zstd-ruby 1.4.9.0 → 1.5.0.0

data/ext/zstdruby/libzstd/compress/zstd_lazy.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -26,6 +26,7 @@ extern "C" {
 #define ZSTD_LAZY_DDSS_BUCKET_LOG 2
 
 U32 ZSTD_insertAndFindFirstIndex(ZSTD_matchState_t* ms, const BYTE* ip);
+void ZSTD_row_update(ZSTD_matchState_t* const ms, const BYTE* ip);
 
 void ZSTD_dedicatedDictSearch_lazy_loadDictionary(ZSTD_matchState_t* ms, const BYTE* const ip);
 
@@ -43,6 +44,15 @@ size_t ZSTD_compressBlock_lazy(
 size_t ZSTD_compressBlock_greedy(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
 
 size_t ZSTD_compressBlock_btlazy2_dictMatchState(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -56,6 +66,15 @@ size_t ZSTD_compressBlock_lazy_dictMatchState(
 size_t ZSTD_compressBlock_greedy_dictMatchState(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_dictMatchState_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_dictMatchState_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_dictMatchState_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
 
 size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -66,6 +85,15 @@ size_t ZSTD_compressBlock_lazy_dedicatedDictSearch(
 size_t ZSTD_compressBlock_greedy_dedicatedDictSearch(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_dedicatedDictSearch_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_dedicatedDictSearch_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_dedicatedDictSearch_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
 
 size_t ZSTD_compressBlock_greedy_extDict(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
@@ -76,9 +104,19 @@ size_t ZSTD_compressBlock_lazy_extDict(
 size_t ZSTD_compressBlock_lazy2_extDict(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_greedy_extDict_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy_extDict_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
+size_t ZSTD_compressBlock_lazy2_extDict_row(
+        ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+        void const* src, size_t srcSize);
 size_t ZSTD_compressBlock_btlazy2_extDict(
         ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
         void const* src, size_t srcSize);
+        
 
 #if defined (__cplusplus)
 }

data/ext/zstdruby/libzstd/compress/zstd_ldm.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -57,6 +57,33 @@ static void ZSTD_ldm_gear_init(ldmRollingHashState_t* state, ldmParams_t const*
     }
 }
 
+/** ZSTD_ldm_gear_reset()
+ * Feeds [data, data + minMatchLength) into the hash without registering any
+ * splits. This effectively resets the hash state. This is used when skipping
+ * over data, either at the beginning of a block, or skipping sections.
+ */
+static void ZSTD_ldm_gear_reset(ldmRollingHashState_t* state,
+                                BYTE const* data, size_t minMatchLength)
+{
+    U64 hash = state->rolling;
+    size_t n = 0;
+
+#define GEAR_ITER_ONCE() do {                                  \
+        hash = (hash << 1) + ZSTD_ldm_gearTab[data[n] & 0xff]; \
+        n += 1;                                                \
+    } while (0)
+    while (n + 3 < minMatchLength) {
+        GEAR_ITER_ONCE();
+        GEAR_ITER_ONCE();
+        GEAR_ITER_ONCE();
+        GEAR_ITER_ONCE();
+    }
+    while (n < minMatchLength) {
+        GEAR_ITER_ONCE();
+    }
+#undef GEAR_ITER_ONCE
+}
+
 /** ZSTD_ldm_gear_feed():
  *
  * Registers in the splits array all the split points found in the first
@@ -255,7 +282,7 @@ void ZSTD_ldm_fillHashTable(
     while (ip < iend) {
         size_t hashed;
         unsigned n;
-        
+
         numSplits = 0;
         hashed = ZSTD_ldm_gear_feed(&hashState, ip, iend - ip, splits, &numSplits);
 
