Merge pull request #1169 from felixhandte/zstd-lazy-in-place-dict

ZSTD_lazy: Support Searching the Dictionary Context In-Place

lib/compress/zstd_compress.c

@@ -1233,18 +1233,18 @@ static size_t ZSTD_resetCCtx_usingCDict(ZSTD_CCtx* cctx,
         8 KB, /* unused */
         8 KB, /* ZSTD_fast */
         16 KB, /* ZSTD_dfast */
-        16 KB, /* ZSTD_greedy */
+        32 KB, /* ZSTD_greedy */
-        16 KB, /* ZSTD_lazy */
+        32 KB, /* ZSTD_lazy */
-        16 KB, /* ZSTD_lazy2 */
+        32 KB, /* ZSTD_lazy2 */
-        16 KB, /* ZSTD_btlazy2 */
+        256 KB, /* ZSTD_btlazy2 */
-        16 KB, /* ZSTD_btopt */
+        256 KB, /* ZSTD_btopt */
-        16 KB /* ZSTD_btultra */
+        256 KB /* ZSTD_btultra */
     };
     const int attachDict = ( pledgedSrcSize <= attachDictSizeCutoffs[cdict->cParams.strategy]
                           || pledgedSrcSize == ZSTD_CONTENTSIZE_UNKNOWN )
                         && !params.forceWindow /* dictMatchState isn't correctly
                                                 * handled in _enforceMaxDist */
-                        && cdict->cParams.strategy <= ZSTD_dfast
+                        && cdict->cParams.strategy <= ZSTD_lazy2
                         && ZSTD_equivalentCParams(cctx->appliedParams.cParams,
                                                   cdict->cParams);
 
@@ -2230,17 +2230,31 @@ ZSTD_blockCompressor ZSTD_selectBlockCompressor(ZSTD_strategy strat, ZSTD_dictMo
 {
     static const ZSTD_blockCompressor blockCompressor[3][(unsigned)ZSTD_btultra+1] = {
         { ZSTD_compressBlock_fast  /* default for 0 */,
-          ZSTD_compressBlock_fast, ZSTD_compressBlock_doubleFast, ZSTD_compressBlock_greedy,
+          ZSTD_compressBlock_fast,
-          ZSTD_compressBlock_lazy, ZSTD_compressBlock_lazy2, ZSTD_compressBlock_btlazy2,
+          ZSTD_compressBlock_doubleFast,
-          ZSTD_compressBlock_btopt, ZSTD_compressBlock_btultra },
+          ZSTD_compressBlock_greedy,
+          ZSTD_compressBlock_lazy,
+          ZSTD_compressBlock_lazy2,
+          ZSTD_compressBlock_btlazy2,
+          ZSTD_compressBlock_btopt,
+          ZSTD_compressBlock_btultra },
         { ZSTD_compressBlock_fast_extDict  /* default for 0 */,
-          ZSTD_compressBlock_fast_extDict, ZSTD_compressBlock_doubleFast_extDict, ZSTD_compressBlock_greedy_extDict,
+          ZSTD_compressBlock_fast_extDict,
-          ZSTD_compressBlock_lazy_extDict,ZSTD_compressBlock_lazy2_extDict, ZSTD_compressBlock_btlazy2_extDict,
+          ZSTD_compressBlock_doubleFast_extDict,
-          ZSTD_compressBlock_btopt_extDict, ZSTD_compressBlock_btultra_extDict },
+          ZSTD_compressBlock_greedy_extDict,
+          ZSTD_compressBlock_lazy_extDict,
+          ZSTD_compressBlock_lazy2_extDict,
+          ZSTD_compressBlock_btlazy2_extDict,
+          ZSTD_compressBlock_btopt_extDict,
+          ZSTD_compressBlock_btultra_extDict },
         { ZSTD_compressBlock_fast_dictMatchState  /* default for 0 */,
           ZSTD_compressBlock_fast_dictMatchState,
           ZSTD_compressBlock_doubleFast_dictMatchState,
-          NULL, NULL, NULL, NULL, NULL, NULL /* unimplemented as of yet */ }
+          ZSTD_compressBlock_greedy_dictMatchState,
+          ZSTD_compressBlock_lazy_dictMatchState,
+          ZSTD_compressBlock_lazy2_dictMatchState,
+          ZSTD_compressBlock_btlazy2_dictMatchState,
+          NULL, NULL /* unimplemented as of yet */ }
     };
     ZSTD_blockCompressor selectedCompressor;
     ZSTD_STATIC_ASSERT((unsigned)ZSTD_fast == 1);

