diff contrib/python-zstandard/zstd/decompress/huf_decompress.c @ 37495:b1fb341d8a61

zstandard: vendor python-zstandard 0.9.0

This was just released. It features a number of goodies. More info at
https://gregoryszorc.com/blog/2018/04/09/release-of-python-zstandard-0.9/.

The clang-format ignore list was updated to reflect the new source
of files.

The project contains a vendored copy of zstandard 1.3.4. The old
version was 1.1.3. One of the changes between those versions is that
zstandard is now dual licensed BSD + GPLv2 and the patent rights grant
has been removed. Good riddance.

The API should be backwards compatible. So no changes in core
should be needed. However, there were a number of changes in the
library that we'll want to adapt to. Those will be addressed in
subsequent commits.

Differential Revision: https://phab.mercurial-scm.org/D3198

--- a/contrib/python-zstandard/zstd/decompress/huf_decompress.c	Sun Apr 08 01:08:43 2018 +0200
+++ b/contrib/python-zstandard/zstd/decompress/huf_decompress.c	Mon Apr 09 10:13:29 2018 -0700
@@ -33,41 +33,35 @@
 ****************************************************************** */
 
 /* **************************************************************
-*  Compiler specifics
-****************************************************************/
-#if defined (__cplusplus) || (defined (__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) /* C99 */)
-/* inline is defined */
-#elif defined(_MSC_VER) || defined(__GNUC__)
-#  define inline __inline
-#else
-#  define inline /* disable inline */
-#endif
-
-#ifdef _MSC_VER    /* Visual Studio */
-#  pragma warning(disable : 4127)        /* disable: C4127: conditional expression is constant */
-#endif
-
-
-/* **************************************************************
 *  Dependencies
 ****************************************************************/
 #include <string.h>     /* memcpy, memset */
 #include "bitstream.h"  /* BIT_* */
+#include "compiler.h"
 #include "fse.h"        /* header compression */
 #define HUF_STATIC_LINKING_ONLY
 #include "huf.h"
+#include "error_private.h"
 
 
 /* **************************************************************
 *  Error Management
 ****************************************************************/
+#define HUF_isError ERR_isError
 #define HUF_STATIC_ASSERT(c) { enum { HUF_static_assert = 1/(int)(!!(c)) }; }   /* use only *after* variable declarations */
+#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_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)
@@ -81,19 +75,27 @@
 /*-***************************/
 /*  single-symbol decoding   */
 /*-***************************/
-
 typedef struct { BYTE byte; BYTE nbBits; } HUF_DEltX2;   /* single-symbol decoding */
 
-size_t HUF_readDTableX2 (HUF_DTable* DTable, const void* src, size_t srcSize)
+size_t HUF_readDTableX2_wksp(HUF_DTable* DTable, const void* src, size_t srcSize, void* workSpace, size_t wkspSize)
 {
-    BYTE huffWeight[HUF_SYMBOLVALUE_MAX + 1];
-    U32 rankVal[HUF_TABLELOG_ABSOLUTEMAX + 1];   /* large enough for values from 0 to 16 */
     U32 tableLog = 0;
     U32 nbSymbols = 0;
     size_t iSize;
     void* const dtPtr = DTable + 1;
     HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
 
+    U32* rankVal;
+    BYTE* huffWeight;
+    size_t spaceUsed32 = 0;
+
+    rankVal = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_ABSOLUTEMAX + 1;
+    huffWeight = (BYTE *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
+
+    if ((spaceUsed32 << 2) > wkspSize) return ERROR(tableLog_tooLarge);
+
     HUF_STATIC_ASSERT(sizeof(DTableDesc) == sizeof(HUF_DTable));
     /* memset(huffWeight, 0, sizeof(huffWeight)); */   /* is not necessary, even though some analyzer complain ... */
 
@@ -102,16 +104,16 @@
 
     /* Table header */
     {   DTableDesc dtd = HUF_getDTableDesc(DTable);
-        if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge);   /* DTable too small, huffman tree cannot fit in */
+        if (tableLog > (U32)(dtd.maxTableLog+1)) return ERROR(tableLog_tooLarge);   /* DTable too small, Huffman tree cannot fit in */
         dtd.tableType = 0;
         dtd.tableLog = (BYTE)tableLog;
         memcpy(DTable, &dtd, sizeof(dtd));
     }
 
