) ; ) {
+        while ( (endSignal==BIT_DStream_unfinished) && (op4<(oend-3)) ) {
             HUF_DECODE_SYMBOLX2_2(op1, &bitD1);
             HUF_DECODE_SYMBOLX2_2(op2, &bitD2);
             HUF_DECODE_SYMBOLX2_2(op3, &bitD3);
@@ -319,10 +276,15 @@ static size_t HUF_decompress4X2_usingDTable_internal(
             HUF_DECODE_SYMBOLX2_0(op2, &bitD2);
             HUF_DECODE_SYMBOLX2_0(op3, &bitD3);
             HUF_DECODE_SYMBOLX2_0(op4, &bitD4);
-            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+            BIT_reloadDStream(&bitD1);
+            BIT_reloadDStream(&bitD2);
+            BIT_reloadDStream(&bitD3);
+            BIT_reloadDStream(&bitD4);
         }
 
         /* check corruption */
+        /* note : should not be necessary : op# advance in lock step, and we control op4.
+         *        but curiously, binary generated by gcc 7.2 & 7.3 with -mbmi2 runs faster when >=1 test is present */
         if (op1 > opStart2) return ERROR(corruption_detected);
         if (op2 > opStart3) return ERROR(corruption_detected);
         if (op3 > opStart4) return ERROR(corruption_detected);
@@ -335,8 +297,8 @@ static size_t HUF_decompress4X2_usingDTable_internal(
         HUF_decodeStreamX2(op4, &bitD4, oend,     dt, dtLog);
 
         /* check */
-        endSignal = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
-        if (!endSignal) return ERROR(corruption_detected);
+        { U32 const endCheck = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+          if (!endCheck) return ERROR(corruption_detected); }
 
         /* decoded size */
         return dstSize;
@@ -344,30 +306,309 @@ static size_t HUF_decompress4X2_usingDTable_internal(
 }
 
 
-size_t HUF_decompress4X2_usingDTable(
+FORCE_INLINE_TEMPLATE U32
+HUF_decodeSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
+{
+    size_t const val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
+    memcpy(op, dt+val, 2);
+    BIT_skipBits(DStream, dt[val].nbBits);
+    return dt[val].length;
+}
+
+FORCE_INLINE_TEMPLATE U32
+HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
+{
+    size_t const val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
+    memcpy(op, dt+val, 1);
+    if (dt[val].length==1) BIT_skipBits(DStream, dt[val].nbBits);
+    else {
+        if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
+            BIT_skipBits(DStream, dt[val].nbBits);
+            if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
+                /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
+                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);
+    }   }
+    return 1;
+}
+
+#define HUF_DECODE_SYMBOLX4_0(ptr, DStreamPtr) \
+    ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX4_1(ptr, DStreamPtr) \
+    if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
+        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+#define HUF_DECODE_SYMBOLX4_2(ptr, DStreamPtr) \
+    if (MEM_64bits()) \
+        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
+
+HINT_INLINE size_t
+HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd,
+                const HUF_DEltX4* const dt, const U32 dtLog)
+{
+    BYTE* const pStart = p;
+
+    /* up to 8 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-(sizeof(bitDPtr->bitContainer)-1))) {
+        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_1(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
+    }
+
+    /* closer to end : up to 2 symbols at a time */
+    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p <= pEnd-2))
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
+
+    while (p <= pEnd-2)
+        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);   /* no need to reload : reached the end of DStream */
+
+    if (p < pEnd)
+        p += HUF_decodeLastSymbolX4(p, bitDPtr, dt, dtLog);
+
+    return p-pStart;
+}
+
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress1X4_usingDTable_internal_body(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    BIT_DStream_t bitD;
+
+    /* Init */
+    CHECK_F( BIT_initDStream(&bitD, cSrc, cSrcSize) );
+
+    /* decode */
+    {   BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+        const void* const dtPtr = DTable+1;   /* force compiler to not use strict-aliasing */
+        const HUF_DEltX4* const dt = (const HUF_DEltX4*)dtPtr;
+        DTableDesc const dtd = HUF_getDTableDesc(DTable);
+        HUF_decodeStreamX4(ostart, &bitD, oend, dt, dtd.tableLog);
+    }
+
+    /* check */
+    if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
+
+    /* decoded size */
+    return dstSize;
+}
+
+
+FORCE_INLINE_TEMPLATE size_t
+HUF_decompress4X4_usingDTable_internal_body(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    if (cSrcSize < 10) return ERROR(corruption_detected);   /* strict minimum : jump table + 1 byte per stream */
+
+    {   const BYTE* const istart = (const BYTE*) cSrc;
+        BYTE* const ostart = (BYTE*) dst;
+        BYTE* const oend = ostart + dstSize;
+        const void* const dtPtr = DTable+1;
+        const HUF_DEltX4* const dt = (const HUF_DEltX4*)dtPtr;
+
+        /* Init */
+        BIT_DStream_t bitD1;
+        BIT_DStream_t bitD2;
+        BIT_DStream_t bitD3;
+        BIT_DStream_t bitD4;
+        size_t const length1 = MEM_readLE16(istart);
+        size_t const length2 = MEM_readLE16(istart+2);
+        size_t const length3 = MEM_readLE16(istart+4);
+        size_t const length4 = cSrcSize - (length1 + length2 + length3 + 6);
+        const BYTE* const istart1 = istart + 6;  /* jumpTable */
+        const BYTE* const istart2 = istart1 + length1;
+        const BYTE* const istart3 = istart2 + length2;
+        const BYTE* const istart4 = istart3 + length3;
+        size_t const segmentSize = (dstSize+3) / 4;
+        BYTE* const opStart2 = ostart + segmentSize;
+        BYTE* const opStart3 = opStart2 + segmentSize;
+        BYTE* const opStart4 = opStart3 + segmentSize;
+        BYTE* op1 = ostart;
+        BYTE* op2 = opStart2;
+        BYTE* op3 = opStart3;
+        BYTE* op4 = opStart4;
+        U32 endSignal;
+        DTableDesc const dtd = HUF_getDTableDesc(DTable);
+        U32 const dtLog = dtd.tableLog;
+
+        if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
+        CHECK_F( BIT_initDStream(&bitD1, istart1, length1) );
+        CHECK_F( BIT_initDStream(&bitD2, istart2, length2) );
+        CHECK_F( BIT_initDStream(&bitD3, istart3, length3) );
+        CHECK_F( BIT_initDStream(&bitD4, istart4, length4) );
+
+        /* 16-32 symbols per loop (4-8 symbols per stream) */
+        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        for ( ; (endSignal==BIT_DStream_unfinished) & (op4<(oend-(sizeof(bitD4.bitContainer)-1))) ; ) {
+            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_1(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_1(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_1(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_1(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
+            HUF_DECODE_SYMBOLX4_0(op1, &bitD1);
+            HUF_DECODE_SYMBOLX4_0(op2, &bitD2);
+            HUF_DECODE_SYMBOLX4_0(op3, &bitD3);
+            HUF_DECODE_SYMBOLX4_0(op4, &bitD4);
+
+            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
+        }
+
+        /* check corruption */
+        if (op1 > opStart2) return ERROR(corruption_detected);
+        if (op2 > opStart3) return ERROR(corruption_detected);
+        if (op3 > opStart4) return ERROR(corruption_detected);
+        /* note : op4 already verified within main loop */
+
+        /* finish bitStreams one by one */
+        HUF_decodeStreamX4(op1, &bitD1, opStart2, dt, dtLog);
+        HUF_decodeStreamX4(op2, &bitD2, opStart3, dt, dtLog);
+        HUF_decodeStreamX4(op3, &bitD3, opStart4, dt, dtLog);
+        HUF_decodeStreamX4(op4, &bitD4, oend,     dt, dtLog);
+
+        /* check */
+        { U32 const endCheck = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
+          if (!endCheck) return ERROR(corruption_detected); }
+
+        /* decoded size */
+        return dstSize;
+    }
+}
+
+
+typedef size_t (*HUF_decompress_usingDTable_t)(void *dst, size_t dstSize,
+                                               const void *cSrc,
+                                               size_t cSrcSize,
+                                               const HUF_DTable *DTable);
+#if DYNAMIC_BMI2
+
+#define X(fn)                                                               \
+                                                                            \
+    static size_t fn##_default(                                             \
+                  void* dst,  size_t dstSize,                               \
+            const void* cSrc, size_t cSrcSize,                              \
+            const HUF_DTable* DTable)                                       \
+    {                                                                       \
+        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
+    }                                                                       \
+                                                                            \
+    static TARGET_ATTRIBUTE("bmi2") size_t fn##_bmi2(                       \
+                  void* dst,  size_t dstSize,                               \
+            const void* cSrc, size_t cSrcSize,                              \
+            const HUF_DTable* DTable)                                       \
+    {                                                                       \
+        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
+    }                                                                       \
+                                                                            \
+    static size_t fn(void* dst, size_t dstSize, void const* cSrc,           \
+                     size_t cSrcSize, HUF_DTable const* DTable, int bmi2)   \
+    {                                                                       \
+        if (bmi2) {                                                         \
+            return fn##_bmi2(dst, dstSize, cSrc, cSrcSize, DTable);         \
+        }                                                                   \
+        return fn##_default(dst, dstSize, cSrc, cSrcSize, DTable);          \
+    }
+
+#else
+
+#define X(fn)                                                               \
+    static size_t fn(void* dst, size_t dstSize, void const* cSrc,           \
+                     size_t cSrcSize, HUF_DTable const* DTable, int bmi2)   \
+    {                                                                       \
+        (void)bmi2;                                                         \
+        return fn##_body(dst, dstSize, cSrc, cSrcSize, DTable);             \
+    }
+
+#endif
+
+X(HUF_decompress1X2_usingDTable_internal)
+X(HUF_decompress4X2_usingDTable_internal)
+X(HUF_decompress1X4_usingDTable_internal)
+X(HUF_decompress4X4_usingDTable_internal)
+
+#undef X
+
+
+size_t HUF_decompress1X2_usingDTable(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
     const HUF_DTable* DTable)
 {
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     if (dtd.tableType != 0) return ERROR(GENERIC);
-    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+    return HUF_decompress1X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
 }
 
-
-size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+size_t HUF_decompress1X2_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
                                    const void* cSrc, size_t cSrcSize,
                                    void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
+    size_t const hSize = HUF_readDTableX2_wksp(DCtx, cSrc, cSrcSize, workSpace, wkspSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress1X2_usingDTable_internal(dst, dstSize, ip, cSrcSize, DCtx, /* bmi2 */ 0);
+}
+
+
+size_t HUF_decompress1X2_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X2_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
+}
+
+size_t HUF_decompress1X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
+    return HUF_decompress1X2_DCtx (DTable, dst, dstSize, cSrc, cSrcSize);
+}
+
+size_t HUF_decompress4X2_usingDTable(
+          void* dst,  size_t dstSize,
+    const void* cSrc, size_t cSrcSize,
+    const HUF_DTable* DTable)
+{
+    DTableDesc dtd = HUF_getDTableDesc(DTable);
+    if (dtd.tableType != 0) return ERROR(GENERIC);
+    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+}
+
+static size_t HUF_decompress4X2_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize, int bmi2)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
     size_t const hSize = HUF_readDTableX2_wksp (dctx, cSrc, cSrcSize,
                                                 workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
 
-    return HUF_decompress4X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, dctx);
+    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, bmi2);
+}
+
+size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
+{
+    return HUF_decompress4X2_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, 0);
 }
 
 
@@ -387,8 +628,6 @@ size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cS
 /* *************************/
 /* double-symbols decoding */
 /* *************************/
-typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX4;  /* double-symbols decoding */
-
 typedef struct { BYTE symbol; BYTE weight; } sortedSymbol_t;
 
 /* HUF_fillDTableX4Level2() :
@@ -508,10 +747,7 @@ size_t HUF_readDTableX4_wksp(HUF_DTable* DTable, const void* src,
     weightList = (BYTE *)((U32 *)workSpace + spaceUsed32);
     spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
 
-    if ((spaceUsed32 << 2) > wkspSize)
-        return ERROR(tableLog_tooLarge);
-    workSpace = (U32 *)workSpace + spaceUsed32;
-    wkspSize -= (spaceUsed32 << 2);
+    if ((spaceUsed32 << 2) > wkspSize) return ERROR(tableLog_tooLarge);
 
     rankStart = rankStart0 + 1;
     memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
@@ -588,95 +824,6 @@ size_t HUF_readDTableX4(HUF_DTable* DTable, const void* src, size_t srcSize)
                                workSpace, sizeof(workSpace));
 }
 
-static U32 HUF_decodeSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
-{
-    size_t const val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
-    memcpy(op, dt+val, 2);
-    BIT_skipBits(DStream, dt[val].nbBits);
-    return dt[val].length;
-}
-
-static U32 HUF_decodeLastSymbolX4(void* op, BIT_DStream_t* DStream, const HUF_DEltX4* dt, const U32 dtLog)
-{
-    size_t const val = BIT_lookBitsFast(DStream, dtLog);   /* note : dtLog >= 1 */
-    memcpy(op, dt+val, 1);
-    if (dt[val].length==1) BIT_skipBits(DStream, dt[val].nbBits);
-    else {
-        if (DStream->bitsConsumed < (sizeof(DStream->bitContainer)*8)) {
-            BIT_skipBits(DStream, dt[val].nbBits);
-            if (DStream->bitsConsumed > (sizeof(DStream->bitContainer)*8))
-                /* ugly hack; works only because it's the last symbol. Note : can't easily extract nbBits from just this symbol */
-                DStream->bitsConsumed = (sizeof(DStream->bitContainer)*8);
-    }   }
-    return 1;
-}
-
-
-#define HUF_DECODE_SYMBOLX4_0(ptr, DStreamPtr) \
-    ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
-
-#define HUF_DECODE_SYMBOLX4_1(ptr, DStreamPtr) \
-    if (MEM_64bits() || (HUF_TABLELOG_MAX<=12)) \
-        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
-
-#define HUF_DECODE_SYMBOLX4_2(ptr, DStreamPtr) \
-    if (MEM_64bits()) \
-        ptr += HUF_decodeSymbolX4(ptr, DStreamPtr, dt, dtLog)
-
-HINT_INLINE size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
-{
-    BYTE* const pStart = p;
-
-    /* up to 8 symbols at a time */
-    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p < pEnd-(sizeof(bitDPtr->bitContainer)-1))) {
-        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
-        HUF_DECODE_SYMBOLX4_1(p, bitDPtr);
-        HUF_DECODE_SYMBOLX4_2(p, bitDPtr);
-        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
-    }
-
-    /* closer to end : up to 2 symbols at a time */
-    while ((BIT_reloadDStream(bitDPtr) == BIT_DStream_unfinished) & (p <= pEnd-2))
-        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);
-
-    while (p <= pEnd-2)
-        HUF_DECODE_SYMBOLX4_0(p, bitDPtr);   /* no need to reload : reached the end of DStream */
-
-    if (p < pEnd)
-        p += HUF_decodeLastSymbolX4(p, bitDPtr, dt, dtLog);
-
-    return p-pStart;
-}
-
-
-static size_t HUF_decompress1X4_usingDTable_internal(
-          void* dst,  size_t dstSize,
-    const void* cSrc, size_t cSrcSize,
-    const HUF_DTable* DTable)
-{
-    BIT_DStream_t bitD;
-
-    /* Init */
-    {   size_t const errorCode = BIT_initDStream(&bitD, cSrc, cSrcSize);
-        if (HUF_isError(errorCode)) return errorCode;
-    }
-
-    /* decode */
-    {   BYTE* const ostart = (BYTE*) dst;
-        BYTE* const oend = ostart + dstSize;
-        const void* const dtPtr = DTable+1;   /* force compiler to not use strict-aliasing */
-        const HUF_DEltX4* const dt = (const HUF_DEltX4*)dtPtr;
-        DTableDesc const dtd = HUF_getDTableDesc(DTable);
-        HUF_decodeStreamX4(ostart, &bitD, oend, dt, dtd.tableLog);
-    }
-
-    /* check */
-    if (!BIT_endOfDStream(&bitD)) return ERROR(corruption_detected);
-
-    /* decoded size */
-    return dstSize;
-}
-
 size_t HUF_decompress1X4_usingDTable(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
@@ -684,7 +831,7 @@ size_t HUF_decompress1X4_usingDTable(
 {
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     if (dtd.tableType != 1) return ERROR(GENERIC);
-    return HUF_decompress1X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+    return HUF_decompress1X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
 }
 
 size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
@@ -699,7 +846,7 @@ size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
 
-    return HUF_decompress1X4_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
+    return HUF_decompress1X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, DCtx, /* bmi2 */ 0);
 }
 
 
@@ -717,99 +864,6 @@ size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cS
     return HUF_decompress1X4_DCtx(DTable, dst, dstSize, cSrc, cSrcSize);
 }
 