lib/compress/zstd_lazy.c

@@ -62,7 +62,7 @@ void ZSTD_updateDUBT(
 static void ZSTD_insertDUBT1(
                  ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                  U32 current, const BYTE* inputEnd,
-                 U32 nbCompares, U32 btLow, int extDict)
+                 U32 nbCompares, U32 btLow, const ZSTD_dictMode_e dictMode)
 {
     U32*   const bt = ms->chainTable;
     U32    const btLog  = cParams->chainLog - 1;
@@ -92,10 +92,12 @@ static void ZSTD_insertDUBT1(
         size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
         assert(matchIndex < current);
 
-        if ( (!extDict)
+        if ( (dictMode != ZSTD_extDict)
           || (matchIndex+matchLength >= dictLimit)  /* both in current segment*/
           || (current < dictLimit) /* both in extDict */) {
-            const BYTE* const mBase = !extDict || ((matchIndex+matchLength) >= dictLimit) ? base : dictBase;
+            const BYTE* const mBase = ( (dictMode != ZSTD_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;
@@ -143,7 +145,7 @@ static size_t ZSTD_DUBT_findBestMatch (
                             const BYTE* const ip, const BYTE* const iend,
                             size_t* offsetPtr,
                             U32 const mls,
-                            U32 const extDict)
+                            const ZSTD_dictMode_e dictMode)
 {
     U32*   const hashTable = ms->hashTable;
     U32    const hashLog = cParams->hashLog;
@@ -196,7 +198,7 @@ static size_t ZSTD_DUBT_findBestMatch (
         U32* const nextCandidateIdxPtr = bt + 2*(matchIndex&btMask) + 1;
         U32 const nextCandidateIdx = *nextCandidateIdxPtr;
         ZSTD_insertDUBT1(ms, cParams, matchIndex, iend,
-                         nbCandidates, unsortLimit, extDict);
+                         nbCandidates, unsortLimit, dictMode);
         matchIndex = nextCandidateIdx;
         nbCandidates++;
     }
@@ -221,7 +223,7 @@ static size_t ZSTD_DUBT_findBestMatch (
             size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
             const BYTE* match;
 
-            if ((!extDict) || (matchIndex+matchLength >= dictLimit)) {
+            if ((dictMode != ZSTD_extDict) || (matchIndex+matchLength >= dictLimit)) {
                 match = base + matchIndex;
                 matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
             } else {
@@ -276,12 +278,13 @@ static size_t ZSTD_BtFindBestMatch (
                         ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* const ip, const BYTE* const iLimit,
                         size_t* offsetPtr,
-                        const U32 mls /* template */)
+                        const U32 mls /* template */,
+                        const ZSTD_dictMode_e dictMode)
 {
     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);
+    return ZSTD_DUBT_findBestMatch(ms, cParams, ip, iLimit, offsetPtr, mls, dictMode);
 }
 
 
@@ -293,29 +296,15 @@ static size_t ZSTD_BtFindBestMatch_selectMLS (
     switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4);
+    case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
-    case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5);
+    case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
     case 7 :
-    case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6);
+    case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
     }
 }
 
 
-/** Tree updater, providing best match */
+static size_t ZSTD_BtFindBestMatch_dictMatchState_selectMLS (
-static size_t ZSTD_BtFindBestMatch_extDict (
-                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
-                        const BYTE* const ip, const BYTE* const iLimit,
-                        size_t* offsetPtr,
-                        const U32 mls)
-{
-    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_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
                         size_t* offsetPtr)
@@ -323,10 +312,26 @@ static size_t ZSTD_BtFindBestMatch_selectMLS_extDict (
     switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 4);
+    case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
-    case 5 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 5);
+    case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
     case 7 :
-    case 6 : return ZSTD_BtFindBestMatch_extDict(ms, cParams, ip, iLimit, offsetPtr, 6);
+    case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
+    }
+}
+
+
+static size_t ZSTD_BtFindBestMatch_extDict_selectMLS (
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr)
+{
+    switch(cParams->searchLength)
+    {
+    default : /* includes case 3 */
+    case 4 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
+    case 5 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
+    case 7 :
+    case 6 : return ZSTD_BtFindBestMatch(ms, cParams, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
     }
 }
 
