btlazy2

diff --git a/lib/zstdhc.c b/lib/zstdhc.c
index 693c713b2a05a5e6ddc1ec1f94008460d6f41e14..d053b2138f86ff12022cb3c0e75ad49ac3358c47 100644 (file)
--- a/lib/zstdhc.c
+++ b/lib/zstdhc.c
@@ -173,14 +173,14 @@ static const U64 prime6bytes = 227718039650203ULL;
 static size_t ZSTD_HC_hash6(U64 u, U32 h) { return (size_t)((u * prime6bytes) << (64-48) >> (64-h)) ; }
 static size_t ZSTD_HC_hash6Ptr(const void* p, U32 h) { return ZSTD_HC_hash6(MEM_read64(p), h); }
 
-static size_t ZSTD_HC_hashPtr(const void* p, U32 h, U32 mls)
+static size_t ZSTD_HC_hashPtr(const void* p, U32 hBits, U32 mls)
 {
     switch(mls)
     {
     default:
-    case 4: return ZSTD_HC_hash4Ptr(p,h);
-    case 5: return ZSTD_HC_hash5Ptr(p,h);
-    case 6: return ZSTD_HC_hash6Ptr(p,h);
+    case 4: return ZSTD_HC_hash4Ptr(p, hBits);
+    case 5: return ZSTD_HC_hash5Ptr(p, hBits);
+    case 6: return ZSTD_HC_hash6Ptr(p, hBits);
     }
 }
 
@@ -188,8 +188,308 @@ static size_t ZSTD_HC_hashPtr(const void* p, U32 h, U32 mls)
 