-static size_t HUF_decompress4X4_usingDTable_internal(
-          void* dst,  size_t dstSize,
-    const void* cSrc, size_t cSrcSize,
-    const HUF_DTable* DTable)
-{
-    if (cSrcSize < 10) return ERROR(corruption_detected);   /* strict minimum : jump table + 1 byte per stream */
-
-    {   const BYTE* const istart = (const BYTE*) cSrc;
-        BYTE* const ostart = (BYTE*) dst;
-        BYTE* const oend = ostart + dstSize;
-        const void* const dtPtr = DTable+1;
-        const HUF_DEltX4* const dt = (const HUF_DEltX4*)dtPtr;
-
-        /* Init */
-        BIT_DStream_t bitD1;
-        BIT_DStream_t bitD2;
-        BIT_DStream_t bitD3;
-        BIT_DStream_t bitD4;
-        size_t const length1 = MEM_readLE16(istart);
-        size_t const length2 = MEM_readLE16(istart+2);
-        size_t const length3 = MEM_readLE16(istart+4);
-        size_t const length4 = cSrcSize - (length1 + length2 + length3 + 6);
-        const BYTE* const istart1 = istart + 6;  /* jumpTable */
-        const BYTE* const istart2 = istart1 + length1;
-        const BYTE* const istart3 = istart2 + length2;
-        const BYTE* const istart4 = istart3 + length3;
-        size_t const segmentSize = (dstSize+3) / 4;
-        BYTE* const opStart2 = ostart + segmentSize;
-        BYTE* const opStart3 = opStart2 + segmentSize;
-        BYTE* const opStart4 = opStart3 + segmentSize;
-        BYTE* op1 = ostart;
-        BYTE* op2 = opStart2;
-        BYTE* op3 = opStart3;
-        BYTE* op4 = opStart4;
-        U32 endSignal;
-        DTableDesc const dtd = HUF_getDTableDesc(DTable);
-        U32 const dtLog = dtd.tableLog;
-
-        if (length4 > cSrcSize) return ERROR(corruption_detected);   /* overflow */
-        { size_t const errorCode = BIT_initDStream(&bitD1, istart1, length1);
-          if (HUF_isError(errorCode)) return errorCode; }
-        { size_t const errorCode = BIT_initDStream(&bitD2, istart2, length2);
-          if (HUF_isError(errorCode)) return errorCode; }
-        { size_t const errorCode = BIT_initDStream(&bitD3, istart3, length3);
-          if (HUF_isError(errorCode)) return errorCode; }
-        { size_t const errorCode = BIT_initDStream(&bitD4, istart4, length4);
-          if (HUF_isError(errorCode)) return errorCode; }
-
-        /* 16-32 symbols per loop (4-8 symbols per stream) */
-        endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
-        for ( ; (endSignal==BIT_DStream_unfinished) & (op4<(oend-(sizeof(bitD4.bitContainer)-1))) ; ) {
-            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
-            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
-            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
-            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
-            HUF_DECODE_SYMBOLX4_1(op1, &bitD1);
-            HUF_DECODE_SYMBOLX4_1(op2, &bitD2);
-            HUF_DECODE_SYMBOLX4_1(op3, &bitD3);
-            HUF_DECODE_SYMBOLX4_1(op4, &bitD4);
-            HUF_DECODE_SYMBOLX4_2(op1, &bitD1);
-            HUF_DECODE_SYMBOLX4_2(op2, &bitD2);
-            HUF_DECODE_SYMBOLX4_2(op3, &bitD3);
-            HUF_DECODE_SYMBOLX4_2(op4, &bitD4);
-            HUF_DECODE_SYMBOLX4_0(op1, &bitD1);
-            HUF_DECODE_SYMBOLX4_0(op2, &bitD2);
-            HUF_DECODE_SYMBOLX4_0(op3, &bitD3);
-            HUF_DECODE_SYMBOLX4_0(op4, &bitD4);
-
-            endSignal = BIT_reloadDStream(&bitD1) | BIT_reloadDStream(&bitD2) | BIT_reloadDStream(&bitD3) | BIT_reloadDStream(&bitD4);
-        }
-
-        /* check corruption */
-        if (op1 > opStart2) return ERROR(corruption_detected);
-        if (op2 > opStart3) return ERROR(corruption_detected);
-        if (op3 > opStart4) return ERROR(corruption_detected);
-        /* note : op4 already verified within main loop */
-
-        /* finish bitStreams one by one */
-        HUF_decodeStreamX4(op1, &bitD1, opStart2, dt, dtLog);
-        HUF_decodeStreamX4(op2, &bitD2, opStart3, dt, dtLog);
-        HUF_decodeStreamX4(op3, &bitD3, opStart4, dt, dtLog);
-        HUF_decodeStreamX4(op4, &bitD4, oend,     dt, dtLog);
-
-        /* check */
-        { U32 const endCheck = BIT_endOfDStream(&bitD1) & BIT_endOfDStream(&bitD2) & BIT_endOfDStream(&bitD3) & BIT_endOfDStream(&bitD4);
-          if (!endCheck) return ERROR(corruption_detected); }
-
-        /* decoded size */
-        return dstSize;
-    }
-}
-
-
 size_t HUF_decompress4X4_usingDTable(
           void* dst,  size_t dstSize,
     const void* cSrc, size_t cSrcSize,
@@ -817,13 +871,12 @@ size_t HUF_decompress4X4_usingDTable(
 {
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     if (dtd.tableType != 1) return ERROR(GENERIC);
-    return HUF_decompress4X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+    return HUF_decompress4X4_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
 }
 
-
-size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+static size_t HUF_decompress4X4_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize,
                                    const void* cSrc, size_t cSrcSize,
-                                   void* workSpace, size_t wkspSize)
+                                   void* workSpace, size_t wkspSize, int bmi2)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
@@ -833,7 +886,14 @@ size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
 
-    return HUF_decompress4X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx);
+    return HUF_decompress4X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, bmi2);
+}
+
+size_t HUF_decompress4X4_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
+{
+    return HUF_decompress4X4_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, /* bmi2 */ 0);
 }
 
 
@@ -861,8 +921,8 @@ size_t HUF_decompress1X_usingDTable(void* dst, size_t maxDstSize,
                                     const HUF_DTable* DTable)
 {
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
-    return dtd.tableType ? HUF_decompress1X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable) :
-                           HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable);
+    return dtd.tableType ? HUF_decompress1X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0) :
+                           HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
 }
 
 size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize,
@@ -870,8 +930,8 @@ size_t HUF_decompress4X_usingDTable(void* dst, size_t maxDstSize,
                                     const HUF_DTable* DTable)
 {
     DTableDesc const dtd = HUF_getDTableDesc(DTable);
-    return dtd.tableType ? HUF_decompress4X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable) :
-                           HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable);
+    return dtd.tableType ? HUF_decompress4X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0) :
+                           HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
 }
 
 
@@ -898,21 +958,22 @@ static const algo_time_t algoTime[16 /* Quantization */][3 /* single, double, qu
 };
 
 /** HUF_selectDecoder() :
-*   Tells which decoder is likely to decode faster,
-*   based on a set of pre-determined metrics.
-*   @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
-*   Assumption : 0 < cSrcSize, dstSize <= 128 KB */
+ *  Tells which decoder is likely to decode faster,
+ *  based on a set of pre-computed metrics.
+ * @return : 0==HUF_decompress4X2, 1==HUF_decompress4X4 .
+ *  Assumption : 0 < dstSize <= 128 KB */
 U32 HUF_selectDecoder (size_t dstSize, size_t cSrcSize)
 {
+    assert(dstSize > 0);
+    assert(dstSize <= 128 KB);
     /* decoder timing evaluation */
-    U32 const Q = cSrcSize >= dstSize ? 15 : (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 */
-    U32 const D256 = (U32)(dstSize >> 8);
-    U32 const DTime0 = algoTime[Q][0].tableTime + (algoTime[Q][0].decode256Time * D256);
-    U32 DTime1 = algoTime[Q][1].tableTime + (algoTime[Q][1].decode256Time * D256);
-    DTime1 += DTime1 >> 3;  /* advantage to algorithm using less memory, for cache eviction */
-
-    return DTime1 < DTime0;
-}
+    {   U32 const Q = (cSrcSize >= dstSize) ? 15 : (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 */
+        U32 const D256 = (U32)(dstSize >> 8);
+        U32 const DTime0 = algoTime[Q][0].tableTime + (algoTime[Q][0].decode256Time * D256);
+        U32 DTime1 = algoTime[Q][1].tableTime + (algoTime[Q][1].decode256Time * D256);
+        DTime1 += DTime1 >> 3;  /* advantage to algorithm using less memory, to reduce cache eviction */
+        return DTime1 < DTime0;
+}   }
 
 
 typedef size_t (*decompressionAlgo)(void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize);
@@ -994,3 +1055,42 @@ size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
     return HUF_decompress1X_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
                                       workSpace, sizeof(workSpace));
 }
+
+
+size_t HUF_decompress1X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2)
+{
+    DTableDesc const dtd = HUF_getDTableDesc(DTable);
+    return dtd.tableType ? HUF_decompress1X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2) :
+                           HUF_decompress1X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+}
+
+size_t HUF_decompress1X2_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t const hSize = HUF_readDTableX2_wksp(dctx, cSrc, cSrcSize, workSpace, wkspSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress1X2_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, bmi2);
+}
+
+size_t HUF_decompress4X_usingDTable_bmi2(void* dst, size_t maxDstSize, const void* cSrc, size_t cSrcSize, const HUF_DTable* DTable, int bmi2)
+{
+    DTableDesc const dtd = HUF_getDTableDesc(DTable);
+    return dtd.tableType ? HUF_decompress4X4_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2) :
+                           HUF_decompress4X2_usingDTable_internal(dst, maxDstSize, cSrc, cSrcSize, DTable, bmi2);
+}
+
+size_t HUF_decompress4X_hufOnly_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize, void* workSpace, size_t wkspSize, int bmi2)
+{
+    /* validation checks */
+    if (dstSize == 0) return ERROR(dstSize_tooSmall);
+    if (cSrcSize == 0) return ERROR(corruption_detected);
+
+    {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
+        return algoNb ? HUF_decompress4X4_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, bmi2) :
+                        HUF_decompress4X2_DCtx_wksp_bmi2(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize, bmi2);
+    }
+}

src/borg/algorithms/zstd/lib/deprecated/zbuff_compress.c

@@ -72,6 +72,7 @@ size_t ZBUFF_compressInit_advanced(ZBUFF_CCtx* zbc,
                                    const void* dict, size_t dictSize,
                                    ZSTD_parameters params, unsigned long long pledgedSrcSize)
 {
+    if (pledgedSrcSize==0) pledgedSrcSize = ZSTD_CONTENTSIZE_UNKNOWN;  /* preserve "0 == unknown" behavior */
     return ZSTD_initCStream_advanced(zbc, dict, dictSize, params, pledgedSrcSize);
 }
 

src/borg/algorithms/zstd/lib/dictBuilder/cover.c

@@ -537,8 +537,8 @@ static int COVER_ctx_init(COVER_ctx_t *ctx, const void *samplesBuffer,
   /* Checks */
   if (totalSamplesSize < MAX(d, sizeof(U64)) ||
       totalSamplesSize >= (size_t)COVER_MAX_SAMPLES_SIZE) {
-    DISPLAYLEVEL(1, "Total samples size is too large, maximum size is %u MB\n",
-                 (COVER_MAX_SAMPLES_SIZE >> 20));
+    DISPLAYLEVEL(1, "Total samples size is too large (%u MB), maximum size is %u MB\n",
+                 (U32)(totalSamplesSize>>20), (COVER_MAX_SAMPLES_SIZE >> 20));
     return 0;
   }
   /* Zero the context */
@@ -651,12 +651,16 @@ static size_t COVER_buildDictionary(const COVER_ctx_t *ctx, U32 *freqs,
 }
 
 ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
-    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
-    const size_t *samplesSizes, unsigned nbSamples,
-    ZDICT_cover_params_t parameters) {
-  BYTE *const dict = (BYTE *)dictBuffer;
+    void *dictBuffer, size_t dictBufferCapacity,
+    const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_cover_params_t parameters)
+{
+  BYTE* const dict = (BYTE*)dictBuffer;
   COVER_ctx_t ctx;
   COVER_map_t activeDmers;
+
+  /* Initialize global data */
+  g_displayLevel = parameters.zParams.notificationLevel;
   /* Checks */
   if (!COVER_checkParameters(parameters, dictBufferCapacity)) {
     DISPLAYLEVEL(1, "Cover parameters incorrect\n");
@@ -671,8 +675,6 @@ ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
                  ZDICT_DICTSIZE_MIN);
     return ERROR(dstSize_tooSmall);
   }
-  /* Initialize global data */
-  g_displayLevel = parameters.zParams.notificationLevel;
   /* Initialize context and activeDmers */
   if (!COVER_ctx_init(&ctx, samplesBuffer, samplesSizes, nbSamples,
                       parameters.d)) {
@@ -947,6 +949,7 @@ ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
   unsigned k;
   COVER_best_t best;
   POOL_ctx *pool = NULL;
+
   /* Checks */
   if (kMinK < kMaxD || kMaxK < kMinK) {
     LOCALDISPLAYLEVEL(displayLevel, 1, "Incorrect parameters\n");

src/borg/algorithms/zstd/lib/dictBuilder/zdict.c

@@ -103,7 +103,7 @@ unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize)
 /*-********************************************************
 *  Dictionary training functions
 **********************************************************/
-static unsigned ZDICT_NbCommonBytes (register size_t val)
+static unsigned ZDICT_NbCommonBytes (size_t val)
 {
     if (MEM_isLittleEndian()) {
         if (MEM_64bits()) {
@@ -207,7 +207,6 @@ static dictItem ZDICT_analyzePos(
     U32 cumulLength[LLIMIT] = {0};
     U32 savings[LLIMIT] = {0};
     const BYTE* b = (const BYTE*)buffer;
-    size_t length;
     size_t maxLength = LLIMIT;
     size_t pos = suffix[start];
     U32 end = start;
@@ -222,26 +221,30 @@ static dictItem ZDICT_analyzePos(
        ||(MEM_read16(b+pos+1) == MEM_read16(b+pos+3))
        ||(MEM_read16(b+pos+2) == MEM_read16(b+pos+4)) ) {
         /* skip and mark segment */
-        U16 u16 = MEM_read16(b+pos+4);
-        U32 u, e = 6;
-        while (MEM_read16(b+pos+e) == u16) e+=2 ;
-        if (b[pos+e] == b[pos+e-1]) e++;
-        for (u=1; u<e; u++)
+        U16 const pattern16 = MEM_read16(b+pos+4);
+        U32 u, patternEnd = 6;
+        while (MEM_read16(b+pos+patternEnd) == pattern16) patternEnd+=2 ;
+        if (b[pos+patternEnd] == b[pos+patternEnd-1]) patternEnd++;
+        for (u=1; u<patternEnd; u++)
             doneMarks[pos+u] = 1;
         return solution;
     }
 
     /* look forward */
-    do {
-        end++;
-        length = ZDICT_count(b + pos, b + suffix[end]);
-    } while (length >=MINMATCHLENGTH);
+    {   size_t length;
+        do {
+            end++;
+            length = ZDICT_count(b + pos, b + suffix[end]);
+        } while (length >= MINMATCHLENGTH);
+    }
 
     /* look backward */
-    do {
-        length = ZDICT_count(b + pos, b + *(suffix+start-1));
-        if (length >=MINMATCHLENGTH) start--;
-    } while(length >= MINMATCHLENGTH);
+    {   size_t length;
+        do {
+            length = ZDICT_count(b + pos, b + *(suffix+start-1));
+            if (length >=MINMATCHLENGTH) start--;
+        } while(length >= MINMATCHLENGTH);
+    }
 
     /* exit if not found a minimum nb of repetitions */
     if (end-start < minRatio) {
@@ -268,7 +271,7 @@ static dictItem ZDICT_analyzePos(
             U32 selectedCount = 0;
             U32 selectedID = currentID;
             for (id =refinedStart; id < refinedEnd; id++) {
-                if (b[ suffix[id] + searchLength] != currentChar) {
+                if (b[suffix[id] + searchLength] != currentChar) {
                     if (currentCount > selectedCount) {
                         selectedCount = currentCount;
                         selectedID = currentID;
@@ -297,20 +300,23 @@ static dictItem ZDICT_analyzePos(
         memset(lengthList, 0, sizeof(lengthList));
 
         /* look forward */
-        do {
-            end++;
-            length = ZDICT_count(b + pos, b + suffix[end]);
-            if (length >= LLIMIT) length = LLIMIT-1;
-            lengthList[length]++;
-        } while (length >=MINMATCHLENGTH);
+        {   size_t length;
+            do {
+                end++;
+                length = ZDICT_count(b + pos, b + suffix[end]);
+                if (length >= LLIMIT) length = LLIMIT-1;
+                lengthList[length]++;
+            } while (length >=MINMATCHLENGTH);
+        }
 
         /* look backward */
-        length = MINMATCHLENGTH;
-        while ((length >= MINMATCHLENGTH) & (start > 0)) {
-            length = ZDICT_count(b + pos, b + suffix[start - 1]);
-            if (length >= LLIMIT) length = LLIMIT - 1;
-            lengthList[length]++;
-            if (length >= MINMATCHLENGTH) start--;
+        {   size_t length = MINMATCHLENGTH;
+            while ((length >= MINMATCHLENGTH) & (start > 0)) {
+                length = ZDICT_count(b + pos, b + suffix[start - 1]);
+                if (length >= LLIMIT) length = LLIMIT - 1;
+                lengthList[length]++;
+                if (length >= MINMATCHLENGTH) start--;
+            }
         }
 
         /* largest useful length */
@@ -345,12 +351,12 @@ static dictItem ZDICT_analyzePos(
         /* mark positions done */
         {   U32 id;
             for (id=start; id<end; id++) {
-                U32 p, pEnd;
+                U32 p, pEnd, length;
                 U32 const testedPos = suffix[id];
                 if (testedPos == pos)
                     length = solution.length;
                 else {
-                    length = ZDICT_count(b+pos, b+testedPos);
+                    length = (U32)ZDICT_count(b+pos, b+testedPos);
                     if (length > solution.length) length = solution.length;
                 }
                 pEnd = (U32)(testedPos + length);
@@ -575,29 +581,30 @@ static void ZDICT_fillNoise(void* buffer, size_t length)
 
 typedef struct
 {
-    ZSTD_CCtx* ref;
-    ZSTD_CCtx* zc;
+    ZSTD_CCtx* ref;    /* contains reference to dictionary */
+    ZSTD_CCtx* zc;     /* working context */
     void* workPlace;   /* must be ZSTD_BLOCKSIZE_MAX allocated */
 } EStats_ress_t;
 
 #define MAXREPOFFSET 1024
 
 static void ZDICT_countEStats(EStats_ress_t esr, ZSTD_parameters params,
-                            U32* countLit, U32* offsetcodeCount, U32* matchlengthCount, U32* litlengthCount, U32* repOffsets,
-                            const void* src, size_t srcSize, U32 notificationLevel)
+                              U32* countLit, U32* offsetcodeCount, U32* matchlengthCount, U32* litlengthCount, U32* repOffsets,
+                              const void* src, size_t srcSize,
+                              U32 notificationLevel)
 {
     size_t const blockSizeMax = MIN (ZSTD_BLOCKSIZE_MAX, 1 << params.cParams.windowLog);
     size_t cSize;
 
     if (srcSize > blockSizeMax) srcSize = blockSizeMax;   /* protection vs large samples */
-    {  size_t const errorCode = ZSTD_copyCCtx(esr.zc, esr.ref, 0);
-            if (ZSTD_isError(errorCode)) { DISPLAYLEVEL(1, "warning : ZSTD_copyCCtx failed \n"); return; }
+    {   size_t const errorCode = ZSTD_copyCCtx(esr.zc, esr.ref, 0);
+        if (ZSTD_isError(errorCode)) { DISPLAYLEVEL(1, "warning : ZSTD_copyCCtx failed \n"); return; }
     }
     cSize = ZSTD_compressBlock(esr.zc, esr.workPlace, ZSTD_BLOCKSIZE_MAX, src, srcSize);
     if (ZSTD_isError(cSize)) { DISPLAYLEVEL(3, "warning : could not compress sample size %u \n", (U32)srcSize); return; }
 
     if (cSize) {  /* if == 0; block is not compressible */
-        const seqStore_t* seqStorePtr = ZSTD_getSeqStore(esr.zc);
+        const seqStore_t* const seqStorePtr = ZSTD_getSeqStore(esr.zc);
 
