Merge pull request #18905 from volzhs/zstd-1.3.4

Update zstd to 1.3.4

thirdparty/README.md

@@ -480,7 +480,7 @@ Files extracted from upstream source:
 ## zstd
 
 - Upstream: https://github.com/facebook/zstd
-- Version: 1.3.3
+- Version: 1.3.4
 - License: BSD-3-Clause
 
 Files extracted from upstream source:

thirdparty/zstd/common/bitstream.h

@@ -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 */

thirdparty/zstd/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 */

thirdparty/zstd/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 */

thirdparty/zstd/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";

thirdparty/zstd/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`).
 */

thirdparty/zstd/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);
     }   }
 

thirdparty/zstd/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()
+ *  HUF detailed API
+ * ****************************************/
 
-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_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)

thirdparty/zstd/common/pool.c

@@ -12,6 +12,7 @@
 /* ======   Dependencies   ======= */
 #include <stddef.h>  /* size_t */
 #include "pool.h"
+#include "zstd_internal.h"  /* ZSTD_malloc, ZSTD_free */
 
 /* ======   Compiler specifics   ====== */
 #if defined(_MSC_VER)
@@ -193,32 +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; }
 
-    ZSTD_pthread_mutex_lock(&ctx->queueMutex);
-    {   POOL_job const job = {function, opaque};
+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;
 
-        /* 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;
-        }
-    }
-    ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
+    ctx->queueEmpty = 0;
+    ctx->queue[ctx->queueTail] = job;
+    ctx->queueTail = (ctx->queueTail + 1) % ctx->queueSize;
     ZSTD_pthread_cond_signal(&ctx->queuePopCond);
 }
 
-#else  /* ZSTD_MULTITHREAD  not defined */
-/* No multi-threading support */
+void POOL_add(POOL_ctx* ctx, POOL_function function, void* opaque)
+{
+    assert(ctx != NULL);
+    ZSTD_pthread_mutex_lock(&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);
+    }
+    POOL_add_internal(ctx, function, opaque);
+    ZSTD_pthread_mutex_unlock(&ctx->queueMutex);
+}
 
-/* We don't need any data, but if it is empty malloc() might return NULL. */
+
+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);
+    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. */
 struct POOL_ctx_s {
     int dummy;
 };
@@ -240,11 +263,17 @@ 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) {
     if (ctx==NULL) return 0;  /* supports sizeof NULL */
     assert(ctx == &g_ctx);

thirdparty/zstd/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)

thirdparty/zstd/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 */
 

thirdparty/zstd/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 */

thirdparty/zstd/common/zstd_internal.h

@@ -132,14 +132,15 @@ 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,
@@ -228,8 +229,6 @@ 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;
 
 const seqStore_t* ZSTD_getSeqStore(const ZSTD_CCtx* ctx);   /* compress & dictBuilder */

thirdparty/zstd/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;
         }   }

thirdparty/zstd/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);
 }

thirdparty/zstd/compress/zstd_compress_internal.h

@@ -30,8 +30,14 @@ extern "C" {
 /*-*************************************
 *  Constants
 ***************************************/
-static const U32 g_searchStrength = 8;
-#define HASH_READ_SIZE 8
+#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 */
 
 
 /*-*************************************
@@ -43,7 +49,7 @@ 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_dictContentType_e dictContentType;
 } ZSTD_prefixDict;
 
 typedef struct {
@@ -51,7 +57,6 @@ typedef struct {
     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;
@@ -93,12 +98,44 @@ typedef struct {
     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.
@@ -111,60 +148,68 @@ typedef struct {
     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;
-    U32 forceWindow;           /* force back-references to respect limit of
+    int disableLiteralCompression;
+    int forceWindow;           /* force back-references to respect limit of
                                 * 1<<wLog, even for dictionary */
 
     /* Multithreading: used to pass parameters to mtctx */
-    U32 nbThreads;
+    unsigned nbWorkers;
     unsigned jobSize;
     unsigned overlapSizeLog;
 
     /* Long distance matching parameters */
     ldmParams_t ldmParams;
 
-    /* For use with createCCtxParams() and freeCCtxParams() only */
+    /* Internal use, for createCCtxParams() and freeCCtxParams() only */
     ZSTD_customMem customMem;
-
 };  /* typedef'd to ZSTD_CCtx_params within "zstd.h" */
 
 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;
+    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;
-    U64 pledgedSrcSizePlusOne;  /* this way, 0 (default) == unknown */
-    U64 consumedSrcSize;
+    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 */
-    optState_t optState;
-    ldmState_t ldmState;    /* long distance matching state */
-    U32* hashTable;
-    U32* hashTable3;
-    U32* chainTable;
-    ZSTD_entropyCTables_t* entropy;
+    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;
@@ -191,6 +236,12 @@ struct ZSTD_CCtx_s {
 };
 
 
+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,
@@ -359,10 +410,12 @@ MEM_STATIC size_t ZSTD_count(const BYTE* pIn, const BYTE* pMatch, const BYTE* co
 }
 
 /** 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)
+ *  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);
@@ -372,8 +425,8 @@ MEM_STATIC size_t ZSTD_count_2segments(const BYTE* ip, const BYTE* match, const
 
 
 /*-*************************************
-*  Hashes
-***************************************/
+ *  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 */
@@ -411,6 +464,171 @@ MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
     }
 }
 