@@ -327,16 +354,8 @@ static size_t ZSTD_ldm_generateSequences_internal(
 
     /* Initialize the rolling hash state with the first minMatchLength bytes */
     ZSTD_ldm_gear_init(&hashState, params);
-    {
-        size_t n = 0;
-
-        while (n < minMatchLength) {
-            numSplits = 0;
-            n += ZSTD_ldm_gear_feed(&hashState, ip + n, minMatchLength - n,
-                                    splits, &numSplits);
-        }
-        ip += minMatchLength;
-    }
+    ZSTD_ldm_gear_reset(&hashState, ip, minMatchLength);
+    ip += minMatchLength;
 
     while (ip < ilimit) {
         size_t hashed;
@@ -361,6 +380,7 @@ static size_t ZSTD_ldm_generateSequences_internal(
         for (n = 0; n < numSplits; n++) {
             size_t forwardMatchLength = 0, backwardMatchLength = 0,
                    bestMatchLength = 0, mLength;
+            U32 offset;
             BYTE const* const split = candidates[n].split;
             U32 const checksum = candidates[n].checksum;
             U32 const hash = candidates[n].hash;
@@ -428,9 +448,9 @@ static size_t ZSTD_ldm_generateSequences_internal(
             }
 
             /* Match found */
+            offset = (U32)(split - base) - bestEntry->offset;
             mLength = forwardMatchLength + backwardMatchLength;
             {
-                U32 const offset = (U32)(split - base) - bestEntry->offset;
                 rawSeq* const seq = rawSeqStore->seq + rawSeqStore->size;
 
                 /* Out of sequence storage */
@@ -447,6 +467,21 @@ static size_t ZSTD_ldm_generateSequences_internal(
             ZSTD_ldm_insertEntry(ldmState, hash, newEntry, *params);
 
             anchor = split + forwardMatchLength;
+
+            /* If we find a match that ends after the data that we've hashed
+             * then we have a repeating, overlapping, pattern. E.g. all zeros.
+             * If one repetition of the pattern matches our `stopMask` then all
+             * repetitions will. We don't need to insert them all into out table,
+             * only the first one. So skip over overlapping matches.
+             * This is a major speed boost (20x) for compressing a single byte
+             * repeated, when that byte ends up in the table.
+             */
+            if (anchor > ip + hashed) {
+                ZSTD_ldm_gear_reset(&hashState, anchor - minMatchLength, minMatchLength);
+                /* Continue the outter loop at anchor (ip + hashed == anchor). */
+                ip = anchor - hashed;
+                break;
+            }
         }
 
         ip += hashed;
@@ -500,7 +535,7 @@ size_t ZSTD_ldm_generateSequences(
 
         assert(chunkStart < iend);
         /* 1. Perform overflow correction if necessary. */
-        if (ZSTD_window_needOverflowCorrection(ldmState->window, chunkEnd)) {
+        if (ZSTD_window_needOverflowCorrection(ldmState->window, 0, maxDist, ldmState->loadedDictEnd, chunkStart, chunkEnd)) {
             U32 const ldmHSize = 1U << params->hashLog;
             U32 const correction = ZSTD_window_correctOverflow(
                 &ldmState->window, /* cycleLog */ 0, maxDist, chunkStart);
@@ -622,12 +657,13 @@ void ZSTD_ldm_skipRawSeqStoreBytes(rawSeqStore_t* rawSeqStore, size_t nbBytes) {
 
 size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
     ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+    ZSTD_useRowMatchFinderMode_e useRowMatchFinder,
     void const* src, size_t srcSize)
 {
     const ZSTD_compressionParameters* const cParams = &ms->cParams;
     unsigned const minMatch = cParams->minMatch;
     ZSTD_blockCompressor const blockCompressor =
-        ZSTD_selectBlockCompressor(cParams->strategy, ZSTD_matchState_dictMode(ms));
+        ZSTD_selectBlockCompressor(cParams->strategy, useRowMatchFinder, ZSTD_matchState_dictMode(ms));
     /* Input bounds */
     BYTE const* const istart = (BYTE const*)src;
     BYTE const* const iend = istart + srcSize;

data/ext/zstdruby/libzstd/compress/zstd_ldm.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -66,6 +66,7 @@ size_t ZSTD_ldm_generateSequences(
  */
 size_t ZSTD_ldm_blockCompress(rawSeqStore_t* rawSeqStore,
             ZSTD_matchState_t* ms, seqStore_t* seqStore, U32 rep[ZSTD_REP_NUM],
+            ZSTD_useRowMatchFinderMode_e useRowMatchFinder,
             void const* src, size_t srcSize);
 