-    /* Prepare ranks */
+    /* Calculate starting value for each rank */
     {   U32 n, nextRankStart = 0;
         for (n=1; n<tableLog+1; n++) {
-            U32 current = nextRankStart;
+            U32 const current = nextRankStart;
             nextRankStart += (rankVal[n] << (n-1));
             rankVal[n] = current;
     }   }
@@ -121,19 +123,28 @@
         for (n=0; n<nbSymbols; n++) {
             U32 const w = huffWeight[n];
             U32 const length = (1 << w) >> 1;
-            U32 i;
+            U32 u;
             HUF_DEltX2 D;
             D.byte = (BYTE)n; D.nbBits = (BYTE)(tableLog + 1 - w);
-            for (i = rankVal[w]; i < rankVal[w] + length; i++)
-                dt[i] = D;
+            for (u = rankVal[w]; u < rankVal[w] + length; u++)
+                dt[u] = D;
             rankVal[w] += length;
     }   }
 
     return iSize;
 }
 
+size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_readDTableX2_wksp(DTable, src, srcSize,
+                                 workSpace, sizeof(workSpace));
+}
 
-static BYTE HUF_decodeSymbolX2(BIT_DStream_t* Dstream, const HUF_DEltX2* dt, const U32 dtLog)
+typedef struct { U16 sequence; BYTE nbBits; BYTE length; } HUF_DEltX4;  /* double-symbols decoding */
+
+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;
@@ -144,7 +155,7 @@
 #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)
 
@@ -152,30 +163,33 @@
     if (MEM_64bits()) \
         HUF_DECODE_SYMBOLX2_0(ptr, DStreamPtr)
 
-static 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)
@@ -188,47 +202,17 @@
     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 (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
-{
-    const BYTE* ip = (const BYTE*) cSrc;
-
-    size_t const hSize = HUF_readDTableX2 (DCtx, cSrc, cSrcSize);
-    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 (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)
@@ -263,23 +247,19 @@
         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);
@@ -296,10 +276,15 @@
             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);
@@ -312,8 +297,8 @@
         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;
@@ -321,6 +306,279 @@
 }
 
 
+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,
@@ -328,22 +586,38 @@
 {
     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);
+}
+
+static size_t HUF_decompress4X2_DCtx_wksp_bmi2(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize, int bmi2)
+{
+    const BYTE* ip = (const BYTE*) cSrc;
+
+    size_t const hSize = HUF_readDTableX2_wksp (dctx, cSrc, cSrcSize,
+                                                workSpace, wkspSize);
+    if (HUF_isError(hSize)) return hSize;
+    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
+    ip += hSize; cSrcSize -= hSize;
+
+    return HUF_decompress4X2_usingDTable_internal(dst, dstSize, ip, cSrcSize, dctx, 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);
 }
 
 
 size_t HUF_decompress4X2_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
-    const BYTE* ip = (const BYTE*) cSrc;
-
-    size_t const hSize = HUF_readDTableX2 (dctx, cSrc, cSrcSize);
-    if (HUF_isError(hSize)) return hSize;
-    if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
-    ip += hSize; cSrcSize -= hSize;
-
-    return HUF_decompress4X2_usingDTable_internal (dst, dstSize, ip, cSrcSize, dctx);
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
 }
-
 size_t HUF_decompress4X2 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
 {
     HUF_CREATE_STATIC_DTABLEX2(DTable, HUF_TABLELOG_MAX);
@@ -354,8 +628,6 @@
 /* *************************/
 /* 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() :
@@ -400,7 +672,8 @@
     }   }
 }
 
-typedef U32 rankVal_t[HUF_TABLELOG_MAX][HUF_TABLELOG_MAX + 1];
+typedef U32 rankValCol_t[HUF_TABLELOG_MAX + 1];
+typedef rankValCol_t rankVal_t[HUF_TABLELOG_MAX];
 
 static void HUF_fillDTableX4(HUF_DEltX4* DTable, const U32 targetLog,
                            const sortedSymbol_t* sortedList, const U32 sortedListSize,
@@ -444,22 +717,42 @@
     }
 }
 