+/*-*************************************
+*  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
@@ -421,6 +639,13 @@ MEM_STATIC size_t ZSTD_hashPtr(const void* p, U32 hBits, U32 mls)
  * 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.
@@ -446,7 +671,7 @@ ZSTD_compressionParameters ZSTD_getCParamsFromCDict(const ZSTD_CDict* cdict);
  * 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_dictContentType_e dictContentType,
                                     const ZSTD_CDict* cdict,
                                     ZSTD_CCtx_params params,
                                     unsigned long long pledgedSrcSize);
@@ -459,4 +684,26 @@ size_t ZSTD_compress_advanced_internal(ZSTD_CCtx* cctx,
                                  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 */

thirdparty/zstd/compress/zstd_double_fast.c

@@ -12,44 +12,58 @@
 #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;
+    const U32 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;
+    /* 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 */
@@ -76,7 +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)) ) {
@@ -99,14 +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 */
@@ -129,61 +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) */
@@ -209,7 +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;
@@ -220,7 +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);
@@ -245,10 +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;
         }   }
 
@@ -272,7 +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;
@@ -283,27 +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);
     }
 }

thirdparty/zstd/compress/zstd_double_fast.h

@@ -16,11 +16,18 @@ extern "C" {
 #endif
 
 #include "mem.h"      /* U32 */
-#include "zstd.h"     /* ZSTD_CCtx, size_t */
+#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);
 
-void ZSTD_fillDoubleHashTable(ZSTD_CCtx* cctx, const void* end, const U32 mls);
-size_t ZSTD_compressBlock_doubleFast(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
-size_t ZSTD_compressBlock_doubleFast_extDict(ZSTD_CCtx* ctx, const void* src, size_t srcSize);
 
 #if defined (__cplusplus)
 }

thirdparty/zstd/compress/zstd_fast.c

@@ -12,39 +12,48 @@
 #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 */
@@ -57,7 +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;
@@ -66,21 +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;
@@ -88,8 +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)
@@ -97,65 +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;
@@ -171,11 +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;
@@ -186,7 +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 */
@@ -195,8 +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);
@@ -207,8 +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;
@@ -217,27 +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);
     }
 }

thirdparty/zstd/compress/zstd_fast.h

@@ -16,13 +16,17 @@ extern "C" {
 #endif
 
 #include "mem.h"      /* U32 */
-#include "zstd.h"     /* ZSTD_CCtx, size_t */
+#include "zstd_compress_internal.h"
 
-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);
+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)
 }

thirdparty/zstd/compress/zstd_lazy.c

@@ -15,76 +15,90 @@
 /*-*************************************
 *  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 BYTE* const iend,
-                U32 nbCompares, U32 const mls, U32 const 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+1;
-    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;
 
-    DEBUGLOG(8, "ZSTD_insertBt1 (%u)", current);
-
-    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 */
         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;
+        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;
@@ -93,11 +107,8 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
                 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 */
             break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt tree */
@@ -108,6 +119,8 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
             *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 */
+            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 {
@@ -115,184 +128,205 @@ static U32 ZSTD_insertBt1(ZSTD_CCtx* zc,
             *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 */
-    assert(matchEndIdx > current + 8);
-    return matchEndIdx - (current + 8);
-}
-
-FORCE_INLINE_TEMPLATE
-void ZSTD_updateTree_internal(ZSTD_CCtx* zc,
-                const BYTE* const ip, const BYTE* const iend,
-                const U32 nbCompares, const U32 mls, const U32 extDict)
-{
-    const BYTE* const base = zc->base;
-    U32 const target = (U32)(ip - base);
-    U32 idx = zc->nextToUpdate;
-    DEBUGLOG(7, "ZSTD_updateTree_internal, from %u to %u  (extDict:%u)",
-                idx, target, extDict);
-
-    while(idx < target)
-        idx += ZSTD_insertBt1(zc, base+idx, iend, nbCompares, mls, extDict);
-    zc->nextToUpdate = target;
-}
-
-void ZSTD_updateTree(ZSTD_CCtx* zc,
-                const BYTE* const ip, const BYTE* const iend,
-                const U32 nbCompares, const U32 mls)
-{
-    ZSTD_updateTree_internal(zc, ip, iend, nbCompares, mls, 0 /*extDict*/);
-}
-
-void ZSTD_updateTree_extDict(ZSTD_CCtx* zc,
-                const BYTE* const ip, const BYTE* const iend,
-                const U32 nbCompares, const U32 mls)
-{
-    ZSTD_updateTree_internal(zc, ip, iend, nbCompares, mls, 1 /*extDict*/);
 }
 
 
-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+1;
-    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);
 