 /**

data/ext/zstdruby/libzstd/compress/zstd_ldm_geartab.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the

data/ext/zstdruby/libzstd/compress/zstd_opt.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
+ * Copyright (c) Przemyslaw Skibinski, Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the

data/ext/zstdruby/libzstd/compress/zstd_opt.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the

data/ext/zstdruby/libzstd/compress/zstdmt_compress.c

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -512,7 +512,7 @@ ZSTDMT_serialState_reset(serialState_t* serialState,
         if (dictSize > 0) {
             if (dictContentType == ZSTD_dct_rawContent) {
                 BYTE const* const dictEnd = (const BYTE*)dict + dictSize;
-                ZSTD_window_update(&serialState->ldmState.window, dict, dictSize);
+                ZSTD_window_update(&serialState->ldmState.window, dict, dictSize, /* forceNonContiguous */ 0);
                 ZSTD_ldm_fillHashTable(&serialState->ldmState, (const BYTE*)dict, dictEnd, &params.ldmParams);
                 serialState->ldmState.loadedDictEnd = params.forceWindow ? 0 : (U32)(dictEnd - serialState->ldmState.window.base);
             } else {
@@ -569,7 +569,7 @@ static void ZSTDMT_serialState_update(serialState_t* serialState,
             assert(seqStore.seq != NULL && seqStore.pos == 0 &&
                    seqStore.size == 0 && seqStore.capacity > 0);
             assert(src.size <= serialState->params.jobSize);
-            ZSTD_window_update(&serialState->ldmState.window, src.start, src.size);
+            ZSTD_window_update(&serialState->ldmState.window, src.start, src.size, /* forceNonContiguous */ 0);
             error = ZSTD_ldm_generateSequences(
                 &serialState->ldmState, &seqStore,
                 &serialState->params.ldmParams, src.start, src.size);
@@ -695,6 +695,10 @@ static void ZSTDMT_compressionJob(void* jobDescription)
         {   size_t const forceWindowError = ZSTD_CCtxParams_setParameter(&jobParams, ZSTD_c_forceMaxWindow, !job->firstJob);
             if (ZSTD_isError(forceWindowError)) JOB_ERROR(forceWindowError);
         }
+        if (!job->firstJob) {
+            size_t const err = ZSTD_CCtxParams_setParameter(&jobParams, ZSTD_c_deterministicRefPrefix, 0);
+            if (ZSTD_isError(err)) JOB_ERROR(err);
+        }
         {   size_t const initError = ZSTD_compressBegin_advanced_internal(cctx,
                                         job->prefix.start, job->prefix.size, ZSTD_dct_rawContent, /* load dictionary in "content-only" mode (no header analysis) */
                                         ZSTD_dtlm_fast,
@@ -750,6 +754,12 @@ static void ZSTDMT_compressionJob(void* jobDescription)
             if (ZSTD_isError(cSize)) JOB_ERROR(cSize);
             lastCBlockSize = cSize;
     }   }
+    if (!job->firstJob) {
+        /* Double check that we don't have an ext-dict, because then our
+         * repcode invalidation doesn't work.
+         */
+        assert(!ZSTD_window_hasExtDict(cctx->blockState.matchState.window));
+    }
     ZSTD_CCtx_trace(cctx, 0);
 