@@ -376,7 +381,7 @@ size_t ZSTD_HcFindBestMatch_generic (
                         ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* const ip, const BYTE* const iLimit,
                         size_t* offsetPtr,
-                        const U32 mls, const U32 extDict)
+                        const U32 mls, const ZSTD_dictMode_e dictMode)
 {
     U32* const chainTable = ms->chainTable;
     const U32 chainSize = (1 << cParams->chainLog);
@@ -397,7 +402,7 @@ size_t ZSTD_HcFindBestMatch_generic (
 
     for ( ; (matchIndex>lowLimit) & (nbAttempts>0) ; nbAttempts--) {
         size_t currentMl=0;
-        if ((!extDict) || matchIndex >= dictLimit) {
+        if ((dictMode != ZSTD_extDict) || matchIndex >= dictLimit) {
             const BYTE* const match = base + matchIndex;
             if (match[ml] == ip[ml])   /* potentially better */
                 currentMl = ZSTD_count(ip, match, iLimit);
@@ -419,6 +424,37 @@ size_t ZSTD_HcFindBestMatch_generic (
         matchIndex = NEXT_IN_CHAIN(matchIndex, chainMask);
     }
 
+    if (dictMode == ZSTD_dictMatchState) {
+        const ZSTD_matchState_t* const dms = ms->dictMatchState;
+        const U32* const dmsChainTable = dms->chainTable;
+        const U32 dmsLowestIndex       = dms->window.dictLimit;
+        const BYTE* const dmsBase      = dms->window.base;
+        const BYTE* const dmsEnd       = dms->window.nextSrc;
+        const U32 dmsSize              = (U32)(dmsEnd - dmsBase);
+        const U32 dmsIndexDelta        = dictLimit - dmsSize;
+        const U32 dmsMinChain = dmsSize > chainSize ? dmsSize - chainSize : 0;
+
+        matchIndex = dms->hashTable[ZSTD_hashPtr(ip, cParams->hashLog, mls)];
+
+        for ( ; (matchIndex>dmsLowestIndex) & (nbAttempts>0) ; nbAttempts--) {
+            size_t currentMl=0;
+            const BYTE* const match = dmsBase + matchIndex;
+            assert(match+4 <= dmsEnd);
+            if (MEM_read32(match) == MEM_read32(ip))   /* assumption : matchIndex <= dictLimit-4 (by table construction) */
+                currentMl = ZSTD_count_2segments(ip+4, match+4, iLimit, dmsEnd, prefixStart) + 4;
+
+            /* save best solution */
+            if (currentMl > ml) {
+                ml = currentMl;
+                *offsetPtr = current - (matchIndex + dmsIndexDelta) + ZSTD_REP_MOVE;
+                if (ip+currentMl == iLimit) break; /* best possible, avoids read overflow on next attempt */
+            }
+
+            if (matchIndex <= dmsMinChain) break;
+            matchIndex = dmsChainTable[matchIndex & chainMask];
+        }
+    }
+
     return ml;
 }
 
@@ -431,10 +467,26 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
     switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 0);
+    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, ZSTD_noDict);
-    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 0);
+    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, ZSTD_noDict);
     case 7 :
-    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 0);
+    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, ZSTD_noDict);
+    }
+}
+
+
+static size_t ZSTD_HcFindBestMatch_dictMatchState_selectMLS (
+                        ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr)
+{
+    switch(cParams->searchLength)
+    {
+    default : /* includes case 3 */
+    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, ZSTD_dictMatchState);
+    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, ZSTD_dictMatchState);
+    case 7 :
+    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, ZSTD_dictMatchState);
     }
 }
 
@@ -442,15 +494,15 @@ FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_selectMLS (
 FORCE_INLINE_TEMPLATE size_t ZSTD_HcFindBestMatch_extDict_selectMLS (
                         ZSTD_matchState_t* ms, ZSTD_compressionParameters const* cParams,
                         const BYTE* ip, const BYTE* const iLimit,
-                        size_t* const offsetPtr)
+                        size_t* offsetPtr)
 {
     switch(cParams->searchLength)
     {
     default : /* includes case 3 */
-    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, 1);
+    case 4 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 4, ZSTD_extDict);
-    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, 1);
+    case 5 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 5, ZSTD_extDict);
     case 7 :
-    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, 1);
+    case 6 : return ZSTD_HcFindBestMatch_generic(ms, cParams, ip, iLimit, offsetPtr, 6, ZSTD_extDict);
     }
 }
 