         /* literals stats */
         {   const BYTE* bytePtr;
@@ -659,6 +666,18 @@ static void ZDICT_insertSortCount(offsetCount_t table[ZSTD_REP_NUM+1], U32 val,
     }
 }
 
+/* ZDICT_flatLit() :
+ * rewrite `countLit` to contain a mostly flat but still compressible distribution of literals.
+ * necessary to avoid generating a non-compressible distribution that HUF_writeCTable() cannot encode.
+ */
+static void ZDICT_flatLit(U32* countLit)
+{
+    int u;
+    for (u=1; u<256; u++) countLit[u] = 2;
+    countLit[0]   = 4;
+    countLit[253] = 1;
+    countLit[254] = 1;
+}
 
 #define OFFCODE_MAX 30  /* only applicable to first block */
 static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
@@ -688,6 +707,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
     BYTE* dstPtr = (BYTE*)dstBuffer;
 
     /* init */
+    DEBUGLOG(4, "ZDICT_analyzeEntropy");
     esr.ref = ZSTD_createCCtx();
     esr.zc = ZSTD_createCCtx();
     esr.workPlace = malloc(ZSTD_BLOCKSIZE_MAX);
@@ -713,7 +733,7 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
             goto _cleanup;
     }   }
 
-    /* collect stats on all files */
+    /* collect stats on all samples */
     for (u=0; u<nbFiles; u++) {
         ZDICT_countEStats(esr, params,
                           countLit, offcodeCount, matchLengthCount, litLengthCount, repOffset,
@@ -722,14 +742,21 @@ static size_t ZDICT_analyzeEntropy(void*  dstBuffer, size_t maxDstSize,
         pos += fileSizes[u];
     }
 
-    /* analyze */
-    errorCode = HUF_buildCTable (hufTable, countLit, 255, huffLog);
-    if (HUF_isError(errorCode)) {
-        eSize = ERROR(GENERIC);
-        DISPLAYLEVEL(1, "HUF_buildCTable error \n");
-        goto _cleanup;
+    /* analyze, build stats, starting with literals */
+    {   size_t maxNbBits = HUF_buildCTable (hufTable, countLit, 255, huffLog);
+        if (HUF_isError(maxNbBits)) {
+            eSize = ERROR(GENERIC);
+            DISPLAYLEVEL(1, " HUF_buildCTable error \n");
+            goto _cleanup;
+        }
+        if (maxNbBits==8) {  /* not compressible : will fail on HUF_writeCTable() */
+            DISPLAYLEVEL(2, "warning : pathological dataset : literals are not compressible : samples are noisy or too regular \n");
+            ZDICT_flatLit(countLit);  /* replace distribution by a fake "mostly flat but still compressible" distribution, that HUF_writeCTable() can encode */
+            maxNbBits = HUF_buildCTable (hufTable, countLit, 255, huffLog);
+            assert(maxNbBits==9);
+        }
+        huffLog = (U32)maxNbBits;
     }
-    huffLog = (U32)errorCode;
 
     /* looking for most common first offsets */
     {   U32 offset;
@@ -850,6 +877,7 @@ size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
     U32 const notificationLevel = params.notificationLevel;
 
     /* check conditions */
+    DEBUGLOG(4, "ZDICT_finalizeDictionary");
     if (dictBufferCapacity < dictContentSize) return ERROR(dstSize_tooSmall);
     if (dictContentSize < ZDICT_CONTENTSIZE_MIN) return ERROR(srcSize_wrong);
     if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) return ERROR(dstSize_tooSmall);
@@ -1025,8 +1053,9 @@ size_t ZDICT_trainFromBuffer_unsafe_legacy(
 }
 
 
-/* issue : samplesBuffer need to be followed by a noisy guard band.
-*  work around : duplicate the buffer, and add the noise */
+/* ZDICT_trainFromBuffer_legacy() :
+ * issue : samplesBuffer need to be followed by a noisy guard band.
+ * work around : duplicate the buffer, and add the noise */
 size_t ZDICT_trainFromBuffer_legacy(void* dictBuffer, size_t dictBufferCapacity,
                               const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
                               ZDICT_legacy_params_t params)
@@ -1054,18 +1083,22 @@ size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
                              const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
 {
     ZDICT_cover_params_t params;
+    DEBUGLOG(3, "ZDICT_trainFromBuffer");
     memset(&params, 0, sizeof(params));
     params.d = 8;
     params.steps = 4;
-    /* Default to level 6 since no compression level information is avaialble */
+    /* Default to level 6 since no compression level information is available */
     params.zParams.compressionLevel = 6;
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
+    params.zParams.notificationLevel = ZSTD_DEBUG;
+#endif
     return ZDICT_optimizeTrainFromBuffer_cover(dictBuffer, dictBufferCapacity,
-                                               samplesBuffer, samplesSizes,
-                                               nbSamples, &params);
+                                               samplesBuffer, samplesSizes, nbSamples,
+                                               &params);
 }
 
 size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
-                                        const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
+                                  const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples)
 {
     ZDICT_params_t params;
     memset(&params, 0, sizeof(params));

src/borg/algorithms/zstd/lib/dictBuilder/zdict.h

@@ -38,21 +38,21 @@ extern "C" {
 
 
 /*! ZDICT_trainFromBuffer():
- * Train a dictionary from an array of samples.
- * Uses ZDICT_optimizeTrainFromBuffer_cover() single-threaded, with d=8 and steps=4.
- * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
- * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
- * The resulting dictionary will be saved into `dictBuffer`.
+ *  Train a dictionary from an array of samples.
+ *  Redirect towards ZDICT_optimizeTrainFromBuffer_cover() single-threaded, with d=8 and steps=4.
+ *  Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ *  supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ *  The resulting dictionary will be saved into `dictBuffer`.
  * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
- *           or an error code, which can be tested with ZDICT_isError().
- * Note: ZDICT_trainFromBuffer() requires about 9 bytes of memory for each input byte.
- * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
- *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
- *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *          or an error code, which can be tested with ZDICT_isError().
+ *  Note: ZDICT_trainFromBuffer() requires about 9 bytes of memory for each input byte.
+ *  Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, though this can vary a lot.
  *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
  */
 ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
-                       const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
+                                    const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
 
 
 /*======   Helper functions   ======*/
@@ -72,14 +72,14 @@ ZDICTLIB_API const char* ZDICT_getErrorName(size_t errorCode);
  * ==================================================================================== */
 
 typedef struct {
-    int      compressionLevel;   /* 0 means default; target a specific zstd compression level */
-    unsigned notificationLevel;  /* Write to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
-    unsigned dictID;             /* 0 means auto mode (32-bits random value); other : force dictID value */
+    int      compressionLevel;   /* optimize for a specific zstd compression level; 0 means default */
+    unsigned notificationLevel;  /* Write log to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
+    unsigned dictID;             /* force dictID value; 0 means auto mode (32-bits random value) */
 } ZDICT_params_t;
 
 /*! ZDICT_cover_params_t:
- *  For all values 0 means default.
  *  k and d are the only required parameters.
+ *  For others, value 0 means default.
  */
 typedef struct {
     unsigned k;                  /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
@@ -91,28 +91,28 @@ typedef struct {
 
 
 /*! ZDICT_trainFromBuffer_cover():
- * Train a dictionary from an array of samples using the COVER algorithm.
- * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
- * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
- * The resulting dictionary will be saved into `dictBuffer`.
+ *  Train a dictionary from an array of samples using the COVER algorithm.
+ *  Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ *  supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ *  The resulting dictionary will be saved into `dictBuffer`.
  * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
- *           or an error code, which can be tested with ZDICT_isError().
- * Note: ZDICT_trainFromBuffer_cover() requires about 9 bytes of memory for each input byte.
- * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
- *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
- *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *          or an error code, which can be tested with ZDICT_isError().
+ *  Note: ZDICT_trainFromBuffer_cover() requires about 9 bytes of memory for each input byte.
+ *  Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, though this can vary a lot.
  *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
  */
 ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
-    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
-    const size_t *samplesSizes, unsigned nbSamples,
-    ZDICT_cover_params_t parameters);
+          void *dictBuffer, size_t dictBufferCapacity,
+    const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
+          ZDICT_cover_params_t parameters);
 
 /*! ZDICT_optimizeTrainFromBuffer_cover():
  * The same requirements as above hold for all the parameters except `parameters`.
  * This function tries many parameter combinations and picks the best parameters.
- * `*parameters` is filled with the best parameters found, and the dictionary
- * constructed with those parameters is stored in `dictBuffer`.
+ * `*parameters` is filled with the best parameters found,
+ * dictionary constructed with those parameters is stored in `dictBuffer`.
  *
  * All of the parameters d, k, steps are optional.
  * If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8, 10, 12, 14, 16}.
@@ -125,9 +125,9 @@ ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
  * Note: ZDICT_optimizeTrainFromBuffer_cover() requires about 8 bytes of memory for each input byte and additionally another 5 bytes of memory for each byte of memory for each thread.
  */
 ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
-    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
-    const size_t *samplesSizes, unsigned nbSamples,
-    ZDICT_cover_params_t *parameters);
+          void* dictBuffer, size_t dictBufferCapacity,
+    const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
+          ZDICT_cover_params_t* parameters);
 
 /*! ZDICT_finalizeDictionary():
  * Given a custom content as a basis for dictionary, and a set of samples,
@@ -157,22 +157,23 @@ typedef struct {
 } ZDICT_legacy_params_t;
 
 /*! ZDICT_trainFromBuffer_legacy():
- * Train a dictionary from an array of samples.
- * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
- * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
- * The resulting dictionary will be saved into `dictBuffer`.
+ *  Train a dictionary from an array of samples.
+ *  Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
+ *  supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
+ *  The resulting dictionary will be saved into `dictBuffer`.
  * `parameters` is optional and can be provided with values set to 0 to mean "default".
  * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
- *           or an error code, which can be tested with ZDICT_isError().
- * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
- *        It's obviously possible to target smaller or larger ones, just by specifying different `dictBufferCapacity`.
- *        In general, it's recommended to provide a few thousands samples, but this can vary a lot.
+ *          or an error code, which can be tested with ZDICT_isError().
+ *  Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
+ *        It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
+ *        In general, it's recommended to provide a few thousands samples, though this can vary a lot.
  *        It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
- * Note: ZDICT_trainFromBuffer_legacy() will send notifications into stderr if instructed to, using notificationLevel>0.
+ *  Note: ZDICT_trainFromBuffer_legacy() will send notifications into stderr if instructed to, using notificationLevel>0.
  */
 ZDICTLIB_API size_t ZDICT_trainFromBuffer_legacy(
-    void *dictBuffer, size_t dictBufferCapacity, const void *samplesBuffer,
-    const size_t *samplesSizes, unsigned nbSamples, ZDICT_legacy_params_t parameters);
+    void *dictBuffer, size_t dictBufferCapacity,
+    const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
+    ZDICT_legacy_params_t parameters);
 
 /* Deprecation warnings */
 /* It is generally possible to disable deprecation warnings from compiler,

src/borg/algorithms/zstd/lib/legacy/zstd_legacy.h

@@ -246,6 +246,7 @@ MEM_STATIC size_t ZSTD_freeLegacyStreamContext(void* legacyContext, U32 version)
 MEM_STATIC size_t ZSTD_initLegacyStream(void** legacyContext, U32 prevVersion, U32 newVersion,
                                         const void* dict, size_t dictSize)
 {
+    DEBUGLOG(5, "ZSTD_initLegacyStream for v0.%u", newVersion);
     if (prevVersion != newVersion) ZSTD_freeLegacyStreamContext(*legacyContext, prevVersion);
     switch(newVersion)
     {
@@ -304,6 +305,7 @@ MEM_STATIC size_t ZSTD_initLegacyStream(void** legacyContext, U32 prevVersion, U
 MEM_STATIC size_t ZSTD_decompressLegacyStream(void* legacyContext, U32 version,
                                               ZSTD_outBuffer* output, ZSTD_inBuffer* input)
 {
+    DEBUGLOG(5, "ZSTD_decompressLegacyStream for v0.%u", version);
     switch(version)
     {
         default :

src/borg/algorithms/zstd/lib/legacy/zstd_v01.c

@@ -339,7 +339,7 @@ typedef U32 DTable_max_t[FSE_DTABLE_SIZE_U32(FSE_MAX_TABLELOG)];
 /****************************************************************
 *  Internal functions
 ****************************************************************/
-FORCE_INLINE unsigned FSE_highbit32 (register U32 val)
+FORCE_INLINE unsigned FSE_highbit32 (U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r;

src/borg/algorithms/zstd/lib/legacy/zstd_v02.c

@@ -330,18 +330,6 @@ MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD);
 MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* bitD);
 
 
-/*
-* Start by invoking BIT_initDStream().
-* A chunk of the bitStream is then stored into a local register.
-* Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
-* You can then retrieve bitFields stored into the local register, **in reverse order**.
-* Local register is manually filled from memory by the BIT_reloadDStream() method.
-* A reload guarantee a minimum of ((8*sizeof(size_t))-7) bits when its result is BIT_DStream_unfinished.
-* Otherwise, it can be less than that, so proceed accordingly.
-* Checking if DStream has reached its end can be performed with BIT_endOfDStream()
-*/
-
-
 /******************************************
 *  unsafe API
 ******************************************/
@@ -353,7 +341,7 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
 /****************************************************************
 *  Helper functions
 ****************************************************************/
-MEM_STATIC unsigned BIT_highbit32 (register U32 val)
+MEM_STATIC unsigned BIT_highbit32 (U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r=0;
@@ -427,13 +415,6 @@ MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, si
     return srcSize;
 }
 
-/*!BIT_lookBits
- * Provides next n bits from local register
- * local register is not modified (bits are still present for next read/look)
- * On 32-bits, maxNbBits==25
- * On 64-bits, maxNbBits==57
- * @return : value extracted
- */
 MEM_STATIC size_t BIT_lookBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
@@ -453,11 +434,6 @@ MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-/*!BIT_readBits
- * Read next n bits from local register.
- * pay attention to not read more than nbBits contained into local register.
- * @return : extracted value.
- */
 MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     size_t value = BIT_lookBits(bitD, nbBits);
@@ -695,55 +671,6 @@ static unsigned char FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bi
 
 static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr);
 
-/*
-Let's now decompose FSE_decompress_usingDTable() into its unitary components.
-You will decode FSE-encoded symbols from the bitStream,
-and also any other bitFields you put in, **in reverse order**.
-
-You will need a few variables to track your bitStream. They are :
-
-BIT_DStream_t DStream;    // Stream context
-FSE_DState_t  DState;     // State context. Multiple ones are possible
-FSE_DTable*   DTablePtr;  // Decoding table, provided by FSE_buildDTable()
-
-The first thing to do is to init the bitStream.
-    errorCode = BIT_initDStream(&DStream, srcBuffer, srcSize);
-
-You should then retrieve your initial state(s)
-(in reverse flushing order if you have several ones) :
-    errorCode = FSE_initDState(&DState, &DStream, DTablePtr);
-
-You can then decode your data, symbol after symbol.
-For information the maximum number of bits read by FSE_decodeSymbol() is 'tableLog'.
-Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
-    unsigned char symbol = FSE_decodeSymbol(&DState, &DStream);
-
-You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
-Note : maximum allowed nbBits is 25, for 32-bits compatibility
-    size_t bitField = BIT_readBits(&DStream, nbBits);
-
-All above operations only read from local register (which size depends on size_t).
-Refueling the register from memory is manually performed by the reload method.
-    endSignal = FSE_reloadDStream(&DStream);
-
-BIT_reloadDStream() result tells if there is still some more data to read from DStream.
-BIT_DStream_unfinished : there is still some data left into the DStream.
-BIT_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
-BIT_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
-BIT_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
-
-When reaching end of buffer (BIT_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
-to properly detect the exact end of stream.
-After each decoded symbol, check if DStream is fully consumed using this simple test :
-    BIT_reloadDStream(&DStream) >= BIT_DStream_completed
-
-When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
-Checking if DStream has reached its end is performed by :
-    BIT_endOfDStream(&DStream);
-Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
-    FSE_endOfDState(&DState);
-*/
-
 
 /******************************************
 *  FSE unsafe API

src/borg/algorithms/zstd/lib/legacy/zstd_v03.c

@@ -332,17 +332,6 @@ MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD);
 MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* bitD);
 
 
-/*
-* Start by invoking BIT_initDStream().
-* A chunk of the bitStream is then stored into a local register.
-* Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
-* You can then retrieve bitFields stored into the local register, **in reverse order**.
-* Local register is manually filled from memory by the BIT_reloadDStream() method.
-* A reload guarantee a minimum of ((8*sizeof(size_t))-7) bits when its result is BIT_DStream_unfinished.
-* Otherwise, it can be less than that, so proceed accordingly.
-* Checking if DStream has reached its end can be performed with BIT_endOfDStream()
-*/
-
 
 /******************************************
 *  unsafe API
@@ -355,7 +344,7 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
 /****************************************************************
 *  Helper functions
 ****************************************************************/
-MEM_STATIC unsigned BIT_highbit32 (register U32 val)
+MEM_STATIC unsigned BIT_highbit32 (U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r=0;
@@ -428,14 +417,6 @@ MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, si
 
     return srcSize;
 }
-
-/*!BIT_lookBits
- * Provides next n bits from local register
- * local register is not modified (bits are still present for next read/look)
- * On 32-bits, maxNbBits==25
- * On 64-bits, maxNbBits==57
- * @return : value extracted
- */
 MEM_STATIC size_t BIT_lookBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
@@ -455,11 +436,6 @@ MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-/*!BIT_readBits
- * Read next n bits from local register.
- * pay attention to not read more than nbBits contained into local register.
- * @return : extracted value.
- */
 MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     size_t value = BIT_lookBits(bitD, nbBits);
@@ -697,55 +673,6 @@ static unsigned char FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bi
 
 static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr);
 