 
 /* *************************************
-*  HC Compression
+*  Binary Tree search
 ***************************************/
+#define BT_SHORTCUT 256
+
+/** ZSTD_HC_insertBt1 : add one ptr to tree
+    @ip : assumed <= iend-8 */
+static void ZSTD_HC_insertBt1(ZSTD_HC_CCtx* zc, const BYTE* const ip, const U32 mls, const BYTE* const iend, U32 nbCompares)
+{
+    U32* const hashTable = zc->hashTable;
+    const U32 hashLog = zc->params.hashLog;
+    const size_t h  = ZSTD_HC_hashPtr(ip, hashLog, mls);
+    U32* const bt   = zc->chainTable;
+    const U32 btLog = zc->params.chainLog - 1;
+    const U32 btMask= (1 << btLog) - 1;
+    U32 matchIndex  = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = zc->base;
+    const U32 current = (U32)(ip-base);
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = bt + 2*(current&btMask) + 1;
+       U32 dummy32;   /* to be nullified at the end */
+    const U32 windowSize = 1 << zc->params.windowLog;
+    const U32 windowLow = windowSize >= current ? 0 : current - windowSize;
+
+    hashTable[h] = (U32)(ip-base);   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow))
+    {
+        U32* nextPtr = bt + 2*(matchIndex & btMask);
+        const BYTE* match = base + matchIndex;
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+
+        matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+
+               if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+                       break;   /* just drop , to guarantee consistency (miss a bit of compression; if someone knows better, please tell) */
+
+        if (match[matchLength] < ip[matchLength])
+               {
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+                       if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+               }
+        else
+               {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+                       if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+               }
+    }
+
+       *smallerPtr = *largerPtr = 0;
+}
+
+
+FORCE_INLINE /* inlining is important to hardwire a hot branch (template emulation) */
+size_t ZSTD_HC_insertBtAndFindBestMatch (
+                        ZSTD_HC_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iend,
+                        size_t* offsetPtr,
+                        U32 nbCompares, const U32 mls)
+{
+    U32* const hashTable = zc->hashTable;
+    const U32 hashLog = zc->params.hashLog;
+    const size_t h  = ZSTD_HC_hashPtr(ip, hashLog, mls);
+    U32* const bt   = zc->chainTable;
+    const U32 btLog = zc->params.chainLog - 1;
+    const U32 btMask= (1 << btLog) - 1;
+    U32 matchIndex  = hashTable[h];
+    size_t commonLengthSmaller=0, commonLengthLarger=0;
+    const BYTE* const base = zc->base;
+    const U32 current = (U32)(ip-base);
+    const U32 btLow = btMask >= current ? 0 : current - btMask;
+    const U32 windowSize = 1 << zc->params.windowLog;
+    const U32 windowLow = windowSize >= current ? 0 : current - windowSize;
+    U32* smallerPtr = bt + 2*(current&btMask);
+    U32* largerPtr  = bt + 2*(current&btMask) + 1;
+    U32 bestLength = 0;
+       U32 dummy32;   /* to be nullified at the end */
+
+       hashTable[h] = (U32)(ip-base);   /* Update Hash Table */
+
+    while (nbCompares-- && (matchIndex > windowLow))
+    {
+        U32* nextPtr = bt + 2*(matchIndex & btMask);
+        const BYTE* match = base + matchIndex;
+        size_t matchLength = MIN(commonLengthSmaller, commonLengthLarger);   /* guaranteed minimum nb of common bytes */
+
+        matchLength += ZSTD_count(ip+matchLength, match+matchLength, iend);
+
+        if (matchLength > bestLength)
+        {
+            bestLength = (U32)matchLength;
+            *offsetPtr = current - matchIndex;
+                       if (ip+matchLength == iend)   /* equal : no way to know if inf or sup */
+                               break;   /* drop, next to null, to guarantee consistency (is there a way to do better ?) */
+        }
+
+        if (match[matchLength] < ip[matchLength])
+               {
+            /* match is smaller than current */
+            *smallerPtr = matchIndex;             /* update smaller idx */
+            commonLengthSmaller = matchLength;    /* all smaller will now have at least this guaranteed common length */
+                       if (matchIndex <= btLow) { smallerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+                       smallerPtr = nextPtr+1;               /* new "smaller" => larger of match */
+            matchIndex = nextPtr[1];              /* new matchIndex larger than previous (closer to current) */
+               }
+        else
+               {
+            /* match is larger than current */
+            *largerPtr = matchIndex;
+            commonLengthLarger = matchLength;
+                       if (matchIndex <= btLow) { largerPtr=&dummy32; break; }   /* beyond tree size, stop the search */
+            largerPtr = nextPtr;
+            matchIndex = nextPtr[0];
+               }
+    }
+
+       *smallerPtr = *largerPtr = 0;
+
+    zc->nextToUpdate = current+1;   /* current has been inserted */
+       if (bestLength < MINMATCH) return 0;
+    return bestLength;
+}
+
+
+static void ZSTD_HC_updateTree(ZSTD_HC_CCtx* zc, const BYTE* const ip, const BYTE* const iend, const U32 nbCompares, const U32 mls)
+{
+    const BYTE* const base = zc->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = zc->nextToUpdate;
+       //size_t dummy;
+
+    for( ; idx < target ; idx++)
+        ZSTD_HC_insertBt1(zc, base+idx, mls, iend, nbCompares);
+        //ZSTD_HC_insertBtAndFindBestMatch(zc, base+idx, iend, &dummy, nbCompares, mls);
+
+    zc->nextToUpdate = target;
+}
+
+
+/** Tree updater, providing best match */
+FORCE_INLINE /* inlining is important to hardwire a hot branch (template emulation) */
+size_t ZSTD_HC_BtFindBestMatch (
+                        ZSTD_HC_CCtx* zc,
+                        const BYTE* const ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 mls)
+{
+    ZSTD_HC_updateTree(zc, ip, iLimit, maxNbAttempts, mls);
+    return ZSTD_HC_insertBtAndFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, mls);
+}
+
+
+FORCE_INLINE size_t ZSTD_HC_BtFindBestMatch_selectMLS (