 _endJob:
@@ -1240,9 +1250,8 @@ size_t ZSTDMT_initCStream_internal(
 
     if (params.rsyncable) {
         /* Aim for the targetsectionSize as the average job size. */
-        U32 const jobSizeMB = (U32)(mtctx->targetSectionSize >> 20);
-        U32 const rsyncBits = ZSTD_highbit32(jobSizeMB) + 20;
-        assert(jobSizeMB >= 1);
+        U32 const jobSizeKB = (U32)(mtctx->targetSectionSize >> 10);
+        U32 const rsyncBits = (assert(jobSizeKB >= 1), ZSTD_highbit32(jobSizeKB) + 10);
         DEBUGLOG(4, "rsyncLog = %u", rsyncBits);
         mtctx->rsync.hash = 0;
         mtctx->rsync.hitMask = (1ULL << rsyncBits) - 1;

data/ext/zstdruby/libzstd/compress/zstdmt_compress.h

@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2016-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  * All rights reserved.
  *
  * This source code is licensed under both the BSD-style license (found in the
@@ -32,11 +32,11 @@
 
 
 /* ===   Constants   === */
-#ifndef ZSTDMT_NBWORKERS_MAX
-#  define ZSTDMT_NBWORKERS_MAX 200
+#ifndef ZSTDMT_NBWORKERS_MAX /* a different value can be selected at compile time */
+#  define ZSTDMT_NBWORKERS_MAX ((sizeof(void*)==4) /*32-bit*/ ? 64 : 256)
 #endif
-#ifndef ZSTDMT_JOBSIZE_MIN
-#  define ZSTDMT_JOBSIZE_MIN (1 MB)
+#ifndef ZSTDMT_JOBSIZE_MIN   /* a different value can be selected at compile time */
+#  define ZSTDMT_JOBSIZE_MIN (512 KB)
 #endif
 #define ZSTDMT_JOBLOG_MAX   (MEM_32bits() ? 29 : 30)
 #define ZSTDMT_JOBSIZE_MAX  (MEM_32bits() ? (512 MB) : (1024 MB))

data/ext/zstdruby/libzstd/decompress/huf_decompress.c

@@ -1,7 +1,7 @@
 /* ******************************************************************
  * huff0 huffman decoder,
  * part of Finite State Entropy library
- * Copyright (c) 2013-2021, Yann Collet, Facebook, Inc.
+ * Copyright (c) Yann Collet, Facebook, Inc.
  *
  *  You can contact the author at :
  *  - FSE+HUF source repository : https://github.com/Cyan4973/FiniteStateEntropy
@@ -528,13 +528,15 @@ typedef rankValCol_t rankVal_t[HUF_TABLELOG_MAX];
 static void HUF_fillDTableX2Level2(HUF_DEltX2* DTable, U32 sizeLog, const U32 consumed,
                            const U32* rankValOrigin, const int minWeight,
                            const sortedSymbol_t* sortedSymbols, const U32 sortedListSize,
-                           U32 nbBitsBaseline, U16 baseSeq)
+                           U32 nbBitsBaseline, U16 baseSeq, U32* wksp, size_t wkspSize)
 {
     HUF_DEltX2 DElt;
-    U32 rankVal[HUF_TABLELOG_MAX + 1];
+    U32* rankVal = wksp;
 
+    assert(wkspSize >= HUF_TABLELOG_MAX + 1);
+    (void)wkspSize;
     /* get pre-calculated rankVal */
-    ZSTD_memcpy(rankVal, rankValOrigin, sizeof(rankVal));
+    ZSTD_memcpy(rankVal, rankValOrigin, sizeof(U32) * (HUF_TABLELOG_MAX + 1));
 
     /* fill skipped values */
     if (minWeight>1) {
@@ -569,14 +571,18 @@ static void HUF_fillDTableX2Level2(HUF_DEltX2* DTable, U32 sizeLog, const U32 co
 static void HUF_fillDTableX2(HUF_DEltX2* DTable, const U32 targetLog,
                            const sortedSymbol_t* sortedList, const U32 sortedListSize,
                            const U32* rankStart, rankVal_t rankValOrigin, const U32 maxWeight,
-                           const U32 nbBitsBaseline)
+                           const U32 nbBitsBaseline, U32* wksp, size_t wkspSize)
 {
-    U32 rankVal[HUF_TABLELOG_MAX + 1];
+    U32* rankVal = wksp;
     const int scaleLog = nbBitsBaseline - targetLog;   /* note : targetLog >= srcLog, hence scaleLog <= 1 */
     const U32 minBits  = nbBitsBaseline - maxWeight;
     U32 s;
 