-/*
-Let's now decompose FSE_decompress_usingDTable() into its unitary components.
-You will decode FSE-encoded symbols from the bitStream,
-and also any other bitFields you put in, **in reverse order**.
-
-You will need a few variables to track your bitStream. They are :
-
-BIT_DStream_t DStream;    // Stream context
-FSE_DState_t  DState;     // State context. Multiple ones are possible
-FSE_DTable*   DTablePtr;  // Decoding table, provided by FSE_buildDTable()
-
-The first thing to do is to init the bitStream.
-    errorCode = BIT_initDStream(&DStream, srcBuffer, srcSize);
-
-You should then retrieve your initial state(s)
-(in reverse flushing order if you have several ones) :
-    errorCode = FSE_initDState(&DState, &DStream, DTablePtr);
-
-You can then decode your data, symbol after symbol.
-For information the maximum number of bits read by FSE_decodeSymbol() is 'tableLog'.
-Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
-    unsigned char symbol = FSE_decodeSymbol(&DState, &DStream);
-
-You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
-Note : maximum allowed nbBits is 25, for 32-bits compatibility
-    size_t bitField = BIT_readBits(&DStream, nbBits);
-
-All above operations only read from local register (which size depends on size_t).
-Refueling the register from memory is manually performed by the reload method.
-    endSignal = FSE_reloadDStream(&DStream);
-
-BIT_reloadDStream() result tells if there is still some more data to read from DStream.
-BIT_DStream_unfinished : there is still some data left into the DStream.
-BIT_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
-BIT_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
-BIT_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
-
-When reaching end of buffer (BIT_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
-to properly detect the exact end of stream.
-After each decoded symbol, check if DStream is fully consumed using this simple test :
-    BIT_reloadDStream(&DStream) >= BIT_DStream_completed
-
-When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
-Checking if DStream has reached its end is performed by :
-    BIT_endOfDStream(&DStream);
-Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
-    FSE_endOfDState(&DState);
-*/
-
 
 /******************************************
 *  FSE unsafe API

src/borg/algorithms/zstd/lib/legacy/zstd_v04.c

@@ -16,37 +16,6 @@
 
 /* ******************************************************************
    mem.h
-   low-level memory access routines
-   Copyright (C) 2013-2015, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - FSE source repository : https://github.com/Cyan4973/FiniteStateEntropy
-    - Public forum : https://groups.google.com/forum/#!forum/lz4c
 ****************************************************************** */
 #ifndef MEM_H_MODULE
 #define MEM_H_MODULE
@@ -103,6 +72,44 @@ extern "C" {
 #endif
 
 
+/*-*************************************
+*  Debug
+***************************************/
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=1)
+#  include <assert.h>
+#else
+#  ifndef assert
+#    define assert(condition) ((void)0)
+#  endif
+#endif
+
+#define ZSTD_STATIC_ASSERT(c) { enum { ZSTD_static_assert = 1/(int)(!!(c)) }; }
+
+#if defined(ZSTD_DEBUG) && (ZSTD_DEBUG>=2)
+#  include <stdio.h>
+extern int g_debuglog_enable;
+/* recommended values for ZSTD_DEBUG display levels :
+ * 1 : no display, enables assert() only
+ * 2 : reserved for currently active debug path
+ * 3 : events once per object lifetime (CCtx, CDict, etc.)
+ * 4 : events once per frame
+ * 5 : events once per block
+ * 6 : events once per sequence (*very* verbose) */
+#  define RAWLOG(l, ...) {                                      \
+                if ((g_debuglog_enable) & (l<=ZSTD_DEBUG)) {    \
+                    fprintf(stderr, __VA_ARGS__);               \
+            }   }
+#  define DEBUGLOG(l, ...) {                                    \
+                if ((g_debuglog_enable) & (l<=ZSTD_DEBUG)) {    \
+                    fprintf(stderr, __FILE__ ": " __VA_ARGS__); \
+                    fprintf(stderr, " \n");                     \
+            }   }
+#else
+#  define RAWLOG(l, ...)      {}    /* disabled */
+#  define DEBUGLOG(l, ...)    {}    /* disabled */
+#endif
+
+
 /****************************************************************
 *  Memory I/O
 *****************************************************************/
@@ -255,34 +262,6 @@ MEM_STATIC size_t MEM_readLEST(const void* memPtr)
 /*
     zstd - standard compression library
     Header File for static linking only
-    Copyright (C) 2014-2015, Yann Collet.
-
-    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-    Redistribution and use in source and binary forms, with or without
-    modification, are permitted provided that the following conditions are
-    met:
-    * Redistributions of source code must retain the above copyright
-    notice, this list of conditions and the following disclaimer.
-    * Redistributions in binary form must reproduce the above
-    copyright notice, this list of conditions and the following disclaimer
-    in the documentation and/or other materials provided with the
-    distribution.
-    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - zstd source repository : https://github.com/Cyan4973/zstd
-    - ztsd public forum : https://groups.google.com/forum/#!forum/lz4c
 */
 #ifndef ZSTD_STATIC_H
 #define ZSTD_STATIC_H
@@ -392,34 +371,6 @@ static size_t ZSTD_decompressContinue(ZSTD_DCtx* dctx, void* dst, size_t maxDstS
 /*
     zstd_internal - common functions to include
     Header File for include
-    Copyright (C) 2014-2015, Yann Collet.
-
-    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-    Redistribution and use in source and binary forms, with or without
-    modification, are permitted provided that the following conditions are
-    met:
-    * Redistributions of source code must retain the above copyright
-    notice, this list of conditions and the following disclaimer.
-    * Redistributions in binary form must reproduce the above
-    copyright notice, this list of conditions and the following disclaimer
-    in the documentation and/or other materials provided with the
-    distribution.
-    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-    You can contact the author at :
-    - zstd source repository : https://github.com/Cyan4973/zstd
-    - ztsd public forum : https://groups.google.com/forum/#!forum/lz4c
 */
 #ifndef ZSTD_CCOMMON_H_MODULE
 #define ZSTD_CCOMMON_H_MODULE
@@ -507,36 +458,6 @@ static void ZSTD_wildcopy(void* dst, const void* src, ptrdiff_t length)
 /* ******************************************************************
    FSE : Finite State Entropy coder
    header file
-   Copyright (C) 2013-2015, Yann Collet.
-
-   BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
-
-   Redistribution and use in source and binary forms, with or without
-   modification, are permitted provided that the following conditions are
-   met:
-
-       * Redistributions of source code must retain the above copyright
-   notice, this list of conditions and the following disclaimer.
-       * Redistributions in binary form must reproduce the above
-   copyright notice, this list of conditions and the following disclaimer
-   in the documentation and/or other materials provided with the
-   distribution.
-
-   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
-   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
-   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
-   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
-   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
-   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
-   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
-   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
-   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
-   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
-   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
-
-   You can contact the author at :
-   - Source repository : https://github.com/Cyan4973/FiniteStateEntropy
-   - Public forum : https://groups.google.com/forum/#!forum/lz4c
 ****************************************************************** */
 #ifndef FSE_H
 #define FSE_H
@@ -738,16 +659,6 @@ MEM_STATIC BIT_DStream_status BIT_reloadDStream(BIT_DStream_t* bitD);
 MEM_STATIC unsigned BIT_endOfDStream(const BIT_DStream_t* bitD);
 
 
-/*
-* Start by invoking BIT_initDStream().
-* A chunk of the bitStream is then stored into a local register.
-* Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
-* You can then retrieve bitFields stored into the local register, **in reverse order**.
-* Local register is manually filled from memory by the BIT_reloadDStream() method.
-* A reload guarantee a minimum of ((8*sizeof(size_t))-7) bits when its result is BIT_DStream_unfinished.
-* Otherwise, it can be less than that, so proceed accordingly.
-* Checking if DStream has reached its end can be performed with BIT_endOfDStream()
-*/
 
 
 /******************************************
@@ -761,7 +672,7 @@ MEM_STATIC size_t BIT_readBitsFast(BIT_DStream_t* bitD, unsigned nbBits);
 /****************************************************************
 *  Helper functions
 ****************************************************************/
-MEM_STATIC unsigned BIT_highbit32 (register U32 val)
+MEM_STATIC unsigned BIT_highbit32 (U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r=0;
@@ -834,13 +745,6 @@ MEM_STATIC size_t BIT_initDStream(BIT_DStream_t* bitD, const void* srcBuffer, si
     return srcSize;
 }
 
-/*!BIT_lookBits
- * Provides next n bits from local register
- * local register is not modified (bits are still present for next read/look)
- * On 32-bits, maxNbBits==25
- * On 64-bits, maxNbBits==57
- * @return : value extracted
- */
 MEM_STATIC size_t BIT_lookBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
@@ -860,11 +764,6 @@ MEM_STATIC void BIT_skipBits(BIT_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-/*!BIT_readBits
- * Read next n bits from local register.
- * pay attention to not read more than nbBits contained into local register.
- * @return : extracted value.
- */
 MEM_STATIC size_t BIT_readBits(BIT_DStream_t* bitD, U32 nbBits)
 {
     size_t value = BIT_lookBits(bitD, nbBits);
@@ -1011,55 +910,6 @@ static unsigned char FSE_decodeSymbol(FSE_DState_t* DStatePtr, BIT_DStream_t* bi
 
 static unsigned FSE_endOfDState(const FSE_DState_t* DStatePtr);
 
-/*!
-Let's now decompose FSE_decompress_usingDTable() into its unitary components.
-You will decode FSE-encoded symbols from the bitStream,
-and also any other bitFields you put in, **in reverse order**.
-
-You will need a few variables to track your bitStream. They are :
-
-BIT_DStream_t DStream;    // Stream context
-FSE_DState_t  DState;     // State context. Multiple ones are possible
-FSE_DTable*   DTablePtr;  // Decoding table, provided by FSE_buildDTable()
-
-The first thing to do is to init the bitStream.
-    errorCode = BIT_initDStream(&DStream, srcBuffer, srcSize);
-
-You should then retrieve your initial state(s)
-(in reverse flushing order if you have several ones) :
-    errorCode = FSE_initDState(&DState, &DStream, DTablePtr);
-
-You can then decode your data, symbol after symbol.
-For information the maximum number of bits read by FSE_decodeSymbol() is 'tableLog'.
-Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
-    unsigned char symbol = FSE_decodeSymbol(&DState, &DStream);
-
-You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
-Note : maximum allowed nbBits is 25, for 32-bits compatibility
-    size_t bitField = BIT_readBits(&DStream, nbBits);
-
-All above operations only read from local register (which size depends on size_t).
-Refueling the register from memory is manually performed by the reload method.
-    endSignal = FSE_reloadDStream(&DStream);
-
-BIT_reloadDStream() result tells if there is still some more data to read from DStream.
-BIT_DStream_unfinished : there is still some data left into the DStream.
-BIT_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
-BIT_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
-BIT_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
-
-When reaching end of buffer (BIT_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
-to properly detect the exact end of stream.
-After each decoded symbol, check if DStream is fully consumed using this simple test :
-    BIT_reloadDStream(&DStream) >= BIT_DStream_completed
-
-When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
-Checking if DStream has reached its end is performed by :
-    BIT_endOfDStream(&DStream);
-Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
-    FSE_endOfDState(&DState);
-*/
-
 
 /* *****************************************
 *  FSE unsafe API
@@ -3599,12 +3449,14 @@ static size_t ZBUFF_decompressContinue(ZBUFF_DCtx* zbc, void* dst, size_t* maxDs
     char* const oend = ostart + *maxDstSizePtr;
     U32 notDone = 1;
 
+    DEBUGLOG(5, "ZBUFF_decompressContinue");
     while (notDone)
     {
         switch(zbc->stage)
         {
 
         case ZBUFFds_init :
+            DEBUGLOG(5, "ZBUFF_decompressContinue: stage==ZBUFFds_init => ERROR(init_missing)");
             return ERROR(init_missing);
 
         case ZBUFFds_readHeader :
@@ -3804,7 +3656,7 @@ size_t ZSTDv04_decompressContinue(ZSTDv04_Dctx* dctx, void* dst, size_t maxDstSi
 
 
 ZBUFFv04_DCtx* ZBUFFv04_createDCtx(void) { return ZBUFF_createDCtx(); }
-size_t      ZBUFFv04_freeDCtx(ZBUFFv04_DCtx* dctx) { return ZBUFF_freeDCtx(dctx); }
+size_t ZBUFFv04_freeDCtx(ZBUFFv04_DCtx* dctx) { return ZBUFF_freeDCtx(dctx); }
 
 size_t ZBUFFv04_decompressInit(ZBUFFv04_DCtx* dctx) { return ZBUFF_decompressInit(dctx); }
 size_t ZBUFFv04_decompressWithDictionary(ZBUFFv04_DCtx* dctx, const void* src, size_t srcSize)
@@ -3812,6 +3664,7 @@ size_t ZBUFFv04_decompressWithDictionary(ZBUFFv04_DCtx* dctx, const void* src, s
 
 size_t ZBUFFv04_decompressContinue(ZBUFFv04_DCtx* dctx, void* dst, size_t* maxDstSizePtr, const void* src, size_t* srcSizePtr)
 {
+    DEBUGLOG(5, "ZBUFFv04_decompressContinue");
     return ZBUFF_decompressContinue(dctx, dst, maxDstSizePtr, src, srcSizePtr);
 }
 

src/borg/algorithms/zstd/lib/legacy/zstd_v05.c

@@ -736,18 +736,6 @@ MEM_STATIC BITv05_DStream_status BITv05_reloadDStream(BITv05_DStream_t* bitD);
 MEM_STATIC unsigned BITv05_endOfDStream(const BITv05_DStream_t* bitD);
 
 
-/*!
-* Start by invoking BITv05_initDStream().
-* A chunk of the bitStream is then stored into a local register.
-* Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
-* You can then retrieve bitFields stored into the local register, **in reverse order**.
-* Local register is explicitly reloaded from memory by the BITv05_reloadDStream() method.
-* A reload guarantee a minimum of ((8*sizeof(size_t))-7) bits when its result is BITv05_DStream_unfinished.
-* Otherwise, it can be less than that, so proceed accordingly.
-* Checking if DStream has reached its end can be performed with BITv05_endOfDStream()
-*/
-
-
 /*-****************************************
 *  unsafe API
 ******************************************/
@@ -759,7 +747,7 @@ MEM_STATIC size_t BITv05_readBitsFast(BITv05_DStream_t* bitD, unsigned nbBits);
 /*-**************************************************************
 *  Helper functions
 ****************************************************************/
-MEM_STATIC unsigned BITv05_highbit32 (register U32 val)
+MEM_STATIC unsigned BITv05_highbit32 (U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r=0;
@@ -829,13 +817,6 @@ MEM_STATIC size_t BITv05_initDStream(BITv05_DStream_t* bitD, const void* srcBuff
     return srcSize;
 }
 
-/*!BITv05_lookBits
- * Provides next n bits from local register
- * local register is not modified (bits are still present for next read/look)
- * On 32-bits, maxNbBits==25
- * On 64-bits, maxNbBits==57
- * @return : value extracted
- */
 MEM_STATIC size_t BITv05_lookBits(BITv05_DStream_t* bitD, U32 nbBits)
 {
     const U32 bitMask = sizeof(bitD->bitContainer)*8 - 1;
@@ -855,11 +836,6 @@ MEM_STATIC void BITv05_skipBits(BITv05_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-/*!BITv05_readBits
- * Read next n bits from local register.
- * pay attention to not read more than nbBits contained into local register.
- * @return : extracted value.
- */
 MEM_STATIC size_t BITv05_readBits(BITv05_DStream_t* bitD, U32 nbBits)
 {
     size_t value = BITv05_lookBits(bitD, nbBits);
@@ -995,54 +971,6 @@ static unsigned char FSEv05_decodeSymbol(FSEv05_DState_t* DStatePtr, BITv05_DStr
 
 static unsigned FSEv05_endOfDState(const FSEv05_DState_t* DStatePtr);
 
-/*!
-Let's now decompose FSEv05_decompress_usingDTable() into its unitary components.
-You will decode FSEv05-encoded symbols from the bitStream,
-and also any other bitFields you put in, **in reverse order**.
-
-You will need a few variables to track your bitStream. They are :
-
-BITv05_DStream_t DStream;    // Stream context
-FSEv05_DState_t  DState;     // State context. Multiple ones are possible
-FSEv05_DTable*   DTablePtr;  // Decoding table, provided by FSEv05_buildDTable()
-
-The first thing to do is to init the bitStream.
-    errorCode = BITv05_initDStream(&DStream, srcBuffer, srcSize);
-
-You should then retrieve your initial state(s)
-(in reverse flushing order if you have several ones) :
-    errorCode = FSEv05_initDState(&DState, &DStream, DTablePtr);
-
-You can then decode your data, symbol after symbol.
-For information the maximum number of bits read by FSEv05_decodeSymbol() is 'tableLog'.
-Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
-    unsigned char symbol = FSEv05_decodeSymbol(&DState, &DStream);
-
-You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
-Note : maximum allowed nbBits is 25, for 32-bits compatibility
-    size_t bitField = BITv05_readBits(&DStream, nbBits);
-
-All above operations only read from local register (which size depends on size_t).
-Refueling the register from memory is manually performed by the reload method.
-    endSignal = FSEv05_reloadDStream(&DStream);
-
-BITv05_reloadDStream() result tells if there is still some more data to read from DStream.
-BITv05_DStream_unfinished : there is still some data left into the DStream.
-BITv05_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
-BITv05_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
-BITv05_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
-
-When reaching end of buffer (BITv05_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
-to properly detect the exact end of stream.
-After each decoded symbol, check if DStream is fully consumed using this simple test :
-    BITv05_reloadDStream(&DStream) >= BITv05_DStream_completed
-
-When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
-Checking if DStream has reached its end is performed by :
-    BITv05_endOfDStream(&DStream);
-Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
-    FSEv05_endOfDState(&DState);
-*/
 