+    U32*         nextCandidate = bt + 2*(matchIndex&btMask);
+    U32*         unsortedMark = bt + 2*(matchIndex&btMask) + 1;
+    U32          nbCompares = 1U << cParams->searchLog;
+    U32          nbCandidates = nbCompares;
+    U32          previousCandidate = 0;
+
+    DEBUGLOG(7, "ZSTD_DUBT_findBestMatch (%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 */
-        const BYTE* match;
+    /* 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 ((!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] */
-        }
+    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 (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) */
+    /* 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++;
+    }
+
+    /* 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] */
             }
+
+            if (matchLength > bestLength) {
+                if (matchLength > matchEndIdx - matchIndex)
+                    matchEndIdx = matchIndex + (U32)matchLength;
+                if ( (4*(int)(matchLength-bestLength)) > (int)(ZSTD_highbit32(current-matchIndex+1) - ZSTD_highbit32((U32)offsetPtr[0]+1)) )
+                    bestLength = matchLength, *offsetPtr = ZSTD_REP_MOVE + current - matchIndex;
+                if (ip+matchLength == iend) {   /* equal : no way to know if inf or sup */
+                    break;   /* drop, to guarantee consistency (miss a little bit of compression) */
+                }
+            }
+
+            if (match[matchLength] < ip[matchLength]) {
+                /* match is smaller than current */
+                *smallerPtr = matchIndex;             /* update smaller idx */
+                commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+                if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+                smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+                matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+            } else {
+                /* match is larger than current */
+                *largerPtr = matchIndex;
+                commonLengthLarger = matchLength;
+                if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+                largerPtr = nextPtr;
+                matchIndex = nextPtr[0];
+        }   }
+
+        *smallerPtr = *largerPtr = 0;
+
+        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);
         }
-
-        if (match[matchLength] < ip[matchLength]) {
-            /* match is smaller than current */
-            *smallerPtr = matchIndex;             /* update smaller idx */
-            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
-            if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
-            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
-        } else {
-            /* match is larger than current */
-            *largerPtr = matchIndex;
-            commonLengthLarger = matchLength;
-            if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
-            largerPtr = nextPtr;
-            matchIndex = nextPtr[0];
-    }   }
-
-    *smallerPtr = *largerPtr = 0;
-
-    assert(matchEndIdx > current+8);
-    zc->nextToUpdate = matchEndIdx - 8;   /* skip repetitive patterns */
-    return bestLength;
+        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);
     }
 }
 
 
 /** 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);
     }
 }
 
@@ -305,15 +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);
@@ -322,35 +358,42 @@ 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--) {
         size_t currentMl=0;
@@ -381,35 +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* const zc,
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
-                        size_t* const offsetPtr,
-                        U32 const maxNbAttempts, U32 const 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);
     }
 }
 
@@ -418,30 +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;
@@ -462,13 +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;
         }
 
@@ -484,7 +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)) {
@@ -503,7 +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)) {
@@ -528,7 +568,7 @@ size_t ZSTD_compressBlock_lazy_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;
         }
 
@@ -538,73 +578,80 @@ _storeSequence:
             /* store sequence */
             matchLength = ZSTD_count(ip+4, ip+4-offset_2, iend) + 4;
             offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset; /* swap repcodes */
-            ZSTD_storeSeq(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 */
@@ -628,13 +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;
         }
 
@@ -661,7 +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)) {
@@ -691,7 +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)) {
@@ -713,7 +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;
         }
 
@@ -728,7 +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 ... (?) */
@@ -737,29 +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);
 }

thirdparty/zstd/compress/zstd_lazy.h

@@ -15,22 +15,39 @@
 extern "C" {
 #endif
 
-#include "mem.h"    /* U32 */
-#include "zstd.h"   /* ZSTD_CCtx, size_t */
+#include "zstd_compress_internal.h"
 
-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);
+U32 ZSTD_insertAndFindFirstIndex(
+        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+        const BYTE* ip);
 
-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);
+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_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);
+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)
 }

thirdparty/zstd/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;
                 }
+                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(ip, pMatch, iend);
-                if (curForwardMatchLength < ldmParams.minMatchLength) {
-                    continue;
+                    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);
+            U32 const offset = current - matchIndex;
+            rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
 
-            /* 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);
+            /* 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])) )) {
+    }
+    return iend - anchor;
+}
 
-            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_ldm_reduceTable() :
+ *  reduce table indexes by `reducerValue` */
+static void ZSTD_ldm_reduceTable(ldmEntry_t* const table, U32 const size,
+                                 U32 const reducerValue)
+{
+    U32 u;
+    for (u = 0; u < size; u++) {
+        if (table[u].offset < reducerValue) table[u].offset = 0;
+        else table[u].offset -= reducerValue;
+    }
+}
 
-            ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, rLength-MINMATCH);
+size_t ZSTD_ldm_generateSequences(
+        ldmState_t* ldmState, rawSeqStore_t* sequences,
+        ldmParams_t const* params, void const* src, size_t srcSize)
+{
+    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;
 