@@ -464,28 +516,56 @@ size_t ZSTD_compressBlock_lazy_generic(
                         U32 rep[ZSTD_REP_NUM],
                         ZSTD_compressionParameters const* cParams,
                         const void* src, size_t srcSize,
-                        const U32 searchMethod, const U32 depth)
+                        const U32 searchMethod, const U32 depth,
+                        ZSTD_dictMode_e const dictMode)
 {
     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 = ms->window.base + ms->window.dictLimit;
+    const BYTE* const base = ms->window.base;
+    const U32 prefixLowestIndex = ms->window.dictLimit;
+    const BYTE* const prefixLowest = base + prefixLowestIndex;
 
     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;
+    searchMax_f const searchMax = dictMode == ZSTD_dictMatchState ?
+        (searchMethod ? ZSTD_BtFindBestMatch_dictMatchState_selectMLS : ZSTD_HcFindBestMatch_dictMatchState_selectMLS) :
+        (searchMethod ? ZSTD_BtFindBestMatch_selectMLS : ZSTD_HcFindBestMatch_selectMLS);
     U32 offset_1 = rep[0], offset_2 = rep[1], savedOffset=0;
 
+    const ZSTD_matchState_t* const dms = ms->dictMatchState;
+    const U32 dictLowestIndex      = dictMode == ZSTD_dictMatchState ?
+                                     dms->window.dictLimit : 0;
+    const BYTE* const dictBase     = dictMode == ZSTD_dictMatchState ?
+                                     dms->window.base : NULL;
+    const BYTE* const dictLowest   = dictMode == ZSTD_dictMatchState ?
+                                     dictBase + dictLowestIndex : NULL;
+    const BYTE* const dictEnd      = dictMode == ZSTD_dictMatchState ?
+                                     dms->window.nextSrc : NULL;
+    const U32 dictIndexDelta       = dictMode == ZSTD_dictMatchState ?
+                                     prefixLowestIndex - (U32)(dictEnd - dictBase) :
+                                     0;
+    const U32 dictAndPrefixLength = (U32)(ip - prefixLowest + dictEnd - dictLowest);
+
+    (void)dictMode;
+
     /* init */
-    ip += (ip==base);
+    ip += (dictAndPrefixLength == 0);
     ms->nextToUpdate3 = ms->nextToUpdate;
-    {   U32 const maxRep = (U32)(ip-base);
+    if (dictMode == ZSTD_noDict) {
+        U32 const maxRep = (U32)(ip - prefixLowest);
         if (offset_2 > maxRep) savedOffset = offset_2, offset_2 = 0;
         if (offset_1 > maxRep) savedOffset = offset_1, offset_1 = 0;
     }
+    if (dictMode == ZSTD_dictMatchState) {
+        /* dictMatchState repCode checks don't currently handle repCode == 0
+         * disabling. */
+        assert(offset_1 <= dictAndPrefixLength);
+        assert(offset_2 <= dictAndPrefixLength);
+    }
 
     /* Match Loop */
     while (ip < ilimit) {
@@ -494,8 +574,21 @@ size_t ZSTD_compressBlock_lazy_generic(
         const BYTE* start=ip+1;
 