 
 /* *****************************************

src/borg/algorithms/zstd/lib/legacy/zstd_v06.c

@@ -189,7 +189,7 @@ MEM_STATIC U32 MEM_swap32(U32 in)
 {
 #if defined(_MSC_VER)     /* Visual Studio */
     return _byteswap_ulong(in);
-#elif defined (__GNUC__)
+#elif defined (__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
     return __builtin_bswap32(in);
 #else
     return  ((in << 24) & 0xff000000 ) |
@@ -203,7 +203,7 @@ MEM_STATIC U64 MEM_swap64(U64 in)
 {
 #if defined(_MSC_VER)     /* Visual Studio */
     return _byteswap_uint64(in);
-#elif defined (__GNUC__)
+#elif defined (__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
     return __builtin_bswap64(in);
 #else
     return  ((in << 56) & 0xff00000000000000ULL) |
@@ -839,16 +839,6 @@ MEM_STATIC BITv06_DStream_status BITv06_reloadDStream(BITv06_DStream_t* bitD);
 MEM_STATIC unsigned BITv06_endOfDStream(const BITv06_DStream_t* bitD);
 
 
-/* Start by invoking BITv06_initDStream().
-*  A chunk of the bitStream is then stored into a local register.
-*  Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
-*  You can then retrieve bitFields stored into the local register, **in reverse order**.
-*  Local register is explicitly reloaded from memory by the BITv06_reloadDStream() method.
-*  A reload guarantee a minimum of ((8*sizeof(bitD->bitContainer))-7) bits when its result is BITv06_DStream_unfinished.
-*  Otherwise, it can be less than that, so proceed accordingly.
-*  Checking if DStream has reached its end can be performed with BITv06_endOfDStream().
-*/
-
 
 /*-****************************************
 *  unsafe API
@@ -861,7 +851,7 @@ MEM_STATIC size_t BITv06_readBitsFast(BITv06_DStream_t* bitD, unsigned nbBits);
 /*-**************************************************************
 *  Internal functions
 ****************************************************************/
-MEM_STATIC unsigned BITv06_highbit32 (register U32 val)
+MEM_STATIC unsigned BITv06_highbit32 ( U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r=0;
@@ -929,13 +919,6 @@ MEM_STATIC size_t BITv06_initDStream(BITv06_DStream_t* bitD, const void* srcBuff
 }
 
 
-/*! BITv06_lookBits() :
- *  Provides next n bits from local register.
- *  local register is not modified.
- *  On 32-bits, maxNbBits==24.
- *  On 64-bits, maxNbBits==56.
- *  @return : value extracted
- */
  MEM_STATIC size_t BITv06_lookBits(const BITv06_DStream_t* bitD, U32 nbBits)
 {
     U32 const bitMask = sizeof(bitD->bitContainer)*8 - 1;
@@ -955,11 +938,6 @@ MEM_STATIC void BITv06_skipBits(BITv06_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-/*! BITv06_readBits() :
- *  Read (consume) next n bits from local register and update.
- *  Pay attention to not read more than nbBits contained into local register.
- *  @return : extracted value.
- */
 MEM_STATIC size_t BITv06_readBits(BITv06_DStream_t* bitD, U32 nbBits)
 {
     size_t const value = BITv06_lookBits(bitD, nbBits);
@@ -976,11 +954,6 @@ MEM_STATIC size_t BITv06_readBitsFast(BITv06_DStream_t* bitD, U32 nbBits)
     return value;
 }
 
-/*! BITv06_reloadDStream() :
-*   Refill `BITv06_DStream_t` from src buffer previously defined (see BITv06_initDStream() ).
-*   This function is safe, it guarantees it will not read beyond src buffer.
-*   @return : status of `BITv06_DStream_t` internal register.
-              if status == unfinished, internal register is filled with >= (sizeof(bitD->bitContainer)*8 - 7) bits */
 MEM_STATIC BITv06_DStream_status BITv06_reloadDStream(BITv06_DStream_t* bitD)
 {
     if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* should never happen */
@@ -1103,55 +1076,6 @@ static void     FSEv06_initDState(FSEv06_DState_t* DStatePtr, BITv06_DStream_t*
 
 static unsigned char FSEv06_decodeSymbol(FSEv06_DState_t* DStatePtr, BITv06_DStream_t* bitD);
 
-/*!
-Let's now decompose FSEv06_decompress_usingDTable() into its unitary components.
-You will decode FSE-encoded symbols from the bitStream,
-and also any other bitFields you put in, **in reverse order**.
-
-You will need a few variables to track your bitStream. They are :
-
-BITv06_DStream_t DStream;    // Stream context
-FSEv06_DState_t  DState;     // State context. Multiple ones are possible
-FSEv06_DTable*   DTablePtr;  // Decoding table, provided by FSEv06_buildDTable()
-
-The first thing to do is to init the bitStream.
-    errorCode = BITv06_initDStream(&DStream, srcBuffer, srcSize);
-
-You should then retrieve your initial state(s)
-(in reverse flushing order if you have several ones) :
-    errorCode = FSEv06_initDState(&DState, &DStream, DTablePtr);
-
-You can then decode your data, symbol after symbol.
-For information the maximum number of bits read by FSEv06_decodeSymbol() is 'tableLog'.
-Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
-    unsigned char symbol = FSEv06_decodeSymbol(&DState, &DStream);
-
-You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
-Note : maximum allowed nbBits is 25, for 32-bits compatibility
-    size_t bitField = BITv06_readBits(&DStream, nbBits);
-
-All above operations only read from local register (which size depends on size_t).
-Refueling the register from memory is manually performed by the reload method.
-    endSignal = FSEv06_reloadDStream(&DStream);
-
-BITv06_reloadDStream() result tells if there is still some more data to read from DStream.
-BITv06_DStream_unfinished : there is still some data left into the DStream.
-BITv06_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
-BITv06_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
-BITv06_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
-
-When reaching end of buffer (BITv06_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
-to properly detect the exact end of stream.
-After each decoded symbol, check if DStream is fully consumed using this simple test :
-    BITv06_reloadDStream(&DStream) >= BITv06_DStream_completed
-
-When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
-Checking if DStream has reached its end is performed by :
-    BITv06_endOfDStream(&DStream);
-Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
-    FSEv06_endOfDState(&DState);
-*/
-
 
 /* *****************************************
 *  FSE unsafe API

src/borg/algorithms/zstd/lib/legacy/zstd_v07.c

@@ -348,7 +348,7 @@ MEM_STATIC U32 MEM_swap32(U32 in)
 {
 #if defined(_MSC_VER)     /* Visual Studio */
     return _byteswap_ulong(in);
-#elif defined (__GNUC__)
+#elif defined (__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
     return __builtin_bswap32(in);
 #else
     return  ((in << 24) & 0xff000000 ) |
@@ -362,7 +362,7 @@ MEM_STATIC U64 MEM_swap64(U64 in)
 {
 #if defined(_MSC_VER)     /* Visual Studio */
     return _byteswap_uint64(in);
-#elif defined (__GNUC__)
+#elif defined (__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 403)
     return __builtin_bswap64(in);
 #else
     return  ((in << 56) & 0xff00000000000000ULL) |
@@ -511,16 +511,6 @@ MEM_STATIC BITv07_DStream_status BITv07_reloadDStream(BITv07_DStream_t* bitD);
 MEM_STATIC unsigned BITv07_endOfDStream(const BITv07_DStream_t* bitD);
 
 
-/* Start by invoking BITv07_initDStream().
-*  A chunk of the bitStream is then stored into a local register.
-*  Local register size is 64-bits on 64-bits systems, 32-bits on 32-bits systems (size_t).
-*  You can then retrieve bitFields stored into the local register, **in reverse order**.
-*  Local register is explicitly reloaded from memory by the BITv07_reloadDStream() method.
-*  A reload guarantee a minimum of ((8*sizeof(bitD->bitContainer))-7) bits when its result is BITv07_DStream_unfinished.
-*  Otherwise, it can be less than that, so proceed accordingly.
-*  Checking if DStream has reached its end can be performed with BITv07_endOfDStream().
-*/
-
 
 /*-****************************************
 *  unsafe API
@@ -533,7 +523,7 @@ MEM_STATIC size_t BITv07_readBitsFast(BITv07_DStream_t* bitD, unsigned nbBits);
 /*-**************************************************************
 *  Internal functions
 ****************************************************************/
-MEM_STATIC unsigned BITv07_highbit32 (register U32 val)
+MEM_STATIC unsigned BITv07_highbit32 (U32 val)
 {
 #   if defined(_MSC_VER)   /* Visual */
     unsigned long r=0;
@@ -599,13 +589,6 @@ MEM_STATIC size_t BITv07_initDStream(BITv07_DStream_t* bitD, const void* srcBuff
 }
 
 
-/*! BITv07_lookBits() :
- *  Provides next n bits from local register.
- *  local register is not modified.
- *  On 32-bits, maxNbBits==24.
- *  On 64-bits, maxNbBits==56.
- *  @return : value extracted
- */
  MEM_STATIC size_t BITv07_lookBits(const BITv07_DStream_t* bitD, U32 nbBits)
 {
     U32 const bitMask = sizeof(bitD->bitContainer)*8 - 1;
@@ -625,11 +608,6 @@ MEM_STATIC void BITv07_skipBits(BITv07_DStream_t* bitD, U32 nbBits)
     bitD->bitsConsumed += nbBits;
 }
 
-/*! BITv07_readBits() :
- *  Read (consume) next n bits from local register and update.
- *  Pay attention to not read more than nbBits contained into local register.
- *  @return : extracted value.
- */
 MEM_STATIC size_t BITv07_readBits(BITv07_DStream_t* bitD, U32 nbBits)
 {
     size_t const value = BITv07_lookBits(bitD, nbBits);
@@ -646,11 +624,6 @@ MEM_STATIC size_t BITv07_readBitsFast(BITv07_DStream_t* bitD, U32 nbBits)
     return value;
 }
 
-/*! BITv07_reloadDStream() :
-*   Refill `BITv07_DStream_t` from src buffer previously defined (see BITv07_initDStream() ).
-*   This function is safe, it guarantees it will not read beyond src buffer.
-*   @return : status of `BITv07_DStream_t` internal register.
-              if status == unfinished, internal register is filled with >= (sizeof(bitD->bitContainer)*8 - 7) bits */
 MEM_STATIC BITv07_DStream_status BITv07_reloadDStream(BITv07_DStream_t* bitD)
 {
     if (bitD->bitsConsumed > (sizeof(bitD->bitContainer)*8))  /* should not happen => corruption detected */
@@ -874,55 +847,6 @@ static void     FSEv07_initDState(FSEv07_DState_t* DStatePtr, BITv07_DStream_t*
 static unsigned char FSEv07_decodeSymbol(FSEv07_DState_t* DStatePtr, BITv07_DStream_t* bitD);
 
 
-/**<
-Let's now decompose FSEv07_decompress_usingDTable() into its unitary components.
-You will decode FSE-encoded symbols from the bitStream,
-and also any other bitFields you put in, **in reverse order**.
-
-You will need a few variables to track your bitStream. They are :
-
-BITv07_DStream_t DStream;    // Stream context
-FSEv07_DState_t  DState;     // State context. Multiple ones are possible
-FSEv07_DTable*   DTablePtr;  // Decoding table, provided by FSEv07_buildDTable()
-
-The first thing to do is to init the bitStream.
-    errorCode = BITv07_initDStream(&DStream, srcBuffer, srcSize);
-
-You should then retrieve your initial state(s)
-(in reverse flushing order if you have several ones) :
-    errorCode = FSEv07_initDState(&DState, &DStream, DTablePtr);
-
-You can then decode your data, symbol after symbol.
-For information the maximum number of bits read by FSEv07_decodeSymbol() is 'tableLog'.
-Keep in mind that symbols are decoded in reverse order, like a LIFO stack (last in, first out).
-    unsigned char symbol = FSEv07_decodeSymbol(&DState, &DStream);
-
-You can retrieve any bitfield you eventually stored into the bitStream (in reverse order)
-Note : maximum allowed nbBits is 25, for 32-bits compatibility
-    size_t bitField = BITv07_readBits(&DStream, nbBits);
-
-All above operations only read from local register (which size depends on size_t).
-Refueling the register from memory is manually performed by the reload method.
-    endSignal = FSEv07_reloadDStream(&DStream);
-
-BITv07_reloadDStream() result tells if there is still some more data to read from DStream.
-BITv07_DStream_unfinished : there is still some data left into the DStream.
-BITv07_DStream_endOfBuffer : Dstream reached end of buffer. Its container may no longer be completely filled.
-BITv07_DStream_completed : Dstream reached its exact end, corresponding in general to decompression completed.
-BITv07_DStream_tooFar : Dstream went too far. Decompression result is corrupted.
-
-When reaching end of buffer (BITv07_DStream_endOfBuffer), progress slowly, notably if you decode multiple symbols per loop,
-to properly detect the exact end of stream.
-After each decoded symbol, check if DStream is fully consumed using this simple test :
-    BITv07_reloadDStream(&DStream) >= BITv07_DStream_completed
-
-When it's done, verify decompression is fully completed, by checking both DStream and the relevant states.
-Checking if DStream has reached its end is performed by :
-    BITv07_endOfDStream(&DStream);
-Check also the states. There might be some symbols left there, if some high probability ones (>50%) are possible.
-    FSEv07_endOfDState(&DState);
-*/
-
 
 /* *****************************************
 *  FSE unsafe API

src/borg/algorithms/zstd/lib/zstd.h

@@ -45,11 +45,11 @@ extern "C" {
   Levels >= 20, labeled `--ultra`, should be used with caution, as they require more memory.
   Compression can be done in:
     - a single step (described as Simple API)
-    - a single step, reusing a context (described as Explicit memory management)
+    - a single step, reusing a context (described as Explicit context)
     - unbounded multiple steps (described as Streaming compression)
   The compression ratio achievable on small data can be highly improved using a dictionary in:
     - a single step (described as Simple dictionary API)
-    - a single step, reusing a dictionary (described as Fast dictionary API)
+    - a single step, reusing a dictionary (described as Bulk-processing dictionary API)
 
   Advanced experimental functions can be accessed using #define ZSTD_STATIC_LINKING_ONLY before including zstd.h.
   Advanced experimental APIs shall never be used with a dynamic library.
@@ -59,7 +59,7 @@ extern "C" {
 /*------   Version   ------*/
 #define ZSTD_VERSION_MAJOR    1
 #define ZSTD_VERSION_MINOR    3
-#define ZSTD_VERSION_RELEASE  2
+#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);   /**< useful to check dll version */
@@ -68,7 +68,7 @@ ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< useful to check dll versio
 #define ZSTD_QUOTE(str) #str
 #define ZSTD_EXPAND_AND_QUOTE(str) ZSTD_QUOTE(str)
 #define ZSTD_VERSION_STRING ZSTD_EXPAND_AND_QUOTE(ZSTD_LIB_VERSION)
-ZSTDLIB_API const char* ZSTD_versionString(void);   /* v1.3.0 */
+ZSTDLIB_API const char* ZSTD_versionString(void);   /* added in v1.3.0 */
 
 
 /***************************************
@@ -92,7 +92,7 @@ ZSTDLIB_API size_t ZSTD_compress( void* dst, size_t dstCapacity,
 ZSTDLIB_API size_t ZSTD_decompress( void* dst, size_t dstCapacity,
                               const void* src, size_t compressedSize);
 
-/*! ZSTD_getFrameContentSize() : v1.3.0
+/*! ZSTD_getFrameContentSize() : added in v1.3.0
  *  `src` should point to the start of a ZSTD encoded frame.
  *  `srcSize` must be at least as large as the frame header.
  *            hint : any size >= `ZSTD_frameHeaderSize_max` is large enough.
@@ -120,26 +120,24 @@ ZSTDLIB_API unsigned long long ZSTD_getFrameContentSize(const void *src, size_t
 
 /*! ZSTD_getDecompressedSize() :
  *  NOTE: This function is now obsolete, in favor of ZSTD_getFrameContentSize().
- *  Both functions work the same way,
- *  but ZSTD_getDecompressedSize() blends
- *  "empty", "unknown" and "error" results in the same return value (0),
- *  while ZSTD_getFrameContentSize() distinguishes them.
- *
- *  'src' is the start of a zstd compressed frame.
- *  @return : content size to be decompressed, as a 64-bits value _if known and not empty_, 0 otherwise. */
+ *  Both functions work the same way, but ZSTD_getDecompressedSize() blends
+ *  "empty", "unknown" and "error" results to the same return value (0),
+ *  while ZSTD_getFrameContentSize() gives them separate return values.
+ * `src` is the start of a zstd compressed frame.
+ * @return : content size to be decompressed, as a 64-bits value _if known and not empty_, 0 otherwise. */
 ZSTDLIB_API unsigned long long ZSTD_getDecompressedSize(const void* src, size_t srcSize);
 
 
 /*======  Helper functions  ======*/
-#define ZSTD_COMPRESSBOUND(srcSize)   ((srcSize) + ((srcSize)>>8) + (((srcSize) < 128 KB) ? ((128 KB - (srcSize)) >> 11) /* margin, from 64 to 0 */ : 0))  /* this formula ensures that bound(A) + bound(B) <= bound(A+B) as long as A and B >= 128 KB */
-ZSTDLIB_API size_t      ZSTD_compressBound(size_t srcSize); /*!< maximum compressed size in worst case scenario */
+#define ZSTD_COMPRESSBOUND(srcSize)   ((srcSize) + ((srcSize)>>8) + (((srcSize) < (128<<10)) ? (((128<<10) - (srcSize)) >> 11) /* margin, from 64 to 0 */ : 0))  /* this formula ensures that bound(A) + bound(B) <= bound(A+B) as long as A and B >= 128 KB */
+ZSTDLIB_API size_t      ZSTD_compressBound(size_t srcSize); /*!< maximum compressed size in worst case single-pass scenario */
 ZSTDLIB_API unsigned    ZSTD_isError(size_t code);          /*!< tells if a `size_t` function result is an error code */
 ZSTDLIB_API const char* ZSTD_getErrorName(size_t code);     /*!< provides readable string from an error code */
 ZSTDLIB_API int         ZSTD_maxCLevel(void);               /*!< maximum compression level available */
 