+                        ZSTD_HC_CCtx* zc,   /* Index table will be updated */
+                        const BYTE* ip, const BYTE* const iLimit,
+                        size_t* offsetPtr,
+                        const U32 maxNbAttempts, const U32 matchLengthSearch)
+{
+    switch(matchLengthSearch)
+    {
+    default :
+    case 4 : return ZSTD_HC_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 4);
+    case 5 : return ZSTD_HC_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 5);
+    case 6 : return ZSTD_HC_BtFindBestMatch(zc, ip, iLimit, offsetPtr, maxNbAttempts, 6);
+    }
+}
+
+
+size_t ZSTD_HC_compressBlock_btLazy2(ZSTD_HC_CCtx* ctx, void* dst, size_t maxDstSize, const void* src, size_t srcSize)
+{
+    seqStore_t* seqStorePtr = &(ctx->seqStore);
+    const BYTE* const istart = (const BYTE*)src;
+    const BYTE* ip = istart;
+    const BYTE* anchor = istart;
+    const BYTE* const iend = istart + srcSize;
+    const BYTE* const ilimit = iend - 8;
+
+    size_t offset_2=REPCODE_STARTVALUE, offset_1=REPCODE_STARTVALUE;
+    const U32 maxSearches = 1 << ctx->params.searchLog;
+    const U32 mls = ctx->params.searchLength;
+
+    /* init */
+    ZSTD_resetSeqStore(seqStorePtr);
+    if (((ip-ctx->base) - ctx->dictLimit) < REPCODE_STARTVALUE) ip += REPCODE_STARTVALUE;
+
+    /* Match Loop */
+    while (ip <= ilimit)
+    {
+        size_t matchLength;
+        size_t offset=999999;
+        const BYTE* start;
+
+        /* try to find a first match */
+        if (MEM_read32(ip) == MEM_read32(ip - offset_2))
+        {
+            /* repcode : we take it*/
+            size_t offtmp = offset_2;
+            size_t litLength = ip - anchor;
+            matchLength = ZSTD_count(ip+MINMATCH, ip+MINMATCH-offset_2, iend);
+            offset_2 = offset_1;
+            offset_1 = offtmp;
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, 0, matchLength);
+            ip += matchLength+MINMATCH;
+            anchor = ip;
+            continue;
+        }
+
+        offset_2 = offset_1;
+        matchLength = ZSTD_HC_BtFindBestMatch_selectMLS(ctx, ip, iend, &offset, maxSearches, mls);
+        if (!matchLength) { ip++; continue; }
+
+        /* let's try to find a better solution */
+        start = ip;
+
+        while (ip<ilimit)
+        {
+            ip ++;
+            if (MEM_read32(ip) == MEM_read32(ip - offset_1))
+            {
+                size_t ml2 = ZSTD_count(ip+MINMATCH, ip+MINMATCH-offset_1, iend) + MINMATCH;
+                int gain2 = (int)(ml2 * 3);
+                int gain1 = (int)(matchLength*3 - ZSTD_highbit((U32)offset+1) + 1);
+                if (gain2 > gain1)
+                    matchLength = ml2, offset = 0, start = ip;
+            }
+            {
+                size_t offset2=999999;
+                size_t ml2 = ZSTD_HC_BtFindBestMatch_selectMLS(ctx, ip, iend, &offset2, maxSearches, mls);
+                int gain2 = (int)(ml2*4 - ZSTD_highbit((U32)offset2+1));   /* raw approx */
+                int gain1 = (int)(matchLength*4 - ZSTD_highbit((U32)offset+1) + 4);
+                if (gain2 > gain1)
+                {
+                    matchLength = ml2, offset = offset2, start = ip;
+                    continue;   /* search a better one */
+                }
+            }
+
+            /* let's find an even better one */
+            if (ip<ilimit)
+            {
+                ip ++;
+                if (MEM_read32(ip) == MEM_read32(ip - offset_1))
+                {
+                    size_t ml2 = ZSTD_count(ip+MINMATCH, ip+MINMATCH-offset_1, iend) + MINMATCH;
+                    int gain2 = (int)(ml2 * 4);
+                    int gain1 = (int)(matchLength*4 - ZSTD_highbit((U32)offset+1) + 1);
+                    if (gain2 > gain1)
+                        matchLength = ml2, offset = 0, start = ip;
+                }
+                {
+                    size_t offset2=999999;
+                    size_t ml2 = ZSTD_HC_BtFindBestMatch_selectMLS(ctx, ip, iend, &offset2, maxSearches, mls);
+                    int gain2 = (int)(ml2*4 - ZSTD_highbit((U32)offset2+1));   /* raw approx */
+                    int gain1 = (int)(matchLength*4 - ZSTD_highbit((U32)offset+1) + 7);
+                    if (gain2 > gain1)
+                    {
+                        matchLength = ml2, offset = offset2, start = ip;
+                        continue;
+                    }
+                }
+            }
+            break;  /* nothing found : store previous solution */
+        }
+
+        /* store sequence */
+        {
+            size_t litLength = start - anchor;
+            if (offset) offset_1 = offset;
+            ZSTD_storeSeq(seqStorePtr, litLength, anchor, offset, matchLength-MINMATCH);
+            ip = start + matchLength;
+            anchor = ip;
+        }
+
+    }
+
+    /* Last Literals */
+    {
+        size_t lastLLSize = iend - anchor;
+        memcpy(seqStorePtr->lit, anchor, lastLLSize);
+        seqStorePtr->lit += lastLLSize;
+    }
+
+    /* Final compression stage */
+    return ZSTD_compressSequences((BYTE*)dst, maxDstSize,
+                                  seqStorePtr, srcSize);
+}
+
+
+
+/* ***********************
+*  Hash Chain
+*************************/
+
 /* Update chains up to ip (excluded) */
 static U32 ZSTD_HC_insertAndFindFirstIndex  (ZSTD_HC_CCtx* zc, const BYTE* ip, U32 mls)
 {
@@ -601,6 +901,8 @@ size_t ZSTD_HC_compressBlock(ZSTD_HC_CCtx* ctx, void* dst, size_t maxDstSize, co
         return ZSTD_HC_compressBlock_lazy(ctx, dst, maxDstSize, src, srcSize);
     case ZSTD_HC_lazydeep:
         return ZSTD_HC_compressBlock_lazydeep(ctx, dst, maxDstSize, src, srcSize);
+    case ZSTD_HC_btlazy2:
+        return ZSTD_HC_compressBlock_btLazy2(ctx, dst, maxDstSize, src, srcSize);
     default :
         return ERROR(GENERIC);   /* unknown block compressor */
     }
@@ -630,6 +932,9 @@ static size_t ZSTD_HC_compress_generic (ZSTD_HC_CCtx* ctxPtr,
     case ZSTD_HC_lazydeep:
         blockCompressor = ZSTD_HC_compressBlock_lazydeep;
         break;
+    case ZSTD_HC_btlazy2:
+        blockCompressor = ZSTD_HC_compressBlock_btLazy2;
+        break;
     default :
         return ERROR(GENERIC);   /* unknown block compressor */
     }