-size_t HUF_readDTableX4 (HUF_DTable* DTable, const void* src, size_t srcSize)
+size_t HUF_readDTableX4_wksp(HUF_DTable* DTable, const void* src,
+                             size_t srcSize, void* workSpace,
+                             size_t wkspSize)
 {
-    BYTE weightList[HUF_SYMBOLVALUE_MAX + 1];
-    sortedSymbol_t sortedSymbol[HUF_SYMBOLVALUE_MAX + 1];
-    U32 rankStats[HUF_TABLELOG_MAX + 1] = { 0 };
-    U32 rankStart0[HUF_TABLELOG_MAX + 2] = { 0 };
-    U32* const rankStart = rankStart0+1;
-    rankVal_t rankVal;
     U32 tableLog, maxW, sizeOfSort, nbSymbols;
     DTableDesc dtd = HUF_getDTableDesc(DTable);
     U32 const maxTableLog = dtd.maxTableLog;
     size_t iSize;
     void* dtPtr = DTable+1;   /* force compiler to avoid strict-aliasing */
     HUF_DEltX4* const dt = (HUF_DEltX4*)dtPtr;
+    U32 *rankStart;
 
-    HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(HUF_DTable));   /* if compilation fails here, assertion is false */
+    rankValCol_t* rankVal;
+    U32* rankStats;
+    U32* rankStart0;
+    sortedSymbol_t* sortedSymbol;
+    BYTE* weightList;
+    size_t spaceUsed32 = 0;
+
+    rankVal = (rankValCol_t *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += (sizeof(rankValCol_t) * HUF_TABLELOG_MAX) >> 2;
+    rankStats = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_MAX + 1;
+    rankStart0 = (U32 *)workSpace + spaceUsed32;
+    spaceUsed32 += HUF_TABLELOG_MAX + 2;
+    sortedSymbol = (sortedSymbol_t *)workSpace + (spaceUsed32 * sizeof(U32)) / sizeof(sortedSymbol_t);
+    spaceUsed32 += HUF_ALIGN(sizeof(sortedSymbol_t) * (HUF_SYMBOLVALUE_MAX + 1), sizeof(U32)) >> 2;
+    weightList = (BYTE *)((U32 *)workSpace + spaceUsed32);
+    spaceUsed32 += HUF_ALIGN(HUF_SYMBOLVALUE_MAX + 1, sizeof(U32)) >> 2;
+
+    if ((spaceUsed32 << 2) > wkspSize) return ERROR(tableLog_tooLarge);
+
+    rankStart = rankStart0 + 1;
+    memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
+
+    HUF_STATIC_ASSERT(sizeof(HUF_DEltX4) == sizeof(HUF_DTable));   /* if compiler fails here, assertion is wrong */
     if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
     /* memset(weightList, 0, sizeof(weightList)); */  /* is not necessary, even though some analyzer complain ... */
 
@@ -524,93 +817,11 @@
     return iSize;
 }
 
-
-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))
-                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 */
-    }   }
-    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)
-
-static inline size_t HUF_decodeStreamX4(BYTE* p, BIT_DStream_t* bitDPtr, BYTE* const pEnd, const HUF_DEltX4* const dt, const U32 dtLog)
+size_t HUF_readDTableX4(HUF_DTable* DTable, const void* src, size_t srcSize)
 {
-    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;
+  U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+  return HUF_readDTableX4_wksp(DTable, src, srcSize,
+                               workSpace, sizeof(workSpace));
 }
 
 size_t HUF_decompress1X4_usingDTable(
@@ -620,19 +831,31 @@
 {
     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 (HUF_DTable* DCtx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X4_DCtx_wksp(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                                   const void* cSrc, size_t cSrcSize,
+                                   void* workSpace, size_t wkspSize)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t const hSize = HUF_readDTableX4 (DCtx, cSrc, cSrcSize);
+    size_t const hSize = HUF_readDTableX4_wksp(DCtx, cSrc, cSrcSize,
+                                               workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
 