-            /* 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;
+    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);
         }
-    }
-
-    /* Overwrite rep */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        seqStorePtr->rep[i] = repToConfirm[i];
-
-    ZSTD_ldm_limitTableUpdate(cctx, anchor);
-    ZSTD_ldm_fillFastTables(cctx, anchor);
-
-    lastLiterals = blockCompressor(cctx, anchor, iend - anchor);
-    cctx->nextToUpdate = (U32)(iend - base);
-
-    /* Restore seqStorePtr->rep */
-    for (i = 0; i < ZSTD_REP_NUM; i++)
-        seqStorePtr->rep[i] = savedRep[i];
-
-    /* Return the last literals size */
-    return lastLiterals;
-}
-
-size_t ZSTD_compressBlock_ldm(ZSTD_CCtx* ctx,
-                              const void* src, size_t srcSize)
-{
-    return ZSTD_compressBlock_ldm_generic(ctx, src, srcSize);
-}
-
-static size_t ZSTD_compressBlock_ldm_extDict_generic(
-                                 ZSTD_CCtx* ctx,
-                                 const void* src, size_t srcSize)
-{
-    ldmState_t* const ldmState = &(ctx->ldmState);
-    const ldmParams_t ldmParams = ctx->appliedParams.ldmParams;
-    const U64 hashPower = ldmState->hashPower;
-    const U32 hBits = ldmParams.hashLog - ldmParams.bucketSizeLog;
-    const U32 ldmBucketSize = ((U32)1 << ldmParams.bucketSizeLog);
-    const U32 ldmTagMask = ((U32)1 << ldmParams.hashEveryLog) - 1;
-    seqStore_t* const seqStorePtr = &(ctx->seqStore);
-    const BYTE* const base = ctx->base;
-    const BYTE* const dictBase = ctx->dictBase;
-    const BYTE* const istart = (const BYTE*)src;
-    const BYTE* ip = istart;
-    const BYTE* anchor = istart;
-    const U32   lowestIndex = ctx->lowLimit;
-    const BYTE* const dictStart = dictBase + lowestIndex;
-    const U32   dictLimit = ctx->dictLimit;
-    const BYTE* const lowPrefixPtr = base + dictLimit;
-    const BYTE* const dictEnd = dictBase + dictLimit;
-    const BYTE* const iend = istart + srcSize;
-    const BYTE* const ilimit = iend - MAX(ldmParams.minMatchLength, HASH_READ_SIZE);
-
-    const ZSTD_blockCompressor blockCompressor =
-        ZSTD_selectBlockCompressor(ctx->appliedParams.cParams.strategy, 1);
-    U32* const repToConfirm = seqStorePtr->repToConfirm;
-    U32 savedRep[ZSTD_REP_NUM];
-    U64 rollingHash = 0;
-    const BYTE* lastHashed = NULL;
-    size_t i, lastLiterals;
-
-    /* Save seqStorePtr->rep and copy repToConfirm */
-    for (i = 0; i < ZSTD_REP_NUM; i++) {
-        savedRep[i] = repToConfirm[i] = seqStorePtr->rep[i];
-    }
-
-    /* Search Loop */
-    while (ip < ilimit) {  /* < instead of <=, because (ip+1) */
-        size_t mLength;
-        const U32 current = (U32)(ip-base);
-        size_t forwardMatchLength = 0, backwardMatchLength = 0;
-        ldmEntry_t* bestEntry = NULL;
-        if (ip != istart) {
-          rollingHash = ZSTD_ldm_updateHash(rollingHash, lastHashed[0],
-                                       lastHashed[ldmParams.minMatchLength],
-                                       hashPower);
+        /* 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);
-        }
-        lastHashed = ip;
-
-        if (ZSTD_ldm_getTag(rollingHash, hBits, ldmParams.hashEveryLog) !=
-                ldmTagMask) {
-            /* Don't insert and don't look for a match */
-           ip++;
-           continue;
-        }
-
-        /* Get the best entry and compute the match lengths */
-        {
-            ldmEntry_t* const bucket =
-                ZSTD_ldm_getBucket(ldmState,
-                                   ZSTD_ldm_getSmallHash(rollingHash, hBits),
-                                   ldmParams);
-            ldmEntry_t* cur;
-            size_t bestMatchLength = 0;
-            U32 const checksum = ZSTD_ldm_getChecksum(rollingHash, hBits);
-
-            for (cur = bucket; cur < bucket + ldmBucketSize; ++cur) {
-                const BYTE* const curMatchBase =
-                    cur->offset < dictLimit ? dictBase : base;
-                const BYTE* const pMatch = curMatchBase + cur->offset;
-                const BYTE* const matchEnd =
-                    cur->offset < dictLimit ? dictEnd : iend;
-                const BYTE* const lowMatchPtr =
-                    cur->offset < dictLimit ? dictStart : lowPrefixPtr;
-                size_t curForwardMatchLength, curBackwardMatchLength,
-                       curTotalMatchLength;
-
-                if (cur->checksum != checksum || cur->offset <= lowestIndex) {
-                    continue;
-                }
-
-                curForwardMatchLength = ZSTD_count_2segments(
-                                            ip, pMatch, iend,
-                                            matchEnd, lowPrefixPtr);
-                if (curForwardMatchLength < ldmParams.minMatchLength) {
-                    continue;
-                }
-                curBackwardMatchLength = ZSTD_ldm_countBackwardsMatch(
-                                             ip, anchor, pMatch, lowMatchPtr);
-                curTotalMatchLength = curForwardMatchLength +
-                                      curBackwardMatchLength;
-
-                if (curTotalMatchLength > bestMatchLength) {
-                    bestMatchLength = curTotalMatchLength;
-                    forwardMatchLength = curForwardMatchLength;
-                    backwardMatchLength = curBackwardMatchLength;
-                    bestEntry = cur;
-                }
-            }
-        }
-
-        /* No match found -- continue searching */
-        if (bestEntry == NULL) {
-            ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
-                                             (U32)(lastHashed - base),
-                                             ldmParams);
-            ip++;
-            continue;
-        }
-
-        /* Match found */
-        mLength = forwardMatchLength + backwardMatchLength;
-        ip -= backwardMatchLength;
-
-        /* Call the block compressor on the remaining literals */
-        {
-            /* ip = current - backwardMatchLength
-             * The match is at (bestEntry->offset - backwardMatchLength) */
-            U32 const matchIndex = bestEntry->offset;
-            U32 const offset = current - matchIndex;
-
-            /* Overwrite rep codes */
-            for (i = 0; i < ZSTD_REP_NUM; i++)
-                seqStorePtr->rep[i] = repToConfirm[i];
-
-            /* Fill the hash table for the block compressor */
-            ZSTD_ldm_limitTableUpdate(ctx, anchor);
-            ZSTD_ldm_fillFastTables(ctx, anchor);
-
-            /* Call block compressor and get remaining literals  */
-            lastLiterals = blockCompressor(ctx, anchor, ip - anchor);
-            ctx->nextToUpdate = (U32)(ip - base);
-
-            /* Update repToConfirm with the new offset */
-            for (i = ZSTD_REP_NUM - 1; i > 0; i--)
-                repToConfirm[i] = repToConfirm[i-1];
-            repToConfirm[0] = offset;
-
-            /* Store the sequence with the leftover literals */
-            ZSTD_storeSeq(seqStorePtr, lastLiterals, ip - lastLiterals,
-                          offset + ZSTD_REP_MOVE, mLength - MINMATCH);
-        }
-
-        /* Insert the current entry into the hash table */
-        ZSTD_ldm_makeEntryAndInsertByTag(ldmState, rollingHash, hBits,
-                                         (U32)(lastHashed - base),
-                                         ldmParams);
-
-        /* Fill the hash table from lastHashed+1 to ip+mLength */
-        assert(ip + backwardMatchLength == lastHashed);
-        if (ip + mLength < ilimit) {
-            rollingHash = ZSTD_ldm_fillLdmHashTable(
-                              ldmState, rollingHash, lastHashed,
-                              ip + mLength, base, hBits,
-                              ldmParams);
-            lastHashed = ip + mLength - 1;
-        }
-        ip += mLength;
-        anchor = ip;
-
-        /* check immediate repcode */
-        while (ip < ilimit) {
-            U32 const current2 = (U32)(ip-base);
-            U32 const repIndex2 = current2 - repToConfirm[1];
-            const BYTE* repMatch2 = repIndex2 < dictLimit ?
-                                    dictBase + repIndex2 : base + repIndex2;
-            if ( (((U32)((dictLimit-1) - repIndex2) >= 3) &
-                        (repIndex2 > lowestIndex))  /* intentional overflow */
-               && (MEM_read32(repMatch2) == MEM_read32(ip)) ) {
-                const BYTE* const repEnd2 = repIndex2 < dictLimit ?
-                                            dictEnd : iend;
-                size_t const repLength2 =
-                        ZSTD_count_2segments(ip+4, repMatch2+4, iend,
-                                             repEnd2, lowPrefixPtr) + 4;
-
-                U32 tmpOffset = repToConfirm[1];
-                repToConfirm[1] = repToConfirm[0];
-                repToConfirm[0] = tmpOffset;
-
-                ZSTD_storeSeq(seqStorePtr, 0, anchor, 0, repLength2-MINMATCH);
-
-                /* Fill the  hash table from lastHashed+1 to ip+repLength2*/
-                if (ip + repLength2 < ilimit) {
-                    rollingHash = ZSTD_ldm_fillLdmHashTable(
-                                      ldmState, rollingHash, lastHashed,
-                                      ip + repLength2, base, hBits,
-                                      ldmParams);
-                    lastHashed = ip + repLength2 - 1;
-                }
-                ip += repLength2;
-                anchor = ip;
-                continue;
-            }
-            break;
+            assert(newLeftoverSize == chunkSize);
+            leftoverSize += chunkSize;
         }
     }
-
-    /* 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;
+    return 0;
 }
 