-    ZSTD_memcpy(rankVal, rankValOrigin, sizeof(rankVal));
+    assert(wkspSize >= HUF_TABLELOG_MAX + 1);
+    wksp += HUF_TABLELOG_MAX + 1;
+    wkspSize -= HUF_TABLELOG_MAX + 1;
+
+    ZSTD_memcpy(rankVal, rankValOrigin, sizeof(U32) * (HUF_TABLELOG_MAX + 1));
 
     /* fill DTable */
     for (s=0; s<sortedListSize; s++) {
@@ -594,7 +600,7 @@ static void HUF_fillDTableX2(HUF_DEltX2* DTable, const U32 targetLog,
             HUF_fillDTableX2Level2(DTable+start, targetLog-nbBits, nbBits,
                            rankValOrigin[nbBits], minWeight,
                            sortedList+sortedRank, sortedListSize-sortedRank,
-                           nbBitsBaseline, symbol);
+                           nbBitsBaseline, symbol, wksp, wkspSize);
         } else {
             HUF_DEltX2 DElt;
             MEM_writeLE16(&(DElt.sequence), symbol);
@@ -608,6 +614,15 @@ static void HUF_fillDTableX2(HUF_DEltX2* DTable, const U32 targetLog,
     }
 }
 
+typedef struct {
+    rankValCol_t rankVal[HUF_TABLELOG_MAX];
+    U32 rankStats[HUF_TABLELOG_MAX + 1];
+    U32 rankStart0[HUF_TABLELOG_MAX + 2];
+    sortedSymbol_t sortedSymbol[HUF_SYMBOLVALUE_MAX + 1];
+    BYTE weightList[HUF_SYMBOLVALUE_MAX + 1];
+    U32 calleeWksp[HUF_READ_STATS_WORKSPACE_SIZE_U32];
+} HUF_ReadDTableX2_Workspace;
+
 size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
                        const void* src, size_t srcSize,
                              void* workSpace, size_t wkspSize)
@@ -620,47 +635,32 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
     HUF_DEltX2* const dt = (HUF_DEltX2*)dtPtr;
     U32 *rankStart;
 
-    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;
-    ZSTD_memset(rankStats, 0, sizeof(U32) * (2 * HUF_TABLELOG_MAX + 2 + 1));
+    HUF_ReadDTableX2_Workspace* const wksp = (HUF_ReadDTableX2_Workspace*)workSpace;
+
+    if (sizeof(*wksp) > wkspSize) return ERROR(GENERIC);
+
+    rankStart = wksp->rankStart0 + 1;
+    ZSTD_memset(wksp->rankStats, 0, sizeof(wksp->rankStats));
+    ZSTD_memset(wksp->rankStart0, 0, sizeof(wksp->rankStart0));
 
     DEBUG_STATIC_ASSERT(sizeof(HUF_DEltX2) == sizeof(HUF_DTable));   /* if compiler fails here, assertion is wrong */
     if (maxTableLog > HUF_TABLELOG_MAX) return ERROR(tableLog_tooLarge);
     /* ZSTD_memset(weightList, 0, sizeof(weightList)); */  /* is not necessary, even though some analyzer complain ... */
 
-    iSize = HUF_readStats(weightList, HUF_SYMBOLVALUE_MAX + 1, rankStats, &nbSymbols, &tableLog, src, srcSize);
+    iSize = HUF_readStats_wksp(wksp->weightList, HUF_SYMBOLVALUE_MAX + 1, wksp->rankStats, &nbSymbols, &tableLog, src, srcSize, wksp->calleeWksp, sizeof(wksp->calleeWksp), /* bmi2 */ 0);
     if (HUF_isError(iSize)) return iSize;
 