 
 /***************************************
-*  Explicit memory management
+*  Explicit context
 ***************************************/
 /*= Compression context
  *  When compressing many times,
@@ -345,7 +343,7 @@ ZSTDLIB_API size_t ZSTD_CStreamOutSize(void);   /**< recommended size for output
 * *******************************************************************************/
 
 typedef ZSTD_DCtx ZSTD_DStream;  /**< DCtx and DStream are now effectively same object (>= v1.3.0) */
-                                 /* Continue to distinguish them for compatibility with versions <= v1.2.0 */
+                                 /* For compatibility with versions <= v1.2.0, continue to consider them separated. */
 /*===== ZSTD_DStream management functions =====*/
 ZSTDLIB_API ZSTD_DStream* ZSTD_createDStream(void);
 ZSTDLIB_API size_t ZSTD_freeDStream(ZSTD_DStream* zds);
@@ -375,23 +373,24 @@ ZSTDLIB_API size_t ZSTD_DStreamOutSize(void);   /*!< recommended size for output
 /* --- Constants ---*/
 #define ZSTD_MAGICNUMBER            0xFD2FB528   /* >= v0.8.0 */
 #define ZSTD_MAGIC_SKIPPABLE_START  0x184D2A50U
-#define ZSTD_MAGIC_DICTIONARY       0xEC30A437   /* v0.7+ */
+#define ZSTD_MAGIC_DICTIONARY       0xEC30A437   /* >= v0.7.0 */
 
 #define ZSTD_WINDOWLOG_MAX_32   30
 #define ZSTD_WINDOWLOG_MAX_64   31
 #define ZSTD_WINDOWLOG_MAX    ((unsigned)(sizeof(size_t) == 4 ? ZSTD_WINDOWLOG_MAX_32 : ZSTD_WINDOWLOG_MAX_64))
 #define ZSTD_WINDOWLOG_MIN      10
-#define ZSTD_HASHLOG_MAX        MIN(ZSTD_WINDOWLOG_MAX, 30)
+#define ZSTD_HASHLOG_MAX      ((ZSTD_WINDOWLOG_MAX < 30) ? ZSTD_WINDOWLOG_MAX : 30)
 #define ZSTD_HASHLOG_MIN         6
-#define ZSTD_CHAINLOG_MAX       MIN(ZSTD_WINDOWLOG_MAX+1, 30)
+#define ZSTD_CHAINLOG_MAX_32    29
+#define ZSTD_CHAINLOG_MAX_64    30
+#define ZSTD_CHAINLOG_MAX     ((unsigned)(sizeof(size_t) == 4 ? ZSTD_CHAINLOG_MAX_32 : ZSTD_CHAINLOG_MAX_64))
 #define ZSTD_CHAINLOG_MIN       ZSTD_HASHLOG_MIN
 #define ZSTD_HASHLOG3_MAX       17
 #define ZSTD_SEARCHLOG_MAX     (ZSTD_WINDOWLOG_MAX-1)
 #define ZSTD_SEARCHLOG_MIN       1
 #define ZSTD_SEARCHLENGTH_MAX    7   /* only for ZSTD_fast, other strategies are limited to 6 */
 #define ZSTD_SEARCHLENGTH_MIN    3   /* only for ZSTD_btopt, other strategies are limited to 4 */
-#define ZSTD_TARGETLENGTH_MIN    4   /* only useful for btopt */
-#define ZSTD_TARGETLENGTH_MAX  999   /* only useful for btopt */
+#define ZSTD_TARGETLENGTH_MIN    1   /* only used by btopt, btultra and btfast */
 #define ZSTD_LDM_MINMATCH_MIN    4
 #define ZSTD_LDM_MINMATCH_MAX 4096
 #define ZSTD_LDM_BUCKETSIZELOG_MAX 8
@@ -432,12 +431,17 @@ typedef struct {
 
 typedef struct ZSTD_CCtx_params_s ZSTD_CCtx_params;
 
-/*= Custom memory allocation functions */
-typedef void* (*ZSTD_allocFunction) (void* opaque, size_t size);
-typedef void  (*ZSTD_freeFunction) (void* opaque, void* address);
-typedef struct { ZSTD_allocFunction customAlloc; ZSTD_freeFunction customFree; void* opaque; } ZSTD_customMem;
-/* use this constant to defer to stdlib's functions */
-static const ZSTD_customMem ZSTD_defaultCMem = { NULL, NULL, NULL };
+typedef enum {
+    ZSTD_dct_auto=0,      /* dictionary is "full" when starting with ZSTD_MAGIC_DICTIONARY, otherwise it is "rawContent" */
+    ZSTD_dct_rawContent,  /* ensures dictionary is always loaded as rawContent, even if it starts with ZSTD_MAGIC_DICTIONARY */
+    ZSTD_dct_fullDict     /* refuses to load a dictionary if it does not respect Zstandard's specification */
+} ZSTD_dictContentType_e;
+
+typedef enum {
+    ZSTD_dlm_byCopy = 0, /**< Copy dictionary content internally */
+    ZSTD_dlm_byRef,      /**< Reference dictionary content -- the dictionary buffer must outlive its users. */
+} ZSTD_dictLoadMethod_e;
+
 
 
 /***************************************
@@ -446,7 +450,7 @@ static const ZSTD_customMem ZSTD_defaultCMem = { NULL, NULL, NULL };
 
 /*! ZSTD_findFrameCompressedSize() :
  *  `src` should point to the start of a ZSTD encoded frame or skippable frame
- *  `srcSize` must be at least as large as the frame
+ *  `srcSize` must be >= first frame size
  *  @return : the compressed size of the first frame starting at `src`,
  *            suitable to pass to `ZSTD_decompress` or similar,
  *            or an error code if input is invalid */
@@ -483,12 +487,12 @@ ZSTDLIB_API size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
 
 
 /***************************************
-*  Context memory usage
+*  Memory management
 ***************************************/
 
 /*! ZSTD_sizeof_*() :
  *  These functions give the current memory usage of selected object.
- *  Object memory usage can evolve when re-used multiple times. */
+ *  Object memory usage can evolve when re-used. */
 ZSTDLIB_API size_t ZSTD_sizeof_CCtx(const ZSTD_CCtx* cctx);
 ZSTDLIB_API size_t ZSTD_sizeof_DCtx(const ZSTD_DCtx* dctx);
 ZSTDLIB_API size_t ZSTD_sizeof_CStream(const ZSTD_CStream* zcs);
@@ -503,8 +507,8 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
  *  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_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbThreads is > 1.
- *  Note : CCtx estimation is only correct for single-threaded compression */
+ *  ZSTD_estimateCCtxSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbWorkers is >= 1.
+ *  Note : CCtx size estimation is only correct for single-threaded compression. */
 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);
@@ -515,8 +519,8 @@ ZSTDLIB_API size_t ZSTD_estimateDCtxSize(void);
  *  It will also consider src size to be arbitrarily "large", which is worst case.
  *  If srcSize is known to always be small, ZSTD_estimateCStreamSize_usingCParams() can provide a tighter estimation.
  *  ZSTD_estimateCStreamSize_usingCParams() can be used in tandem with ZSTD_getCParams() to create cParams from compressionLevel.
- *  ZSTD_estimateCStreamSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbThreads is set to a value > 1.
- *  Note : CStream estimation is only correct for single-threaded compression.
+ *  ZSTD_estimateCStreamSize_usingCCtxParams() can be used in tandem with ZSTD_CCtxParam_setParameter(). Only single-threaded compression is supported. This function will return an error code if ZSTD_p_nbWorkers is >= 1.
+ *  Note : CStream size estimation is only correct for single-threaded compression.
  *  ZSTD_DStream memory budget depends on window Size.
  *  This information can be passed manually, using ZSTD_estimateDStreamSize,
  *  or deducted from a valid frame Header, using ZSTD_estimateDStreamSize_fromFrame();
@@ -529,82 +533,92 @@ ZSTDLIB_API size_t ZSTD_estimateCStreamSize_usingCCtxParams(const ZSTD_CCtx_para
 ZSTDLIB_API size_t ZSTD_estimateDStreamSize(size_t windowSize);
 ZSTDLIB_API size_t ZSTD_estimateDStreamSize_fromFrame(const void* src, size_t srcSize);
 
-typedef enum {
-    ZSTD_dlm_byCopy = 0,     /**< Copy dictionary content internally */
-    ZSTD_dlm_byRef,          /**< Reference dictionary content -- the dictionary buffer must outlive its users. */
-} ZSTD_dictLoadMethod_e;
-
 /*! ZSTD_estimate?DictSize() :
  *  ZSTD_estimateCDictSize() will bet that src size is relatively "small", and content is copied, like ZSTD_createCDict().
- *  ZSTD_estimateCStreamSize_advanced_usingCParams() makes it possible to control precisely compression parameters, like ZSTD_createCDict_advanced().
- *  Note : dictionary created by reference using ZSTD_dlm_byRef are smaller
+ *  ZSTD_estimateCDictSize_advanced() makes it possible to control compression parameters precisely, like ZSTD_createCDict_advanced().
+ *  Note : dictionaries created by reference (`ZSTD_dlm_byRef`) are logically smaller.
  */
 ZSTDLIB_API size_t ZSTD_estimateCDictSize(size_t dictSize, int compressionLevel);
 ZSTDLIB_API size_t ZSTD_estimateCDictSize_advanced(size_t dictSize, ZSTD_compressionParameters cParams, ZSTD_dictLoadMethod_e dictLoadMethod);
 ZSTDLIB_API size_t ZSTD_estimateDDictSize(size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod);
 
-
-/***************************************
-*  Advanced compression functions
-***************************************/
-/*! ZSTD_createCCtx_advanced() :
- *  Create a ZSTD compression context using external alloc and free functions */
-ZSTDLIB_API ZSTD_CCtx* ZSTD_createCCtx_advanced(ZSTD_customMem customMem);
-
-/*! ZSTD_initStaticCCtx() : initialize a fixed-size zstd compression context
- *  workspace: The memory area to emplace the context into.
- *             Provided pointer must 8-bytes aligned.
- *             It must outlive context usage.
- *  workspaceSize: Use ZSTD_estimateCCtxSize() or ZSTD_estimateCStreamSize()
- *                 to determine how large workspace must be to support scenario.
- * @return : pointer to ZSTD_CCtx*, or NULL if error (size too small)
- *  Note : zstd will never resize nor malloc() when using a static cctx.
- *         If it needs more memory than available, it will simply error out.
+/*! ZSTD_initStatic*() :
+ *  Initialize an object using a pre-allocated fixed-size buffer.
+ *  workspace: The memory area to emplace the object into.
+ *             Provided pointer *must be 8-bytes aligned*.
+ *             Buffer must outlive object.
+ *  workspaceSize: Use ZSTD_estimate*Size() to determine
+ *                 how large workspace must be to support target scenario.
+ * @return : pointer to object (same address as workspace, just different type),
+ *           or NULL if error (size too small, incorrect alignment, etc.)
+ *  Note : zstd will never resize nor malloc() when using a static buffer.
+ *         If the object requires more memory than available,
+ *         zstd will just error out (typically ZSTD_error_memory_allocation).
  *  Note 2 : there is no corresponding "free" function.
- *           Since workspace was allocated externally, it must be freed externally too.
- *  Limitation 1 : currently not compatible with internal CDict creation, such as
- *                 ZSTD_CCtx_loadDictionary() or ZSTD_initCStream_usingDict().
- *  Limitation 2 : currently not compatible with multi-threading
+ *           Since workspace is allocated externally, it must be freed externally too.
+ *  Note 3 : cParams : use ZSTD_getCParams() to convert a compression level
+ *           into its associated cParams.
+ *  Limitation 1 : currently not compatible with internal dictionary creation, triggered by
+ *                 ZSTD_CCtx_loadDictionary(), ZSTD_initCStream_usingDict() or ZSTD_initDStream_usingDict().
+ *  Limitation 2 : static cctx currently not compatible with multi-threading.
+ *  Limitation 3 : static dctx is incompatible with legacy support.
  */
-ZSTDLIB_API ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize);
+ZSTDLIB_API ZSTD_CCtx*    ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize);
+ZSTDLIB_API ZSTD_CStream* ZSTD_initStaticCStream(void* workspace, size_t workspaceSize);    /**< same as ZSTD_initStaticCCtx() */
 
+ZSTDLIB_API ZSTD_DCtx*    ZSTD_initStaticDCtx(void* workspace, size_t workspaceSize);
+ZSTDLIB_API ZSTD_DStream* ZSTD_initStaticDStream(void* workspace, size_t workspaceSize);    /**< same as ZSTD_initStaticDCtx() */
 
-/*! ZSTD_createCDict_byReference() :
- *  Create a digested dictionary for compression
- *  Dictionary content is simply referenced, and therefore stays in dictBuffer.
- *  It is important that dictBuffer outlives CDict, it must remain read accessible throughout the lifetime of CDict */
-ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
+ZSTDLIB_API const ZSTD_CDict* ZSTD_initStaticCDict(
+                                        void* workspace, size_t workspaceSize,
+                                        const void* dict, size_t dictSize,
+                                        ZSTD_dictLoadMethod_e dictLoadMethod,
+                                        ZSTD_dictContentType_e dictContentType,
+                                        ZSTD_compressionParameters cParams);
+
+ZSTDLIB_API const ZSTD_DDict* ZSTD_initStaticDDict(
+                                        void* workspace, size_t workspaceSize,
+                                        const void* dict, size_t dictSize,
+                                        ZSTD_dictLoadMethod_e dictLoadMethod,
+                                        ZSTD_dictContentType_e dictContentType);
+
+/*! Custom memory allocation :
+ *  These prototypes make it possible to pass your own allocation/free functions.
+ *  ZSTD_customMem is provided at creation time, using ZSTD_create*_advanced() variants listed below.
+ *  All allocation/free operations will be completed using these custom variants instead of regular <stdlib.h> ones.
+ */
+typedef void* (*ZSTD_allocFunction) (void* opaque, size_t size);
+typedef void  (*ZSTD_freeFunction) (void* opaque, void* address);
+typedef struct { ZSTD_allocFunction customAlloc; ZSTD_freeFunction customFree; void* opaque; } ZSTD_customMem;
+static ZSTD_customMem const ZSTD_defaultCMem = { NULL, NULL, NULL };  /**< this constant defers to stdlib's functions */
+
+ZSTDLIB_API ZSTD_CCtx*    ZSTD_createCCtx_advanced(ZSTD_customMem customMem);
+ZSTDLIB_API ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem);
+ZSTDLIB_API ZSTD_DCtx*    ZSTD_createDCtx_advanced(ZSTD_customMem customMem);
+ZSTDLIB_API ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem);
 
-typedef enum { ZSTD_dm_auto=0,        /* dictionary is "full" if it starts with ZSTD_MAGIC_DICTIONARY, otherwise it is "rawContent" */
-               ZSTD_dm_rawContent,    /* ensures dictionary is always loaded as rawContent, even if it starts with ZSTD_MAGIC_DICTIONARY */
-               ZSTD_dm_fullDict       /* refuses to load a dictionary if it does not respect Zstandard's specification */
-} ZSTD_dictMode_e;
-/*! ZSTD_createCDict_advanced() :
- *  Create a ZSTD_CDict using external alloc and free, and customized compression parameters */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize,
                                                   ZSTD_dictLoadMethod_e dictLoadMethod,
-                                                  ZSTD_dictMode_e dictMode,
+                                                  ZSTD_dictContentType_e dictContentType,
                                                   ZSTD_compressionParameters cParams,
                                                   ZSTD_customMem customMem);
 
-/*! ZSTD_initStaticCDict_advanced() :
- *  Generate a digested dictionary in provided memory area.
- *  workspace: The memory area to emplace the dictionary into.
- *             Provided pointer must 8-bytes aligned.
- *             It must outlive dictionary usage.
- *  workspaceSize: Use ZSTD_estimateCDictSize()
- *                 to determine how large workspace must be.
- *  cParams : use ZSTD_getCParams() to transform a compression level
- *            into its relevants cParams.
- * @return : pointer to ZSTD_CDict*, or NULL if error (size too small)
- *  Note : there is no corresponding "free" function.
- *         Since workspace was allocated externally, it must be freed externally.
- */
-ZSTDLIB_API ZSTD_CDict* ZSTD_initStaticCDict(
-                            void* workspace, size_t workspaceSize,
-                      const void* dict, size_t dictSize,
-                            ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode,
-                            ZSTD_compressionParameters cParams);
+ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
+                                                  ZSTD_dictLoadMethod_e dictLoadMethod,
+                                                  ZSTD_dictContentType_e dictContentType,
+                                                  ZSTD_customMem customMem);
+
+
+
+/***************************************
+*  Advanced compression functions
+***************************************/
+
+/*! ZSTD_createCDict_byReference() :
+ *  Create a digested dictionary for compression
+ *  Dictionary content is simply referenced, and therefore stays in dictBuffer.
+ *  It is important that dictBuffer outlives CDict, it must remain read accessible throughout the lifetime of CDict */
+ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
 
 /*! ZSTD_getCParams() :
 *   @return ZSTD_compressionParameters structure for a selected compression level and estimated srcSize.
@@ -613,7 +627,7 @@ ZSTDLIB_API ZSTD_compressionParameters ZSTD_getCParams(int compressionLevel, uns
 
 /*! ZSTD_getParams() :
 *   same as ZSTD_getCParams(), but @return a full `ZSTD_parameters` object instead of sub-component `ZSTD_compressionParameters`.
-*   All fields of `ZSTD_frameParameters` are set to default (0) */
+*   All fields of `ZSTD_frameParameters` are set to default : contentSize=1, checksum=0, noDictID=0 */
 ZSTDLIB_API ZSTD_parameters ZSTD_getParams(int compressionLevel, unsigned long long estimatedSrcSize, size_t dictSize);
 
 /*! ZSTD_checkCParams() :
@@ -650,27 +664,6 @@ ZSTDLIB_API size_t ZSTD_compress_usingCDict_advanced(ZSTD_CCtx* cctx,
  *  Note 3 : Skippable Frame Identifiers are considered valid. */
 ZSTDLIB_API unsigned ZSTD_isFrame(const void* buffer, size_t size);
 