-    return HUF_decompress1X4_usingDTable_internal (dst, dstSize, ip, cSrcSize, DCtx);
+    return HUF_decompress1X4_usingDTable_internal(dst, dstSize, ip, cSrcSize, DCtx, /* bmi2 */ 0);
+}
+
+
+size_t HUF_decompress1X4_DCtx(HUF_DTable* DCtx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X4_DCtx_wksp(DCtx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
 }
 
 size_t HUF_decompress1X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
@@ -641,99 +864,6 @@
     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,
@@ -741,20 +871,38 @@
 {
     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 (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+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, int bmi2)
 {
     const BYTE* ip = (const BYTE*) cSrc;
 
-    size_t hSize = HUF_readDTableX4 (dctx, cSrc, cSrcSize);
+    size_t hSize = HUF_readDTableX4_wksp(dctx, cSrc, cSrcSize,
+                                         workSpace, wkspSize);
     if (HUF_isError(hSize)) return hSize;
     if (hSize >= cSrcSize) return ERROR(srcSize_wrong);
     ip += hSize; cSrcSize -= hSize;
 
-    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);
+}
+
+
+size_t HUF_decompress4X4_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
+                              const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X4_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                       workSpace, sizeof(workSpace));
 }
 
 size_t HUF_decompress4X4 (void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
@@ -773,8 +921,8 @@
                                     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,
@@ -782,8 +930,8 @@
                                     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);
 }
 
 
@@ -810,21 +958,22 @@
 };
 
 /** 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 = (U32)(cSrcSize * 16 / dstSize);   /* Q < 16 since dstSize > cSrcSize */
-    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);
@@ -858,19 +1007,32 @@
     }
 }
 
-size_t HUF_decompress4X_hufOnly (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress4X_hufOnly(HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress4X_hufOnly_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                         workSpace, sizeof(workSpace));
+}
+
+
+size_t HUF_decompress4X_hufOnly_wksp(HUF_DTable* dctx, void* dst,
+                                     size_t dstSize, const void* cSrc,
+                                     size_t cSrcSize, void* workSpace,
+                                     size_t wkspSize)
 {
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
-    if ((cSrcSize >= dstSize) || (cSrcSize <= 1)) return ERROR(corruption_detected);   /* invalid */
+    if (cSrcSize == 0) return ERROR(corruption_detected);
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress4X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
-                        HUF_decompress4X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+        return algoNb ? HUF_decompress4X4_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize):
+                        HUF_decompress4X2_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize, workSpace, wkspSize);
     }
 }
 
-size_t HUF_decompress1X_DCtx (HUF_DTable* dctx, void* dst, size_t dstSize, const void* cSrc, size_t cSrcSize)
+size_t HUF_decompress1X_DCtx_wksp(HUF_DTable* dctx, void* dst, size_t dstSize,
+                                  const void* cSrc, size_t cSrcSize,
+                                  void* workSpace, size_t wkspSize)
 {
     /* validation checks */
     if (dstSize == 0) return ERROR(dstSize_tooSmall);
@@ -879,7 +1041,56 @@
     if (cSrcSize == 1) { memset(dst, *(const BYTE*)cSrc, dstSize); return dstSize; }   /* RLE */
 
     {   U32 const algoNb = HUF_selectDecoder(dstSize, cSrcSize);
-        return algoNb ? HUF_decompress1X4_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) :
-                        HUF_decompress1X2_DCtx(dctx, dst, dstSize, cSrc, cSrcSize) ;
+        return algoNb ? HUF_decompress1X4_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize):
+                        HUF_decompress1X2_DCtx_wksp(dctx, dst, dstSize, cSrc,
+                                cSrcSize, workSpace, wkspSize);
     }
 }
+
+size_t HUF_decompress1X_DCtx(HUF_DTable* dctx, void* dst, size_t dstSize,
+                             const void* cSrc, size_t cSrcSize)
+{
+    U32 workSpace[HUF_DECOMPRESS_WORKSPACE_SIZE_U32];
+    return HUF_decompress1X_DCtx_wksp(dctx, dst, dstSize, cSrc, cSrcSize,
+                                      workSpace, sizeof(workSpace));
+}
+
+
+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);
+    }
+}