     /* check result */
     if (tableLog > maxTableLog) return ERROR(tableLog_tooLarge);   /* DTable can't fit code depth */
 
     /* find maxWeight */
-    for (maxW = tableLog; rankStats[maxW]==0; maxW--) {}  /* necessarily finds a solution before 0 */
+    for (maxW = tableLog; wksp->rankStats[maxW]==0; maxW--) {}  /* necessarily finds a solution before 0 */
 
     /* Get start index of each weight */
     {   U32 w, nextRankStart = 0;
         for (w=1; w<maxW+1; w++) {
             U32 curr = nextRankStart;
-            nextRankStart += rankStats[w];
+            nextRankStart += wksp->rankStats[w];
             rankStart[w] = curr;
         }
         rankStart[0] = nextRankStart;   /* put all 0w symbols at the end of sorted list*/
@@ -670,37 +670,38 @@ size_t HUF_readDTableX2_wksp(HUF_DTable* DTable,
     /* sort symbols by weight */
     {   U32 s;
         for (s=0; s<nbSymbols; s++) {
-            U32 const w = weightList[s];
+            U32 const w = wksp->weightList[s];
             U32 const r = rankStart[w]++;
-            sortedSymbol[r].symbol = (BYTE)s;
-            sortedSymbol[r].weight = (BYTE)w;
+            wksp->sortedSymbol[r].symbol = (BYTE)s;
+            wksp->sortedSymbol[r].weight = (BYTE)w;
         }
         rankStart[0] = 0;   /* forget 0w symbols; this is beginning of weight(1) */
     }
 
     /* Build rankVal */
-    {   U32* const rankVal0 = rankVal[0];
+    {   U32* const rankVal0 = wksp->rankVal[0];
         {   int const rescale = (maxTableLog-tableLog) - 1;   /* tableLog <= maxTableLog */
             U32 nextRankVal = 0;
             U32 w;
             for (w=1; w<maxW+1; w++) {
                 U32 curr = nextRankVal;
-                nextRankVal += rankStats[w] << (w+rescale);
+                nextRankVal += wksp->rankStats[w] << (w+rescale);
                 rankVal0[w] = curr;
         }   }
         {   U32 const minBits = tableLog+1 - maxW;
             U32 consumed;
             for (consumed = minBits; consumed < maxTableLog - minBits + 1; consumed++) {
-                U32* const rankValPtr = rankVal[consumed];
+                U32* const rankValPtr = wksp->rankVal[consumed];
                 U32 w;
                 for (w = 1; w < maxW+1; w++) {
                     rankValPtr[w] = rankVal0[w] >> consumed;
     }   }   }   }
 
     HUF_fillDTableX2(dt, maxTableLog,
-                   sortedSymbol, sizeOfSort,
-                   rankStart0, rankVal, maxW,
-                   tableLog+1);
+                   wksp->sortedSymbol, sizeOfSort,
+                   wksp->rankStart0, wksp->rankVal, maxW,
+                   tableLog+1,
+                   wksp->calleeWksp, sizeof(wksp->calleeWksp) / sizeof(U32));
 
     dtd.tableLog = (BYTE)maxTableLog;
     dtd.tableType = 1;
@@ -1225,7 +1226,7 @@ size_t HUF_decompress1X1 (void* dst, size_t dstSize, const void* cSrc, size_t cS
     HUF_CREATE_STATIC_DTABLEX1(DTable, HUF_TABLELOG_MAX);
     return HUF_decompress1X1_DCtx (DTable, dst, dstSize, cSrc, cSrcSize);
 }
-#endif 
+#endif
 
 #ifndef HUF_FORCE_DECOMPRESS_X1
 size_t HUF_readDTableX2(HUF_DTable* DTable, const void* src, size_t srcSize)