-/*! ZSTD_createDCtx_advanced() :
- *  Create a ZSTD decompression context using external alloc and free functions */
-ZSTDLIB_API ZSTD_DCtx* ZSTD_createDCtx_advanced(ZSTD_customMem customMem);
-
-/*! ZSTD_initStaticDCtx() : initialize a fixed-size zstd decompression context
- *  workspace: The memory area to emplace the context into.
- *             Provided pointer must 8-bytes aligned.
- *             It must outlive context usage.
- *  workspaceSize: Use ZSTD_estimateDCtxSize() or ZSTD_estimateDStreamSize()
- *                 to determine how large workspace must be to support scenario.
- * @return : pointer to ZSTD_DCtx*, or NULL if error (size too small)
- *  Note : zstd will never resize nor malloc() when using a static dctx.
- *         If it needs more memory than available, it will simply error out.
- *  Note 2 : static dctx is incompatible with legacy support
- *  Note 3 : there is no corresponding "free" function.
- *           Since workspace was allocated externally, it must be freed externally.
- *  Limitation : currently not compatible with internal DDict creation,
- *               such as ZSTD_initDStream_usingDict().
- */
-ZSTDLIB_API ZSTD_DCtx* ZSTD_initStaticDCtx(void* workspace, size_t workspaceSize);
-
 /*! ZSTD_createDDict_byReference() :
  *  Create a digested dictionary, ready to start decompression operation without startup delay.
  *  Dictionary content is referenced, and therefore stays in dictBuffer.
@@ -678,26 +671,6 @@ ZSTDLIB_API ZSTD_DCtx* ZSTD_initStaticDCtx(void* workspace, size_t workspaceSize
  *  it must remain read accessible throughout the lifetime of DDict */
 ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_byReference(const void* dictBuffer, size_t dictSize);
 
-/*! ZSTD_createDDict_advanced() :
- *  Create a ZSTD_DDict using external alloc and free, optionally by reference */
-ZSTDLIB_API ZSTD_DDict* ZSTD_createDDict_advanced(const void* dict, size_t dictSize,
-                                                  ZSTD_dictLoadMethod_e dictLoadMethod,
-                                                  ZSTD_customMem customMem);
-
-/*! ZSTD_initStaticDDict() :
- *  Generate a digested dictionary in provided memory area.
- *  workspace: The memory area to emplace the dictionary into.
- *             Provided pointer must 8-bytes aligned.
- *             It must outlive dictionary usage.
- *  workspaceSize: Use ZSTD_estimateDDictSize()
- *                 to determine how large workspace must be.
- * @return : pointer to ZSTD_DDict*, or NULL if error (size too small)
- *  Note : there is no corresponding "free" function.
- *         Since workspace was allocated externally, it must be freed externally.
- */
-ZSTDLIB_API ZSTD_DDict* ZSTD_initStaticDDict(void* workspace, size_t workspaceSize,
-                                             const void* dict, size_t dictSize,
-                                             ZSTD_dictLoadMethod_e dictLoadMethod);
 
 /*! ZSTD_getDictID_fromDict() :
  *  Provides the dictID stored within dictionary.
@@ -729,28 +702,42 @@ ZSTDLIB_API unsigned ZSTD_getDictID_fromFrame(const void* src, size_t srcSize);
 ********************************************************************/
 
 /*=====   Advanced Streaming compression functions  =====*/
-ZSTDLIB_API ZSTD_CStream* ZSTD_createCStream_advanced(ZSTD_customMem customMem);
-ZSTDLIB_API ZSTD_CStream* ZSTD_initStaticCStream(void* workspace, size_t workspaceSize);    /**< same as ZSTD_initStaticCCtx() */
-ZSTDLIB_API size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize);   /**< pledgedSrcSize must be correct, a size of 0 means unknown.  for a frame size of 0 use initCStream_advanced */
+ZSTDLIB_API size_t ZSTD_initCStream_srcSize(ZSTD_CStream* zcs, int compressionLevel, unsigned long long pledgedSrcSize);   /**< 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. */
 ZSTDLIB_API size_t ZSTD_initCStream_usingDict(ZSTD_CStream* zcs, const void* dict, size_t dictSize, int compressionLevel); /**< creates of an internal CDict (incompatible with static CCtx), except if dict == NULL or dictSize < 8, in which case no dict is used. Note: dict is loaded with ZSTD_dm_auto (treated as a full zstd dictionary if it begins with ZSTD_MAGIC_DICTIONARY, else as raw content) and ZSTD_dlm_byCopy.*/
 ZSTDLIB_API size_t ZSTD_initCStream_advanced(ZSTD_CStream* zcs, const void* dict, size_t dictSize,
-                                             ZSTD_parameters params, unsigned long long pledgedSrcSize);  /**< pledgedSrcSize is optional and can be 0 (meaning unknown). note: if the contentSizeFlag is set, pledgedSrcSize == 0 means the source size is actually 0. dict is loaded with ZSTD_dm_auto and ZSTD_dlm_byCopy. */
+                                             ZSTD_parameters params, unsigned long long pledgedSrcSize);  /**< 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. */
 ZSTDLIB_API size_t ZSTD_initCStream_usingCDict(ZSTD_CStream* zcs, const ZSTD_CDict* cdict);  /**< note : cdict will just be referenced, and must outlive compression session */
-ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize);  /**< same as ZSTD_initCStream_usingCDict(), with control over frame parameters */
+ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const ZSTD_CDict* cdict, ZSTD_frameParameters fParams, unsigned long long pledgedSrcSize);  /**< same as ZSTD_initCStream_usingCDict(), with control over frame parameters. pledgedSrcSize must be correct. If srcSize is not known at init time, use value ZSTD_CONTENTSIZE_UNKNOWN. */
 
 /*! ZSTD_resetCStream() :
  *  start a new compression job, using same parameters from previous job.
  *  This is typically useful to skip dictionary loading stage, since it will re-use it in-place..
  *  Note that zcs must be init at least once before using ZSTD_resetCStream().
- *  pledgedSrcSize==0 means "srcSize unknown".
+ *  If pledgedSrcSize is not known at reset time, use macro ZSTD_CONTENTSIZE_UNKNOWN.
  *  If pledgedSrcSize > 0, its value must be correct, as it will be written in header, and controlled at the end.
- *  @return : 0, or an error code (which can be tested using ZSTD_isError()) */
+ *  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()) */
 ZSTDLIB_API size_t ZSTD_resetCStream(ZSTD_CStream* zcs, unsigned long long pledgedSrcSize);
 
 
+typedef struct {
+    unsigned long long ingested;
+    unsigned long long consumed;
+    unsigned long long produced;
+} ZSTD_frameProgression;
+
+/* ZSTD_getFrameProgression():
+ * tells how much data has been ingested (read from input)
+ * consumed (input actually compressed) and produced (output) for current frame.
+ * Therefore, (ingested - consumed) is amount of input data buffered internally, not yet compressed.
+ * Can report progression inside worker threads (multi-threading and non-blocking mode).
+ */
+ZSTD_frameProgression ZSTD_getFrameProgression(const ZSTD_CCtx* cctx);
+
+
+
 /*=====   Advanced Streaming decompression functions  =====*/
-ZSTDLIB_API ZSTD_DStream* ZSTD_createDStream_advanced(ZSTD_customMem customMem);
-ZSTDLIB_API ZSTD_DStream* ZSTD_initStaticDStream(void* workspace, size_t workspaceSize);    /**< same as ZSTD_initStaticDCtx() */
 typedef enum { DStream_p_maxWindowSize } ZSTD_DStreamParameter_e;
 ZSTDLIB_API size_t ZSTD_setDStreamParameter(ZSTD_DStream* zds, ZSTD_DStreamParameter_e paramType, unsigned paramValue);   /* obsolete : this API will be removed in a future version */
 ZSTDLIB_API size_t ZSTD_initDStream_usingDict(ZSTD_DStream* zds, const void* dict, size_t dictSize); /**< note: no dictionary will be used if dict == NULL or dictSize < 8 */
@@ -800,10 +787,10 @@ ZSTDLIB_API size_t ZSTD_resetDStream(ZSTD_DStream* zds);  /**< re-use decompress
 /*=====   Buffer-less streaming compression functions  =====*/
 ZSTDLIB_API size_t ZSTD_compressBegin(ZSTD_CCtx* cctx, int compressionLevel);
 ZSTDLIB_API size_t ZSTD_compressBegin_usingDict(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, int compressionLevel);
-ZSTDLIB_API size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_parameters params, unsigned long long pledgedSrcSize); /**< pledgedSrcSize is optional and can be 0 (meaning unknown). note: if the contentSizeFlag is set, pledgedSrcSize == 0 means the source size is actually 0 */
+ZSTDLIB_API size_t ZSTD_compressBegin_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_parameters params, unsigned long long pledgedSrcSize); /**< pledgedSrcSize : If srcSize is not known at init time, use ZSTD_CONTENTSIZE_UNKNOWN */
 ZSTDLIB_API size_t ZSTD_compressBegin_usingCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict); /**< note: fails if cdict==NULL */
-ZSTDLIB_API size_t ZSTD_compressBegin_usingCDict_advanced(ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize);   /* compression parameters are already set within cdict. pledgedSrcSize=0 means null-size */
-ZSTDLIB_API size_t ZSTD_copyCCtx(ZSTD_CCtx* cctx, const ZSTD_CCtx* preparedCCtx, unsigned long long pledgedSrcSize); /**<  note: if pledgedSrcSize can be 0, indicating unknown size.  if it is non-zero, it must be accurate.  for 0 size frames, use compressBegin_advanced */
+ZSTDLIB_API size_t ZSTD_compressBegin_usingCDict_advanced(ZSTD_CCtx* const cctx, const ZSTD_CDict* const cdict, ZSTD_frameParameters const fParams, unsigned long long const pledgedSrcSize);   /* compression parameters are already set within cdict. pledgedSrcSize must be correct. If srcSize is not known, use macro ZSTD_CONTENTSIZE_UNKNOWN */
+ZSTDLIB_API size_t ZSTD_copyCCtx(ZSTD_CCtx* cctx, const ZSTD_CCtx* preparedCCtx, unsigned long long pledgedSrcSize); /**<  note: if pledgedSrcSize is not known, use ZSTD_CONTENTSIZE_UNKNOWN */
 
 ZSTDLIB_API size_t ZSTD_compressContinue(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
 ZSTDLIB_API size_t ZSTD_compressEnd(ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
@@ -919,10 +906,8 @@ ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
  *   and then applied on all subsequent compression jobs.
  *   When no parameter is ever provided, CCtx is created with compression level ZSTD_CLEVEL_DEFAULT.
  *
- *   This API is intended to replace all others experimental API.
- *   It can basically do all other use cases, and even new ones.
- *   In constrast with _advanced() variants, it stands a reasonable chance to become "stable",
- *   after a good testing period.
+ *   This API is intended to replace all others advanced / experimental API entry points.
+ *   But it stands a reasonable chance to become "stable", after a reasonable testing period.
  */
 
 /* note on naming convention :
@@ -939,12 +924,12 @@ ZSTDLIB_API ZSTD_nextInputType_e ZSTD_nextInputType(ZSTD_DCtx* dctx);
  * All enum will be pinned to explicit values before reaching "stable API" status */
 
 typedef enum {
-    /* Question : should we have a format ZSTD_f_auto ?
-     * For the time being, it would mean exactly the same as ZSTD_f_zstd1.
-     * But, in the future, should several formats be supported,
+    /* 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 "multi format",
-     * and ZSTD_f_zstd1 could be reserved to mean "accept *only* zstd frames".
+     * 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 */
@@ -962,102 +947,125 @@ typedef enum {
     /* compression parameters */
     ZSTD_p_compressionLevel=100, /* Update all compression parameters according to pre-defined cLevel table
                               * Default level is ZSTD_CLEVEL_DEFAULT==3.
-                              * Special: value 0 means "do not change cLevel". */
+                              * Special: value 0 means "do not change cLevel".
+                              * Note 1 : it's possible to pass a negative compression level by casting it to unsigned type.
+                              * Note 2 : setting a level sets all default values of other compression parameters.
+                              * Note 3 : setting compressionLevel automatically updates ZSTD_p_compressLiterals. */
     ZSTD_p_windowLog,        /* Maximum allowed back-reference distance, expressed as power of 2.
                               * Must be clamped between ZSTD_WINDOWLOG_MIN and ZSTD_WINDOWLOG_MAX.
-                              * Special: value 0 means "do not change windowLog".
+                              * Special: value 0 means "use default windowLog".
                               * Note: Using a window size greater than ZSTD_MAXWINDOWSIZE_DEFAULT (default: 2^27)
-                              * requires setting the maximum window size at least as large during decompression. */
+                              *       requires explicitly allowing such window size during decompression stage. */
     ZSTD_p_hashLog,          /* Size of the probe table, as a power of 2.
                               * Resulting table size is (1 << (hashLog+2)).
                               * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX.
                               * Larger tables improve compression ratio of strategies <= dFast,
                               * and improve speed of strategies > dFast.
-                              * Special: value 0 means "do not change hashLog". */
+                              * Special: value 0 means "use default hashLog". */
     ZSTD_p_chainLog,         /* Size of the full-search table, as a power of 2.
                               * Resulting table size is (1 << (chainLog+2)).
                               * Larger tables result in better and slower compression.
                               * This parameter is useless when using "fast" strategy.
-                              * Special: value 0 means "do not change chainLog". */
+                              * Special: value 0 means "use default chainLog". */
     ZSTD_p_searchLog,        /* Number of search attempts, as a power of 2.
                               * More attempts result in better and slower compression.
                               * This parameter is useless when using "fast" and "dFast" strategies.
-                              * Special: value 0 means "do not change searchLog". */
+                              * Special: value 0 means "use default searchLog". */
     ZSTD_p_minMatch,         /* Minimum size of searched matches (note : repCode matches can be smaller).
                               * Larger values make faster compression and decompression, but decrease ratio.
                               * Must be clamped between ZSTD_SEARCHLENGTH_MIN and ZSTD_SEARCHLENGTH_MAX.
                               * Note that currently, for all strategies < btopt, effective minimum is 4.
-                              * Note that currently, for all strategies > fast, effective maximum is 6.
-                              * Special: value 0 means "do not change minMatchLength". */
-    ZSTD_p_targetLength,     /* Only useful for strategies >= btopt.
-                              * Length of Match considered "good enough" to stop search.
-                              * Larger values make compression stronger and slower.
-                              * Special: value 0 means "do not change targetLength". */
+                              *                    , for all strategies > fast, effective maximum is 6.
+                              * Special: value 0 means "use default minMatchLength". */
+    ZSTD_p_targetLength,     /* Impact of this field depends on strategy.
+                              * For strategies btopt & btultra:
+                              *     Length of Match considered "good enough" to stop search.
+                              *     Larger values make compression stronger, and slower.
+                              * For strategy fast:
+                              *     Distance between match sampling.
+                              *     Larger values make compression faster, and weaker.
+                              * Special: value 0 means "use default targetLength". */
     ZSTD_p_compressionStrategy, /* See ZSTD_strategy enum definition.
                               * Cast selected strategy as unsigned for ZSTD_CCtx_setParameter() compatibility.
                               * The higher the value of selected strategy, the more complex it is,
                               * resulting in stronger and slower compression.
-                              * Special: value 0 means "do not change strategy". */
+                              * Special: value 0 means "use default strategy". */
+
+    ZSTD_p_enableLongDistanceMatching=160, /* Enable long distance matching.
+                                         * This parameter is designed to improve compression ratio
+                                         * for large inputs, by finding large matches at long distance.
+                                         * It increases memory usage and window size.
+                                         * Note: enabling this parameter increases ZSTD_p_windowLog to 128 MB
+                                         * except when expressly set to a different value. */
+    ZSTD_p_ldmHashLog,       /* Size of the table for long distance matching, as a power of 2.
+                              * Larger values increase memory usage and compression ratio,
+                              * but decrease compression speed.
+                              * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
+                              * default: windowlog - 7.
+                              * Special: value 0 means "automatically determine hashlog". */
+    ZSTD_p_ldmMinMatch,      /* Minimum match size for long distance matcher.
+                              * Larger/too small values usually decrease compression ratio.
+                              * Must be clamped between ZSTD_LDM_MINMATCH_MIN and ZSTD_LDM_MINMATCH_MAX.
+                              * Special: value 0 means "use default value" (default: 64). */
+    ZSTD_p_ldmBucketSizeLog, /* Log size of each bucket in the LDM hash table for collision resolution.
+                              * Larger values improve collision resolution but decrease compression speed.
+                              * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX .
+                              * Special: value 0 means "use default value" (default: 3). */
+    ZSTD_p_ldmHashEveryLog,  /* Frequency of inserting/looking up entries in the LDM hash table.
+                              * Must be clamped between 0 and (ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN).
+                              * Default is MAX(0, (windowLog - ldmHashLog)), optimizing hash table usage.
+                              * Larger values improve compression speed.
+                              * Deviating far from default value will likely result in a compression ratio decrease.
+                              * Special: value 0 means "automatically determine hashEveryLog". */
 
     /* frame parameters */
-    ZSTD_p_contentSizeFlag=200, /* Content size is written into frame header _whenever known_ (default:1)
-                              * note that content size must be known at the beginning,
-                              * it is sent using ZSTD_CCtx_setPledgedSrcSize() */
+    ZSTD_p_contentSizeFlag=200, /* Content size will be written into frame header _whenever known_ (default:1)
+                              * Content size must be known at the beginning of compression,
+                              * it is provided using ZSTD_CCtx_setPledgedSrcSize() */
     ZSTD_p_checksumFlag,     /* A 32-bits checksum of content is written at end of frame (default:0) */
-    ZSTD_p_dictIDFlag,       /* When applicable, dictID of dictionary is provided in frame header (default:1) */
+    ZSTD_p_dictIDFlag,       /* When applicable, dictionary's ID is written into frame header (default:1) */
 
     /* multi-threading parameters */
-    ZSTD_p_nbThreads=400,    /* Select how many threads a compression job can spawn (default:1)
-                              * More threads improve speed, but also increase memory usage.
-                              * Can only receive a value > 1 if ZSTD_MULTITHREAD is enabled.
-                              * Special: value 0 means "do not change nbThreads" */
-    ZSTD_p_jobSize,          /* Size of a compression job. Each compression job is completed in parallel.
+    /* These parameters are only useful if multi-threading is enabled (ZSTD_MULTITHREAD).
+     * They return an error otherwise. */
+    ZSTD_p_nbWorkers=400,    /* Select how many threads will be spawned to compress in parallel.
+                              * When nbWorkers >= 1, triggers asynchronous mode :
+                              * ZSTD_compress_generic() consumes some input, flush some output if possible, and immediately gives back control to caller,
+                              * while compression work is performed in parallel, within worker threads.
+                              * (note : a strong exception to this rule is when first invocation sets ZSTD_e_end : it becomes a blocking call).
+                              * More workers improve speed, but also increase memory usage.
+                              * Default value is `0`, aka "single-threaded mode" : no worker is spawned, compression is performed inside Caller's thread, all invocations are blocking */
+    ZSTD_p_jobSize,          /* Size of a compression job. This value is enforced only in non-blocking mode.
+                              * Each compression job is completed in parallel, so this value indirectly controls the nb of active threads.
                               * 0 means default, which is dynamically determined based on compression parameters.
-                              * Job size must be a minimum of overlapSize, or 1 KB, whichever is largest
+                              * Job size must be a minimum of overlapSize, or 1 MB, whichever is largest.
                               * The minimum size is automatically and transparently enforced */
     ZSTD_p_overlapSizeLog,   /* Size of previous input reloaded at the beginning of each job.
                               * 0 => no overlap, 6(default) => use 1/8th of windowSize, >=9 => use full windowSize */
 