         /* check repCode */
-        if ((offset_1>0) & (MEM_read32(ip+1) == MEM_read32(ip+1 - offset_1))) {
+        if (dictMode == ZSTD_dictMatchState) {
-            /* repcode : we take it */
+            const U32 repIndex = (U32)(ip - base) + 1 - offset_1;
+            const BYTE* repMatch = (dictMode == ZSTD_dictMatchState
+                                && repIndex < prefixLowestIndex) ?
+                                   dictBase + (repIndex - dictIndexDelta) :
+                                   base + repIndex;
+            if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+                && (MEM_read32(repMatch) == MEM_read32(ip+1)) ) {
+                const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+                matchLength = ZSTD_count_2segments(ip+1+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+                if (depth==0) goto _storeSequence;
+            }
+        }
+        if ( dictMode == ZSTD_noDict
+          && ((offset_1 > 0) & (MEM_read32(ip+1-offset_1) == MEM_read32(ip+1)))) {
             matchLength = ZSTD_count(ip+1+4, ip+1+4-offset_1, iend) + 4;
             if (depth==0) goto _storeSequence;
         }
@@ -516,13 +609,29 @@ size_t ZSTD_compressBlock_lazy_generic(
         if (depth>=1)
         while (ip<ilimit) {
             ip ++;
-            if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+            if ( (dictMode == ZSTD_noDict)
+              && (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
                 size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
                 int const gain2 = (int)(mlRep * 3);
                 int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
                 if ((mlRep >= 4) && (gain2 > gain1))
                     matchLength = mlRep, offset = 0, start = ip;
             }
+            if (dictMode == ZSTD_dictMatchState) {
+                const U32 repIndex = (U32)(ip - base) - offset_1;
+                const BYTE* repMatch = repIndex < prefixLowestIndex ?
+                               dictBase + (repIndex - dictIndexDelta) :
+                               base + repIndex;
+                if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+                    && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
+                    const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+                    size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+                    int const gain2 = (int)(mlRep * 3);
+                    int const gain1 = (int)(matchLength*3 - ZSTD_highbit32((U32)offset+1) + 1);
+                    if ((mlRep >= 4) && (gain2 > gain1))
+                        matchLength = mlRep, offset = 0, start = ip;
+                }
+            }
             {   size_t offset2=99999999;
                 size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
                 int const gain2 = (int)(ml2*4 - ZSTD_highbit32((U32)offset2+1));   /* raw approx */
@@ -535,12 +644,28 @@ size_t ZSTD_compressBlock_lazy_generic(
             /* let's find an even better one */
             if ((depth==2) && (ip<ilimit)) {
                 ip ++;
-                if ((offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
+                if ( (dictMode == ZSTD_noDict)
-                    size_t const ml2 = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
+                  && (offset) && ((offset_1>0) & (MEM_read32(ip) == MEM_read32(ip - offset_1)))) {
-                    int const gain2 = (int)(ml2 * 4);
+                    size_t const mlRep = ZSTD_count(ip+4, ip+4-offset_1, iend) + 4;
+                    int const gain2 = (int)(mlRep * 4);
                     int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
-                    if ((ml2 >= 4) && (gain2 > gain1))
+                    if ((mlRep >= 4) && (gain2 > gain1))
-                        matchLength = ml2, offset = 0, start = ip;
+                        matchLength = mlRep, offset = 0, start = ip;
+                }
+                if (dictMode == ZSTD_dictMatchState) {
+                    const U32 repIndex = (U32)(ip - base) - offset_1;
+                    const BYTE* repMatch = repIndex < prefixLowestIndex ?
+                                   dictBase + (repIndex - dictIndexDelta) :
+                                   base + repIndex;
+                    if (((U32)((prefixLowestIndex-1) - repIndex) >= 3 /* intentional underflow */)
+                        && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
+                        const BYTE* repMatchEnd = repIndex < prefixLowestIndex ? dictEnd : iend;
+                        size_t const mlRep = ZSTD_count_2segments(ip+4, repMatch+4, iend, repMatchEnd, prefixLowest) + 4;
+                        int const gain2 = (int)(mlRep * 4);
+                        int const gain1 = (int)(matchLength*4 - ZSTD_highbit32((U32)offset+1) + 1);
+                        if ((mlRep >= 4) && (gain2 > gain1))
+                            matchLength = mlRep, offset = 0, start = ip;
+                    }
                 }
                 {   size_t offset2=99999999;
                     size_t const ml2 = searchMax(ms, cParams, ip, iend, &offset2);
@@ -560,9 +685,17 @@ size_t ZSTD_compressBlock_lazy_generic(
          */
         /* catch up */
         if (offset) {
-            while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > base))
+            if (dictMode == ZSTD_noDict) {
+                while ( ((start > anchor) & (start - (offset-ZSTD_REP_MOVE) > prefixLowest))
                      && (start[-1] == (start-(offset-ZSTD_REP_MOVE))[-1]) )  /* only search for offset within prefix */
                     { start--; matchLength++; }
+            }
+            if (dictMode == ZSTD_dictMatchState) {
+                U32 const matchIndex = (U32)((start-base) - (offset - ZSTD_REP_MOVE));
+                const BYTE* match = (matchIndex < prefixLowestIndex) ? dictBase + matchIndex - dictIndexDelta : base + matchIndex;
+                const BYTE* const mStart = (matchIndex < prefixLowestIndex) ? dictLowest : prefixLowest;
+                while ((start>anchor) && (match>mStart) && (start[-1] == match[-1])) { start--; match--; matchLength++; }  /* catch up */
+            }
             offset_2 = offset_1; offset_1 = (U32)(offset - ZSTD_REP_MOVE);
         }
         /* store sequence */
@@ -573,6 +706,29 @@ _storeSequence:
         }
 