-size_t ZSTD_compressBlock_ldm_extDict(ZSTD_CCtx* ctx,
-                                      const void* src, size_t srcSize)
+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;
+        }
+        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++;
+    }
+}
+
+/**
+ * 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)
 {
-    return ZSTD_compressBlock_ldm_extDict_generic(ctx, src, srcSize);
+    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;
+}
+
+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;
+
+        assert(sequence.offset <= (1U << cParams->windowLog));
+        assert(ip + sequence.litLength + sequence.matchLength <= iend);
+
+        /* 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--)
+                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;
+        }
+    }
+    /* 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;
+    }
 }

thirdparty/zstd/compress/zstd_ldm.h

@@ -22,32 +22,71 @@ extern "C" {
 ***************************************/
 
 #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.
- *
- *  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);
+ * 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():
+ *
+ * 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) */
@@ -58,8 +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)
 }

thirdparty/zstd/compress/zstd_opt.c

@@ -10,7 +10,6 @@
 
 #include "zstd_compress_internal.h"
 #include "zstd_opt.h"
-#include "zstd_lazy.h"   /* ZSTD_updateTree, ZSTD_updateTree_extDict */
 
 
 #define ZSTD_LITFREQ_ADD    2   /* scaling factor for litFreq, so that frequencies adapt faster to new stats. Also used for matchSum (?) */
@@ -244,14 +243,15 @@ MEM_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* const cctx, const BYTE* const ip)
+static U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_matchState_t* ms, const BYTE* const ip)
 {
-    U32* const hashTable3  = cctx->hashTable3;
-    U32 const hashLog3  = cctx->hashLog3;
-    const BYTE* const base = cctx->base;
-    U32 idx = cctx->nextToUpdate3;
-    U32 const target = cctx->nextToUpdate3 = (U32)(ip - base);
+    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;
@@ -265,36 +265,173 @@ static U32 ZSTD_insertAndFindFirstIndexHash3 (ZSTD_CCtx* const cctx, const BYTE*
 /*-*************************************
 *  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_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                const BYTE* const ip, const BYTE* const iend,
+                U32 const mls, U32 const extDict)
+{
+    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 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_CCtx* zc,
+                    ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                     const BYTE* const ip, const BYTE* const iLimit, int const extDict,
-                    U32 nbCompares, U32 const mls, U32 const sufficient_len,
                     U32 rep[ZSTD_REP_NUM], U32 const ll0,
-                    ZSTD_match_t* matches, const U32 lengthToBeat)
+                    ZSTD_match_t* matches, const U32 lengthToBeat, U32 const mls /* template */)
 {
-    const BYTE* const base = zc->base;
+    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 = zc->appliedParams.cParams.hashLog;
+    U32 const hashLog = cParams->hashLog;
     U32 const minMatch = (mls==3) ? 3 : 4;
-    U32* const hashTable = zc->hashTable;
+    U32* const hashTable = ms->hashTable;
     size_t const h  = ZSTD_hashPtr(ip, hashLog, mls);
     U32 matchIndex  = hashTable[h];
-    U32* const bt   = zc->chainTable;
-    U32 const btLog = zc->appliedParams.cParams.chainLog - 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;
-    U32 const 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;
     U32 const btLow = btMask >= current ? 0 : current - btMask;
-    U32 const windowLow = zc->lowLimit;
+    U32 const windowLow = ms->window.lowLimit;
     U32* smallerPtr = bt + 2*(current&btMask);
     U32* largerPtr  = bt + 2*(current&btMask) + 1;
     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;
+    U32 nbCompares = 1U << cParams->searchLog;
 