diff --git a/lib/zstdhc_static.h b/lib/zstdhc_static.h
index db185b400a53e5d1834dea0effe20a7fc770b1c9..36e0781f50621072e0815c914633e269a4b098e0 100644 (file)
--- a/lib/zstdhc_static.h
+++ b/lib/zstdhc_static.h
@@ -45,7 +45,7 @@ extern "C" {
 /* *************************************
 *  Types
 ***************************************/
-typedef enum { ZSTD_HC_greedy, ZSTD_HC_lazy, ZSTD_HC_lazydeep } ZSTD_HC_strategy;
+typedef enum { ZSTD_HC_greedy, ZSTD_HC_lazy, ZSTD_HC_lazydeep, ZSTD_HC_btlazy2 } ZSTD_HC_strategy;
 typedef struct
 {
     U32 windowLog;     /* largest match distance : impact decompression buffer size */
@@ -100,13 +100,13 @@ static const ZSTD_HC_parameters ZSTD_HC_defaultParameters[ZSTD_HC_MAX_CLEVEL+1]
     { 19, 13, 17,  3,  5, ZSTD_HC_greedy   },  /* level  3 */
     { 20, 18, 19,  2,  5, ZSTD_HC_greedy   },  /* level  4 */
     { 20, 18, 19,  2,  5, ZSTD_HC_lazy     },  /* level  5 */
-    { 20, 18, 20,  3,  5, ZSTD_HC_lazy     },  /* level  6 */
-    { 20, 18, 20,  4,  5, ZSTD_HC_lazy     },  /* level  7 */
+    { 20, 18, 19,  2,  5, ZSTD_HC_lazydeep },  //{ 20, 18, 20,  3,  5, ZSTD_HC_lazy     },  /* level  6 */
+    { 20, 18, 19,  2,  5, ZSTD_HC_btlazy2  },  //{ 20, 18, 20,  4,  5, ZSTD_HC_lazy     },  /* level  7 */
     { 21, 19, 20,  4,  5, ZSTD_HC_lazy     },  /* level  8 */
     { 21, 19, 20,  5,  5, ZSTD_HC_lazy     },  /* level  9 */
     { 21, 20, 20,  5,  5, ZSTD_HC_lazy     },  /* level 10 */
     { 21, 20, 20,  5,  5, ZSTD_HC_lazydeep },  /* level 11 */
-    { 22, 20, 22,  5,  5, ZSTD_HC_lazydeep },  /* level 12 */
+    { 21, 20, 20,  5,  5, ZSTD_HC_btlazy2  },  //{ 22, 20, 22,  5,  5, ZSTD_HC_lazydeep },  /* level 12 */
     { 22, 20, 22,  6,  5, ZSTD_HC_lazydeep },  /* level 13 */
     { 21, 21, 22,  6,  5, ZSTD_HC_lazydeep },  /* level 14 */
     { 22, 21, 22,  6,  5, ZSTD_HC_lazydeep },  /* level 15 */