         /* check immediate repcode */
+        if (dictMode == ZSTD_dictMatchState) {
+            while (ip <= ilimit) {
+                U32 const current2 = (U32)(ip-base);
+                U32 const repIndex = current2 - offset_2;
+                const BYTE* repMatch = dictMode == ZSTD_dictMatchState
+                    && repIndex < prefixLowestIndex ?
+                        dictBase - dictIndexDelta + repIndex :
+                        base + repIndex;
+                if ( ((U32)((prefixLowestIndex-1) - (U32)repIndex) >= 3 /* intentional overflow */)
+                   && (MEM_read32(repMatch) == MEM_read32(ip)) ) {
+                    const BYTE* const repEnd2 = repIndex < prefixLowestIndex ? dictEnd : iend;
+                    matchLength = ZSTD_count_2segments(ip+4, repMatch+4, iend, repEnd2, prefixLowest) + 4;
+                    offset = offset_2; offset_2 = offset_1; offset_1 = (U32)offset;   /* swap offset_2 <=> offset_1 */
+                    ZSTD_storeSeq(seqStore, 0, anchor, 0, matchLength-MINMATCH);
+                    ip += matchLength;
+                    anchor = ip;
+                    continue;
+                }
+                break;
+            }
+        }
+
+        if (dictMode == ZSTD_noDict) {
             while ( ((ip <= ilimit) & (offset_2>0))
                  && (MEM_read32(ip) == MEM_read32(ip - offset_2)) ) {
                 /* store sequence */
@@ -582,7 +738,7 @@ _storeSequence:
                 ip += matchLength;
                 anchor = ip;
                 continue;   /* faster when present ... (?) */
-    }   }
+    }   }   }
 
     /* Save reps for next block */
     rep[0] = offset_1 ? offset_1 : savedOffset;
@@ -597,28 +753,56 @@ 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(ms, seqStore, rep, cParams, src, srcSize, 1, 2);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 1, 2, ZSTD_noDict);
 }
 
 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(ms, seqStore, rep, cParams, src, srcSize, 0, 2);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 2, ZSTD_noDict);
 }
 
 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(ms, seqStore, rep, cParams, src, srcSize, 0, 1);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 1, ZSTD_noDict);
 }
 
 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(ms, seqStore, rep, cParams, src, srcSize, 0, 0);
+    return ZSTD_compressBlock_lazy_generic(ms, seqStore, rep, cParams, src, srcSize, 0, 0, ZSTD_noDict);
+}
+
+size_t ZSTD_compressBlock_btlazy2_dictMatchState(
+        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(ms, seqStore, rep, cParams, src, srcSize, 1, 2, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_lazy2_dictMatchState(
+        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(ms, seqStore, rep, cParams, src, srcSize, 0, 2, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_lazy_dictMatchState(
+        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(ms, seqStore, rep, cParams, src, srcSize, 0, 1, ZSTD_dictMatchState);
+}
+
+size_t ZSTD_compressBlock_greedy_dictMatchState(
+        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(ms, seqStore, rep, cParams, src, srcSize, 0, 0, ZSTD_dictMatchState);
 }
 
 
@@ -646,7 +830,7 @@ size_t ZSTD_compressBlock_lazy_extDict_generic(
     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;
+    searchMax_f searchMax = searchMethod ? ZSTD_BtFindBestMatch_extDict_selectMLS : ZSTD_HcFindBestMatch_extDict_selectMLS;
 
     U32 offset_1 = rep[0], offset_2 = rep[1];
 

lib/compress/zstd_lazy.h

@@ -36,6 +36,19 @@ 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_btlazy2_dictMatchState(
+        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_dictMatchState(
+        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_dictMatchState(
+        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_dictMatchState(
+        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);