     size_t bestLength = lengthToBeat-1;
     DEBUGLOG(7, "ZSTD_insertBtAndGetAllMatches");
@@ -335,7 +472,7 @@ U32 ZSTD_insertBtAndGetAllMatches (
 
     /* HC3 match finder */
     if ((mls == 3) /*static*/ && (bestLength < mls)) {
-        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3 (zc, ip);
+        U32 const matchIndex3 = ZSTD_insertAndFindFirstIndexHash3(ms, ip);
         if ((matchIndex3 > windowLow)
           & (current - matchIndex3 < (1<<18)) /*heuristic : longer distance likely too expensive*/ ) {
             size_t mlen;
@@ -359,7 +496,7 @@ U32 ZSTD_insertBtAndGetAllMatches (
                 mnum = 1;
                 if ( (mlen > sufficient_len) |
                      (ip+mlen == iLimit) ) {  /* best possible length */
-                    zc->nextToUpdate = current+1;  /* skip insertion */
+                    ms->nextToUpdate = current+1;  /* skip insertion */
                     return 1;
     }   }   }   }
 
@@ -416,30 +553,29 @@ U32 ZSTD_insertBtAndGetAllMatches (
     *smallerPtr = *largerPtr = 0;
 
     assert(matchEndIdx > current+8);
-    zc->nextToUpdate = matchEndIdx - 8;  /* skip repetitive patterns */
+    ms->nextToUpdate = matchEndIdx - 8;  /* skip repetitive patterns */
     return mnum;
 }
 
 
 FORCE_INLINE_TEMPLATE U32 ZSTD_BtGetAllMatches (
-                        ZSTD_CCtx* zc,   /* Index table will be updated */
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iHighLimit, int const extDict,
-                        U32 const maxNbAttempts, U32 const matchLengthSearch, U32 const sufficient_len,
                         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 < zc->base + zc->nextToUpdate) return 0;   /* skipped area */
-    if (extDict) ZSTD_updateTree_extDict(zc, ip, iHighLimit, maxNbAttempts, matchLengthSearch);
-    else ZSTD_updateTree(zc, ip, iHighLimit, maxNbAttempts, matchLengthSearch);
+    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_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 3, sufficient_len, rep, ll0, matches, lengthToBeat);
+    case 3 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 3);
     default :
-    case 4 : return ZSTD_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 4, sufficient_len, rep, ll0, matches, lengthToBeat);
-    case 5 : return ZSTD_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 5, sufficient_len, rep, ll0, matches, lengthToBeat);
+    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_insertBtAndGetAllMatches(zc, ip, iHighLimit, extDict, maxNbAttempts, 6, sufficient_len, rep, ll0, matches, lengthToBeat);
+    case 6 : return ZSTD_insertBtAndGetAllMatches(ms, cParams, ip, iHighLimit, extDict, rep, ll0, matches, lengthToBeat, 6);
     }
 }
 
@@ -527,36 +663,33 @@ static int ZSTD_literalsContribution_cached(
 }
 
 FORCE_INLINE_TEMPLATE
-size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
+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* const seqStorePtr = &(ctx->seqStore);
-    optState_t* const 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 BYTE* const prefixStart = base + ctx->dictLimit;
+    const BYTE* const base = ms->window.base;
+    const BYTE* const prefixStart = base + ms->window.dictLimit;
 
-    U32 const maxSearches = 1U << ctx->appliedParams.cParams.searchLog;
-    U32 const sufficient_len = MIN(ctx->appliedParams.cParams.targetLength, ZSTD_OPT_NUM -1);
-    U32 const mls = ctx->appliedParams.cParams.searchLength;
-    U32 const 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* const opt = optStatePtr->priceTable;
     ZSTD_match_t* const matches = optStatePtr->matchTable;
     cachedLiteralPrice_t cachedLitPrice;
-    U32 rep[ZSTD_REP_NUM];
 
     /* init */
     DEBUGLOG(5, "ZSTD_compressBlock_opt_generic");
-    ctx->nextToUpdate3 = ctx->nextToUpdate;
+    ms->nextToUpdate3 = ms->nextToUpdate;
     ZSTD_rescaleFreqs(optStatePtr, (const BYTE*)src, srcSize);
     ip += (ip==prefixStart);
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) rep[i]=seqStorePtr->rep[i]; }
     memset(&cachedLitPrice, 0, sizeof(cachedLitPrice));
 