diff --git a/programs/bench.c b/programs/bench.c
index 1892ddd60e5a6c3bfaf929ef548c055dfbe14a7f..c19d0fd97ffc941f246d2c8611b050782f00e030 100644 (file)
--- a/programs/bench.c
+++ b/programs/bench.c
@@ -83,7 +83,7 @@
 #define MB *(1 <<20)
 #define GB *(1U<<30)
 
-static const size_t maxMemory = sizeof(size_t)==4  ?  (2 GB - 64 MB) : (size_t)(1ULL << (sizeof(size_t)*8-31));
+static const size_t maxMemory = sizeof(size_t)==4  ?  (2 GB - 64 MB) : (size_t)(1ULL << ((sizeof(size_t)*8)-31));
 #define DEFAULT_CHUNKSIZE   (4 MB)
 
 static U32 g_compressibilityDefault = 50;

diff --git a/programs/paramgrill.c b/programs/paramgrill.c
index d27a22b5f39a361d03bb5c5610d544143faa043a..78f2b1959b7e0df363dbf2ea70ff8883aef24cdf 100644 (file)
--- a/programs/paramgrill.c
+++ b/programs/paramgrill.c
@@ -425,7 +425,7 @@ static size_t BMK_benchParam(BMK_result_t* resultPtr,
     return 0;
 }
 
-const char* g_stratName[] = { "ZSTD_HC_greedy  ", "ZSTD_HC_lazy    ", "ZSTD_HC_lazydeep" };
+const char* g_stratName[] = { "ZSTD_HC_greedy  ", "ZSTD_HC_lazy    ", "ZSTD_HC_lazydeep", "ZSTD_HC_btlazy2 " };
 
 static void BMK_printWinner(FILE* f, U32 cLevel, BMK_result_t result, ZSTD_HC_parameters params, size_t srcSize)
 {
@@ -576,7 +576,7 @@ static BYTE g_alreadyTested[ZSTD_HC_WINDOWLOG_MAX+1-ZSTD_HC_WINDOWLOG_MIN]
                            [ZSTD_HC_HASHLOG_MAX+1-ZSTD_HC_HASHLOG_MIN]
                            [ZSTD_HC_SEARCHLOG_MAX+1-ZSTD_HC_SEARCHLOG_MIN]
                            [ZSTD_HC_SEARCHLENGTH_MAX+1-ZSTD_HC_SEARCHLENGTH_MIN]
-                           [3 /* strategy */ ] = {};   /* init to zero */
+                           [4 /* strategy */ ] = {};   /* init to zero */
 
 #define NB_TESTS_PLAYED(p) \
     g_alreadyTested[p.windowLog-ZSTD_HC_WINDOWLOG_MIN] \
@@ -647,7 +647,7 @@ static void playAround(FILE* f, winnerInfo_t* winners,
         if (p.searchLength > ZSTD_HC_SEARCHLENGTH_MAX) continue;
         if (p.searchLength < ZSTD_HC_SEARCHLENGTH_MIN) continue;
         if (p.strategy < ZSTD_HC_greedy) continue;
-        if (p.strategy > ZSTD_HC_lazydeep) continue;
+        if (p.strategy > ZSTD_HC_btlazy2) continue;
 
         /* exclude faster if already played params */
         if (FUZ_rand(&g_rand) & ((1 << NB_TESTS_PLAYED(p))-1))
@@ -680,7 +680,7 @@ static void BMK_selectRandomStart(
         p.searchLog  = FUZ_rand(&g_rand) % (ZSTD_HC_SEARCHLOG_MAX+1 - ZSTD_HC_SEARCHLOG_MIN) + ZSTD_HC_SEARCHLOG_MIN;
         p.windowLog  = FUZ_rand(&g_rand) % (ZSTD_HC_WINDOWLOG_MAX+1 - ZSTD_HC_WINDOWLOG_MIN) + ZSTD_HC_WINDOWLOG_MIN;
         p.searchLength=FUZ_rand(&g_rand) % (ZSTD_HC_SEARCHLENGTH_MAX+1 - ZSTD_HC_SEARCHLENGTH_MIN) + ZSTD_HC_SEARCHLENGTH_MIN;
-        p.strategy   = (ZSTD_HC_strategy) (FUZ_rand(&g_rand) % 3);
+        p.strategy   = (ZSTD_HC_strategy) (FUZ_rand(&g_rand) % 4);
         playAround(f, winners, p, srcBuffer, srcSize, ctx);
     }
     else