-    /* advanced parameters - may not remain available after API update */
+    /* =================================================================== */
+    /* experimental parameters - no stability guaranteed                   */
+    /* =================================================================== */
+
+    ZSTD_p_compressLiterals=1000, /* control huffman compression of literals (enabled) by default.
+                              * disabling it improves speed and decreases compression ratio by a large amount.
+                              * note : this setting is automatically updated when changing compression level.
+                              *        positive compression levels set ZSTD_p_compressLiterals to 1.
+                              *        negative compression levels set ZSTD_p_compressLiterals to 0. */
+
     ZSTD_p_forceMaxWindow=1100, /* Force back-reference distances to remain < windowSize,
                               * even when referencing into Dictionary content (default:0) */
-    ZSTD_p_enableLongDistanceMatching=1200,  /* Enable long distance matching.
-                                         * This parameter is designed to improve the compression
-                                         * ratio for large inputs with long distance matches.
-                                         * This increases the memory usage as well as window size.
-                                         * Note: setting this parameter sets all the LDM parameters
-                                         * as well as ZSTD_p_windowLog. It should be set after
-                                         * ZSTD_p_compressionLevel and before ZSTD_p_windowLog and
-                                         * other LDM parameters. Setting the compression level
-                                         * after this parameter overrides the window log, though LDM
-                                         * will remain enabled until explicitly disabled. */
-    ZSTD_p_ldmHashLog,   /* Size of the table for long distance matching, as a power of 2.
-                          * Larger values increase memory usage and compression ratio, but decrease
-                          * compression speed.
-                          * Must be clamped between ZSTD_HASHLOG_MIN and ZSTD_HASHLOG_MAX
-                          * (default: windowlog - 7). */
-    ZSTD_p_ldmMinMatch,  /* Minimum size of searched matches for long distance matcher.
-                          * Larger/too small values usually decrease compression ratio.
-                          * Must be clamped between ZSTD_LDM_MINMATCH_MIN
-                          * and ZSTD_LDM_MINMATCH_MAX (default: 64). */
-    ZSTD_p_ldmBucketSizeLog,  /* Log size of each bucket in the LDM hash table for collision resolution.
-                               * Larger values usually improve collision resolution but may decrease
-                               * compression speed.
-                               * The maximum value is ZSTD_LDM_BUCKETSIZELOG_MAX (default: 3). */
-    ZSTD_p_ldmHashEveryLog,  /* Frequency of inserting/looking up entries in the LDM hash table.
-                              * The default is MAX(0, (windowLog - ldmHashLog)) to
-                              * optimize hash table usage.
-                              * Larger values improve compression speed. Deviating far from the
-                              * default value will likely result in a decrease in compression ratio.
-                              * Must be clamped between 0 and ZSTD_WINDOWLOG_MAX - ZSTD_HASHLOG_MIN. */
 
 } ZSTD_cParameter;
 
 
 /*! ZSTD_CCtx_setParameter() :
  *  Set one compression parameter, selected by enum ZSTD_cParameter.
+ *  Setting a parameter is generally only possible during frame initialization (before starting compression),
+ *  except for a few exceptions which can be updated during compression: compressionLevel, hashLog, chainLog, searchLog, minMatch, targetLength and strategy.
  *  Note : when `value` is an enum, cast it to unsigned for proper type checking.
- *  @result : 0, or an error code (which can be tested with ZSTD_isError()). */
+ *  @result : informational value (typically, value being set clamped correctly),
+ *            or an error code (which can be tested with ZSTD_isError()). */
 ZSTDLIB_API size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param, unsigned value);
 
 /*! ZSTD_CCtx_setPledgedSrcSize() :
@@ -1066,32 +1074,30 @@ ZSTDLIB_API size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param
  * @result : 0, or an error code (which can be tested with ZSTD_isError()).
  *  Note 1 : 0 means zero, empty.
  *           In order to mean "unknown content size", pass constant ZSTD_CONTENTSIZE_UNKNOWN.
- *           Note that ZSTD_CONTENTSIZE_UNKNOWN is default value for new compression jobs.
+ *           ZSTD_CONTENTSIZE_UNKNOWN is default value for any new compression job.
  *  Note 2 : If all data is provided and consumed in a single round,
  *           this value is overriden by srcSize instead. */
 ZSTDLIB_API size_t ZSTD_CCtx_setPledgedSrcSize(ZSTD_CCtx* cctx, unsigned long long pledgedSrcSize);
 
 /*! ZSTD_CCtx_loadDictionary() :
- *  Create an internal CDict from dict buffer.
- *  Decompression will have to use same buffer.
+ *  Create an internal CDict from `dict` buffer.
+ *  Decompression will have to use same dictionary.
  * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Special : Adding a NULL (or 0-size) dictionary invalidates any previous dictionary,
- *            meaning "return to no-dictionary mode".
- *  Note 1 : `dict` content will be copied internally. Use
- *            ZSTD_CCtx_loadDictionary_byReference() to reference dictionary
- *            content instead. The dictionary buffer must then outlive its
- *            users.
+ *  Special: Adding a NULL (or 0-size) dictionary invalidates previous dictionary,
+ *           meaning "return to no-dictionary mode".
+ *  Note 1 : Dictionary will be used for all future compression jobs.
+ *           To return to "no-dictionary" situation, load a NULL dictionary
  *  Note 2 : Loading a dictionary involves building tables, which are dependent on compression parameters.
  *           For this reason, compression parameters cannot be changed anymore after loading a dictionary.
- *           It's also a CPU-heavy operation, with non-negligible impact on latency.
- *  Note 3 : Dictionary will be used for all future compression jobs.
- *           To return to "no-dictionary" situation, load a NULL dictionary
- *  Note 5 : Use ZSTD_CCtx_loadDictionary_advanced() to select how dictionary
- *           content will be interpreted.
- */
+ *           It's also a CPU consuming operation, with non-negligible impact on latency.
+ *  Note 3 :`dict` content will be copied internally.
+ *           Use ZSTD_CCtx_loadDictionary_byReference() to reference dictionary content instead.
+ *           In such a case, dictionary buffer must outlive its users.
+ *  Note 4 : Use ZSTD_CCtx_loadDictionary_advanced()
+ *           to precisely select how dictionary content must be interpreted. */
 ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
 ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_byReference(ZSTD_CCtx* cctx, const void* dict, size_t dictSize);
-ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode);
+ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
 
 
 /*! ZSTD_CCtx_refCDict() :
@@ -1103,8 +1109,7 @@ ZSTDLIB_API size_t ZSTD_CCtx_loadDictionary_advanced(ZSTD_CCtx* cctx, const void
  *  Special : adding a NULL CDict means "return to no-dictionary mode".
  *  Note 1 : Currently, only one dictionary can be managed.
  *           Adding a new dictionary effectively "discards" any previous one.
- *  Note 2 : CDict is just referenced, its lifetime must outlive CCtx.
- */
+ *  Note 2 : CDict is just referenced, its lifetime must outlive CCtx. */
 ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
 
 /*! ZSTD_CCtx_refPrefix() :
@@ -1114,20 +1119,29 @@ ZSTDLIB_API size_t ZSTD_CCtx_refCDict(ZSTD_CCtx* cctx, const ZSTD_CDict* cdict);
  *  Subsequent compression jobs will be done without prefix (if none is explicitly referenced).
  *  If there is a need to use same prefix multiple times, consider embedding it into a ZSTD_CDict instead.
  * @result : 0, or an error code (which can be tested with ZSTD_isError()).
- *  Special : Adding any prefix (including NULL) invalidates any previous prefix or dictionary
+ *  Special: Adding any prefix (including NULL) invalidates any previous prefix or dictionary
  *  Note 1 : Prefix buffer is referenced. It must outlive compression job.
  *  Note 2 : Referencing a prefix involves building tables, which are dependent on compression parameters.
- *           It's a CPU-heavy operation, with non-negligible impact on latency.
- *  Note 3 : By default, the prefix is treated as raw content
- *           (ZSTD_dm_rawContent). Use ZSTD_CCtx_refPrefix_advanced() to alter
- *           dictMode. */
+ *           It's a CPU consuming operation, with non-negligible impact on latency.
+ *  Note 3 : By default, the prefix is treated as raw content (ZSTD_dm_rawContent).
+ *           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode. */
 ZSTDLIB_API size_t ZSTD_CCtx_refPrefix(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize);
-ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictMode_e dictMode);
+ZSTDLIB_API size_t ZSTD_CCtx_refPrefix_advanced(ZSTD_CCtx* cctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType);
+
+/*! ZSTD_CCtx_reset() :
+ *  Return a CCtx to clean state.
+ *  Useful after an error, or to interrupt an ongoing compression job and start a new one.
+ *  Any internal data not yet flushed is cancelled.
+ *  Dictionary (if any) is dropped.
+ *  All parameters are back to default values.
+ *  It's possible to modify compression parameters after a reset.
+ */
+ZSTDLIB_API void ZSTD_CCtx_reset(ZSTD_CCtx* cctx);
 
 
 
 typedef enum {
-    ZSTD_e_continue=0, /* collect more data, encoder transparently decides when to output result, for optimal conditions */
+    ZSTD_e_continue=0, /* collect more data, encoder decides when to output compressed result, for optimal conditions */
     ZSTD_e_flush,      /* flush any data provided so far - frame will continue, future data can still reference previous data for better compression */
     ZSTD_e_end         /* flush any remaining data and close current frame. Any additional data starts a new frame. */
 } ZSTD_EndDirective;
@@ -1138,29 +1152,26 @@ typedef enum {
  *  - Compression parameters cannot be changed once compression is started.
  *  - outpot->pos must be <= dstCapacity, input->pos must be <= srcSize
  *  - outpot->pos and input->pos will be updated. They are guaranteed to remain below their respective limit.
- *  - @return provides the minimum amount of data still to flush from internal buffers
+ *  - In single-thread mode (default), function is blocking : it completed its job before returning to caller.
+ *  - In multi-thread mode, function is non-blocking : it just acquires a copy of input, and distribute job to internal worker threads,
+ *                                                     and then immediately returns, just indicating that there is some data remaining to be flushed.
+ *                                                     The function nonetheless guarantees forward progress : it will return only after it reads or write at least 1+ byte.
+ *  - Exception : in multi-threading mode, if the first call requests a ZSTD_e_end directive, it is blocking : it will complete compression before giving back control to caller.
+ *  - @return provides a minimum amount of data remaining to be flushed from internal buffers
  *            or an error code, which can be tested using ZSTD_isError().
- *            if @return != 0, flush is not fully completed, there is some data left within internal buffers.
- *  - after a ZSTD_e_end directive, if internal buffer is not fully flushed,
+ *            if @return != 0, flush is not fully completed, there is still some data left within internal buffers.
+ *            This is useful for ZSTD_e_flush, since in this case more flushes are necessary to empty all buffers.
+ *            For ZSTD_e_end, @return == 0 when internal buffers are fully flushed and frame is completed.
+ *  - after a ZSTD_e_end directive, if internal buffer is not fully flushed (@return != 0),
  *            only ZSTD_e_end or ZSTD_e_flush operations are allowed.
- *            It is necessary to fully flush internal buffers
- *            before starting a new compression job, or changing compression parameters.
+ *            Before starting a new compression job, or changing compression parameters,
+ *            it is required to fully flush internal buffers.
  */
 ZSTDLIB_API size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
                                           ZSTD_outBuffer* output,
                                           ZSTD_inBuffer* input,
                                           ZSTD_EndDirective endOp);
 
-/*! ZSTD_CCtx_reset() :
- *  Return a CCtx to clean state.
- *  Useful after an error, or to interrupt an ongoing compression job and start a new one.
- *  Any internal data not yet flushed is cancelled.
- *  Dictionary (if any) is dropped.
- *  All parameters are back to default values.
- *  It's possible to modify compression parameters after a reset.
- */
-ZSTDLIB_API void ZSTD_CCtx_reset(ZSTD_CCtx* cctx);   /* Not ready yet ! */
-
 
 /*! ZSTD_compress_generic_simpleArgs() :
  *  Same as ZSTD_compress_generic(),
@@ -1194,25 +1205,26 @@ ZSTDLIB_API size_t ZSTD_compress_generic_simpleArgs (
  *  for static allocation for single-threaded compression.
  */
 ZSTDLIB_API ZSTD_CCtx_params* ZSTD_createCCtxParams(void);
+ZSTDLIB_API size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
 
-/*! ZSTD_resetCCtxParams() :
- *  Reset params to default, with the default compression level.
+
+/*! ZSTD_CCtxParams_reset() :
+ *  Reset params to default values.
  */
-ZSTDLIB_API size_t ZSTD_resetCCtxParams(ZSTD_CCtx_params* params);
+ZSTDLIB_API size_t ZSTD_CCtxParams_reset(ZSTD_CCtx_params* params);
 
-/*! ZSTD_initCCtxParams() :
+/*! ZSTD_CCtxParams_init() :
  *  Initializes the compression parameters of cctxParams according to
  *  compression level. All other parameters are reset to their default values.
  */
-ZSTDLIB_API size_t ZSTD_initCCtxParams(ZSTD_CCtx_params* cctxParams, int compressionLevel);
+ZSTDLIB_API size_t ZSTD_CCtxParams_init(ZSTD_CCtx_params* cctxParams, int compressionLevel);
 
-/*! ZSTD_initCCtxParams_advanced() :
+/*! ZSTD_CCtxParams_init_advanced() :
  *  Initializes the compression and frame parameters of cctxParams according to
  *  params. All other parameters are reset to their default values.
  */
-ZSTDLIB_API size_t ZSTD_initCCtxParams_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
+ZSTDLIB_API size_t ZSTD_CCtxParams_init_advanced(ZSTD_CCtx_params* cctxParams, ZSTD_parameters params);
 
-ZSTDLIB_API size_t ZSTD_freeCCtxParams(ZSTD_CCtx_params* params);
 
 /*! ZSTD_CCtxParam_setParameter() :
  *  Similar to ZSTD_CCtx_setParameter.
@@ -1225,9 +1237,10 @@ ZSTDLIB_API size_t ZSTD_CCtxParam_setParameter(ZSTD_CCtx_params* params, ZSTD_cP
 
 /*! ZSTD_CCtx_setParametersUsingCCtxParams() :
  *  Apply a set of ZSTD_CCtx_params to the compression context.
- *  This must be done before the dictionary is loaded.
- *  The pledgedSrcSize is treated as unknown.
- *  Multithreading parameters are applied only if nbThreads > 1.
+ *  This can be done even after compression is started,
+ *    if nbWorkers==0, this will have no impact until a new compression is started.
+ *    if nbWorkers>=1, new parameters will be picked up at next job,
+ *       with a few restrictions (windowLog, pledgedSrcSize, nbWorkers, jobSize, and overlapLog are not updated).
  */
 ZSTDLIB_API size_t ZSTD_CCtx_setParametersUsingCCtxParams(
         ZSTD_CCtx* cctx, const ZSTD_CCtx_params* params);
@@ -1254,9 +1267,9 @@ ZSTDLIB_API size_t ZSTD_CCtx_setParametersUsingCCtxParams(
  *  Note 3 : Use ZSTD_DCtx_loadDictionary_advanced() to select
  *           how dictionary content will be interpreted and loaded.
  */
-ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);   /* not implemented */
-ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);   /* not implemented */
-ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode);   /* not implemented */
+ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_byReference(ZSTD_DCtx* dctx, const void* dict, size_t dictSize);
+ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void* dict, size_t dictSize, ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictContentType_e dictContentType);
 
 
 /*! ZSTD_DCtx_refDDict() :
@@ -1268,7 +1281,7 @@ ZSTDLIB_API size_t ZSTD_DCtx_loadDictionary_advanced(ZSTD_DCtx* dctx, const void
  *  Special : adding a NULL DDict means "return to no-dictionary mode".
  *  Note 2 : DDict is just referenced, its lifetime must outlive its usage from DCtx.
  */
-ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);   /* not implemented */
+ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
 
 
 /*! ZSTD_DCtx_refPrefix() :
@@ -1282,8 +1295,8 @@ ZSTDLIB_API size_t ZSTD_DCtx_refDDict(ZSTD_DCtx* dctx, const ZSTD_DDict* ddict);
  *           Use ZSTD_CCtx_refPrefix_advanced() to alter dictMode.
  *  Note 4 : Referencing a raw content prefix has almost no cpu nor memory cost.
  */
-ZSTDLIB_API size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize);   /* not implemented */
-ZSTDLIB_API size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictMode_e dictMode);   /* not implemented */
+ZSTDLIB_API size_t ZSTD_DCtx_refPrefix(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize);
+ZSTDLIB_API size_t ZSTD_DCtx_refPrefix_advanced(ZSTD_DCtx* dctx, const void* prefix, size_t prefixSize, ZSTD_dictContentType_e dictContentType);
 
 
 /*! ZSTD_DCtx_setMaxWindowSize() :
@@ -1376,7 +1389,7 @@ ZSTDLIB_API void ZSTD_DCtx_reset(ZSTD_DCtx* dctx);
 ZSTDLIB_API size_t ZSTD_getBlockSize   (const ZSTD_CCtx* cctx);
 ZSTDLIB_API size_t ZSTD_compressBlock  (ZSTD_CCtx* cctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
 ZSTDLIB_API size_t ZSTD_decompressBlock(ZSTD_DCtx* dctx, void* dst, size_t dstCapacity, const void* src, size_t srcSize);
-ZSTDLIB_API size_t ZSTD_insertBlock(ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize);  /**< insert uncompressed block into `dctx` history. Useful for multi-blocks decompression */
+ZSTDLIB_API size_t ZSTD_insertBlock    (ZSTD_DCtx* dctx, const void* blockStart, size_t blockSize);  /**< insert uncompressed block into `dctx` history. Useful for multi-blocks decompression. */
 
 
 #endif   /* ZSTD_H_ZSTD_STATIC_LINKING_ONLY */