     /* Match Loop */
@@ -567,7 +700,7 @@ size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
         /* find first match */
         {   U32 const litlen = (U32)(ip - anchor);
             U32 const ll0 = !litlen;
-            U32 const nbMatches = ZSTD_BtGetAllMatches(ctx, ip, iend, extDict, maxSearches, mls, sufficient_len, rep, ll0, matches, minMatch);
+            U32 const nbMatches = ZSTD_BtGetAllMatches(ms, cParams, ip, iend, extDict, rep, ll0, matches, minMatch);
             if (!nbMatches) { ip++; continue; }
 
             /* initialize opt[0] */
@@ -653,7 +786,7 @@ size_t ZSTD_compressBlock_opt_generic(ZSTD_CCtx* ctx,
                 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(ctx, inr, iend, extDict, maxSearches, mls, sufficient_len, opt[cur].rep, ll0, matches, minMatch);
+                U32 const nbMatches = ZSTD_BtGetAllMatches(ms, cParams, inr, iend, extDict, opt[cur].rep, ll0, matches, minMatch);
                 U32 matchNb;
                 if (!nbMatches) continue;
 
@@ -749,37 +882,42 @@ _shortestPath:   /* cur, last_pos, best_mlen, best_off have to be set */
                 }
 
                 ZSTD_updateStats(optStatePtr, llen, anchor, offset, mlen);
-                ZSTD_storeSeq(seqStorePtr, llen, anchor, offset, mlen-MINMATCH);
+                ZSTD_storeSeq(seqStore, llen, anchor, offset, mlen-MINMATCH);
                 anchor = ip;
         }   }
         ZSTD_setLog2Prices(optStatePtr);
     }   /* while (ip < ilimit) */
 
-    /* Save reps for next block */
-    { int i; for (i=0; i<ZSTD_REP_NUM; i++) seqStorePtr->repToConfirm[i] = rep[i]; }
-
     /* Return the last literals size */
     return iend - anchor;
 }
 
 
-size_t ZSTD_compressBlock_btopt(ZSTD_CCtx* ctx, const void* src, size_t srcSize)
+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(ctx, src, srcSize, 0 /*optLevel*/, 0 /*extDict*/);
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 0 /*optLevel*/, 0 /*extDict*/);
 }
 
-size_t ZSTD_compressBlock_btultra(ZSTD_CCtx* ctx, const void* 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)
 {
-    return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 2 /*optLevel*/, 0 /*extDict*/);
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 2 /*optLevel*/, 0 /*extDict*/);
 }
 
-size_t ZSTD_compressBlock_btopt_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)
 {
-    return ZSTD_compressBlock_opt_generic(ctx, src, srcSize, 0 /*optLevel*/, 1 /*extDict*/);
+    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_generic(ctx, src, srcSize, 2 /*optLevel*/, 1 /*extDict*/);
+    return ZSTD_compressBlock_opt_generic(ms, seqStore, rep, cParams, src, srcSize, 2 /*optLevel*/, 1 /*extDict*/);
 }

thirdparty/zstd/compress/zstd_opt.h

@@ -15,13 +15,25 @@
 extern "C" {
 #endif
 
-#include "zstd.h"   /* ZSTD_CCtx, size_t */
+#include "zstd_compress_internal.h"
 
-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);
+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_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(
+        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_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)
 }

thirdparty/zstd/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);  /**< 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 may change in the future, to mean "empty" */
+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);
 
@@ -68,7 +68,7 @@ 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,
@@ -85,7 +85,7 @@ ZSTDLIB_API size_t ZSTDMT_initCStream_usingCDict(ZSTDMT_CCtx* mtctx,
  * List of parameters that can be set using ZSTDMT_setMTCtxParameter() */
 typedef enum {
     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 */
+    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() :
@@ -97,30 +97,46 @@ ZSTDLIB_API size_t ZSTDMT_setMTCtxParameter(ZSTDMT_CCtx* mtctx, ZSTDMT_parameter
 
 
 /*! 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);
 
-/* ZSTDMT_CCtxParam_setNbThreads()
- * Set nbThreads, and clamp it correctly,
- * also reset jobSize and overlapLog */
-size_t ZSTDMT_CCtxParam_setNbThreads(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_getNbThreads():
+/*! 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 */
-size_t ZSTDMT_getNbThreads(const ZSTDMT_CCtx* mtctx);
+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.
@@ -128,7 +144,7 @@ size_t ZSTDMT_getNbThreads(const ZSTDMT_CCtx* mtctx);
  *  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);
 

thirdparty/zstd/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;
@@ -344,6 +306,279 @@ static size_t HUF_decompress4X2_usingDTable_internal(
 }
 
 
+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_decompress1X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
+}
+
+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,
@@ -351,13 +586,12 @@ size_t HUF_decompress4X2_usingDTable(
 {
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     if (dtd.tableType != 0) return ERROR(GENERIC);
-    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable);
+    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, cSrc, cSrcSize, DTable, /* bmi2 */ 0);
 }
 
-
-size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+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)
+                                   void* workSpace, size_t wkspSize, int bmi2)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
@@ -367,7 +601,14 @@ size_t HUF_decompress4X2_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
     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);
+    }
+}

thirdparty/zstd/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  3
+#define ZSTD_VERSION_RELEASE  4
 
 #define ZSTD_VERSION_NUMBER  (ZSTD_VERSION_MAJOR *100*100 + ZSTD_VERSION_MINOR *100 + ZSTD_VERSION_RELEASE)
 ZSTDLIB_API unsigned ZSTD_versionNumber(void);   /**< 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<<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 scenario */
+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 ZSTD_customMem const 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;
+
 
 
 /***************************************
@@ -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();
@@ -529,46 +533,86 @@ 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);
 
+/*! 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 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_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() */
+
+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);
+
+ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize,
+                                                  ZSTD_dictLoadMethod_e dictLoadMethod,
+                                                  ZSTD_dictContentType_e dictContentType,
+                                                  ZSTD_compressionParameters cParams,
+                                                  ZSTD_customMem customMem);
+
+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_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* (same address as workspace, but different type),
- *           or NULL if error (typically size too small)
- *  Note : zstd will never resize nor malloc() when using a static cctx.
- *         If it needs more memory than available, it will simply error out.
- *  Note 2 : there is no corresponding "free" function.
- *           Since workspace was allocated externally, it must be freed externally too.
- *  Limitation 1 : currently not compatible with internal CDict creation, such as
- *                 ZSTD_CCtx_loadDictionary() or ZSTD_initCStream_usingDict().
- *  Limitation 2 : currently not compatible with multi-threading
- */
-ZSTDLIB_API ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize);
-
 
 /*! ZSTD_createCDict_byReference() :
  *  Create a digested dictionary for compression
@@ -576,38 +620,6 @@ ZSTDLIB_API ZSTD_CCtx* ZSTD_initStaticCCtx(void* workspace, size_t workspaceSize
  *  It is important that dictBuffer outlives CDict, it must remain read accessible throughout the lifetime of CDict */
 ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_byReference(const void* dictBuffer, size_t dictSize, int compressionLevel);
 
-typedef enum { ZSTD_dm_auto=0,        /* dictionary is "full" if it starts with ZSTD_MAGIC_DICTIONARY, otherwise it is "rawContent" */
-               ZSTD_dm_rawContent,    /* ensures dictionary is always loaded as rawContent, even if it starts with ZSTD_MAGIC_DICTIONARY */
-               ZSTD_dm_fullDict       /* refuses to load a dictionary if it does not respect Zstandard's specification */
-} ZSTD_dictMode_e;
-/*! ZSTD_createCDict_advanced() :
- *  Create a ZSTD_CDict using external alloc and free, and customized compression parameters */
-ZSTDLIB_API ZSTD_CDict* ZSTD_createCDict_advanced(const void* dict, size_t dictSize,
-                                                  ZSTD_dictLoadMethod_e dictLoadMethod,
-                                                  ZSTD_dictMode_e dictMode,
-                                                  ZSTD_compressionParameters cParams,
-                                                  ZSTD_customMem customMem);
-
-/*! ZSTD_initStaticCDict() :
- *  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* (same address as workspace, but different type),
- *           or NULL if error (typically, size too small).
- *  Note : there is no corresponding "free" function.
- *         Since workspace was allocated externally, it must be freed externally.
- */
-ZSTDLIB_API ZSTD_CDict* ZSTD_initStaticCDict(
-                            void* workspace, size_t workspaceSize,
-                      const void* dict, size_t dictSize,
-                            ZSTD_dictLoadMethod_e dictLoadMethod, ZSTD_dictMode_e dictMode,
-                            ZSTD_compressionParameters cParams);
-
 /*! ZSTD_getCParams() :
 *   @return ZSTD_compressionParameters structure for a selected compression level and estimated srcSize.
 *   `estimatedSrcSize` value is optional, select 0 if not known */
@@ -652,28 +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* (same address as workspace, but different type),
- *           or NULL if error (typically 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.
@@ -681,26 +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.
@@ -732,8 +702,6 @@ 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. 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,
@@ -748,14 +716,28 @@ ZSTDLIB_API size_t ZSTD_initCStream_usingCDict_advanced(ZSTD_CStream* zcs, const
  *  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.
  *  For the time being, pledgedSrcSize==0 is interpreted as "srcSize unknown" for compatibility with older programs,
- *  but it may change to mean "empty" in some future version, so prefer using macro ZSTD_CONTENTSIZE_UNKNOWN.
+ *  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 */
@@ -924,10 +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 :
@@ -944,12 +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 */
@@ -967,42 +947,76 @@ 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 will be written into frame header _whenever known_ (default:1)
@@ -1012,58 +1026,45 @@ typedef enum {
     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. This value is only enforced in streaming (non-blocking) mode.
-                              * Each compression job is completed in parallel, so indirectly controls the nb of active threads.
+    /* 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 : informational value (typically, the one being set, possibly corrected),
+ *  @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);
 
@@ -1079,26 +1080,24 @@ ZSTDLIB_API size_t ZSTD_CCtx_setParameter(ZSTD_CCtx* cctx, ZSTD_cParameter param
 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() :
@@ -1110,8 +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() :
@@ -1121,20 +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;
@@ -1150,10 +1157,11 @@ typedef enum {
  *                                                     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 the minimum amount of data remaining to be flushed from internal buffers
+ *  - @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 still some data left within internal buffers.
- *            This is useful to determine if a ZSTD_e_flush or ZSTD_e_end directive is completed.
+ *            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.
  *            Before starting a new compression job, or changing compression parameters,
@@ -1164,16 +1172,6 @@ ZSTDLIB_API size_t ZSTD_compress_generic (ZSTD_CCtx* cctx,
                                           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(),
@@ -1207,25 +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.
@@ -1238,9 +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);
@@ -1267,9 +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() :
@@ -1281,7 +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() :
@@ -1295,8 +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() :
@@ -1389,7 +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 */