Merge pull request #287 from inikep/dev

opt parser

6 files changed, 449 insertions, 21 deletions

diff --git a/lib/README.md b/lib/README.md
index 18793d4..b40442c 100644
--- a/lib/README.md
+++ b/lib/README.md
@@ -47,7 +47,7 @@ The header files `lz4.h`, `lz4hc.h`, `lz4frame.h` and the dynamic library
 `dll\liblz4.dll` are required to compile a project using gcc/MinGW.
 The dynamic library has to be added to linking options.
 It means that if a project that uses LZ4 consists of a single `test-dll.c`
-file it should be compiled with "liblz4.lib". For example:
+file it should be linked with `dll\liblz4.dll`. For example:
 ```
     gcc $(CFLAGS) -Iinclude/ test-dll.c -o test-dll dll\liblz4.dll
 ```
diff --git a/lib/dll/example/README.md b/lib/dll/example/README.md
index 4908663..223e473 100644
--- a/lib/dll/example/README.md
+++ b/lib/dll/example/README.md
@@ -39,7 +39,7 @@ The header files from `include\` and the dynamic library `dll\liblz4.dll`
 are required to compile a project using gcc/MinGW.
 The dynamic library has to be added to linking options.
 It means that if a project that uses LZ4 consists of a single `test-dll.c`
-file it should be compiled with "liblz4.dll". For example:
+file it should be linked with `dll\liblz4.dll`. For example:
 ```
     gcc $(CFLAGS) -Iinclude\ test-dll.c -o test-dll dll\liblz4.dll
 ```
diff --git a/lib/lz4hc.c b/lib/lz4hc.c
index 298550c..44a1340 100644
--- a/lib/lz4hc.c
+++ b/lib/lz4hc.c
@@ -84,7 +84,7 @@
 *  Local Macros
 **************************************/
 #define HASH_FUNCTION(i)       (((i) * 2654435761U) >> ((MINMATCH*8)-LZ4HC_HASH_LOG))
-/* #define DELTANEXTU16(p)        chainTable[(p) & LZ4HC_MAXD_MASK] */   /* flexible, LZ4HC_MAXD dependent */
+#define DELTANEXTMAXD(p)       chainTable[(p) & LZ4HC_MAXD_MASK]    /* flexible, LZ4HC_MAXD dependent */
 #define DELTANEXTU16(p)        chainTable[(U16)(p)]   /* faster */
 
 static U32 LZ4HC_hashPtr(const void* ptr) { return HASH_FUNCTION(LZ4_read32(ptr)); }
@@ -98,7 +98,7 @@ static void LZ4HC_init (LZ4HC_CCtx_internal* hc4, const BYTE* start)
 {
     MEM_INIT((void*)hc4->hashTable, 0, sizeof(hc4->hashTable));
     MEM_INIT(hc4->chainTable, 0xFF, sizeof(hc4->chainTable));
-    hc4->nextToUpdate = 64 KB;
+    hc4->nextToUpdate = hc4->nextToUpdateBT = 64 KB;
     hc4->base = start - 64 KB;
     hc4->end = start;
     hc4->dictBase = start - 64 KB;
@@ -305,14 +305,15 @@ FORCE_INLINE int LZ4HC_encodeSequence (
     return 0;
 }
 
+#include "lz4opt.h"
 
-static int LZ4HC_compress_generic (
+static int LZ4HC_compress_hashChain (
     LZ4HC_CCtx_internal* const ctx,
     const char* const source,
     char* const dest,
     int const inputSize,
     int const maxOutputSize,
-    int compressionLevel,
+    unsigned maxNbAttempts,
     limitedOutput_directive limit
     )
 {
@@ -325,7 +326,6 @@ static int LZ4HC_compress_generic (
     BYTE* op = (BYTE*) dest;
     BYTE* const oend = op + maxOutputSize;
 
-    unsigned maxNbAttempts;
     int   ml, ml2, ml3, ml0;
     const BYTE* ref = NULL;
     const BYTE* start2 = NULL;
@@ -336,9 +336,6 @@ static int LZ4HC_compress_generic (
     const BYTE* ref0;
 
     /* init */
-    if (compressionLevel > LZ4HC_MAX_CLEVEL) compressionLevel = LZ4HC_MAX_CLEVEL;
-    if (compressionLevel < 1) compressionLevel = LZ4HC_DEFAULT_CLEVEL;
-    maxNbAttempts = 1 << (compressionLevel-1);
     ctx->end += inputSize;
 
     ip++;
@@ -490,6 +487,29 @@ _Search3:
 }
 
 
+static int LZ4HC_compress_generic (
+    LZ4HC_CCtx_internal* const ctx,
+    const char* const source,
+    char* const dest,
+    int const inputSize,
+    int const maxOutputSize,
+    int compressionLevel,
+    limitedOutput_directive limit
+    )
+{
+    if (compressionLevel < 1) compressionLevel = LZ4HC_DEFAULT_CLEVEL;
+    if (compressionLevel > 9) {
+        switch (compressionLevel) {
+            case 10: return LZ4HC_compress_hashChain(ctx, source, dest, inputSize, maxOutputSize, 1 << (16-1), limit);
+            case 11: ctx->searchNum = 128;  return LZ4HC_compress_optimal(ctx, source, dest, inputSize, maxOutputSize, limit, 128, 0);
+            default:
+            case 12: ctx->searchNum = 1<<10; return LZ4HC_compress_optimal(ctx, source, dest, inputSize, maxOutputSize, limit, LZ4_OPT_NUM, 1);
+        }
+    }
+    return LZ4HC_compress_hashChain(ctx, source, dest, inputSize, maxOutputSize, 1 << (compressionLevel-1), limit);
+}
+
+
 int LZ4_sizeofStateHC(void) { return sizeof(LZ4_streamHC_t); }
 
 int LZ4_compress_HC_extStateHC (void* state, const char* src, char* dst, int srcSize, int maxDstSize, int compressionLevel)
@@ -554,14 +574,18 @@ int LZ4_loadDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, const char* dictionary, int
 
 static void LZ4HC_setExternalDict(LZ4HC_CCtx_internal* ctxPtr, const BYTE* newBlock)
 {
-    if (ctxPtr->end >= ctxPtr->base + 4) LZ4HC_Insert (ctxPtr, ctxPtr->end-3);   /* Referencing remaining dictionary content */
+    if (ctxPtr->compressionLevel >= LZ4HC_MIN_CLEVEL_OPT) 
+        LZ4HC_updateBinTree(ctxPtr, ctxPtr->end - MFLIMIT, ctxPtr->end - LASTLITERALS);
+    else
+        if (ctxPtr->end >= ctxPtr->base + 4) LZ4HC_Insert (ctxPtr, ctxPtr->end-3);   /* Referencing remaining dictionary content */
+
     /* Only one memory segment for extDict, so any previous extDict is lost at this stage */
     ctxPtr->lowLimit  = ctxPtr->dictLimit;
     ctxPtr->dictLimit = (U32)(ctxPtr->end - ctxPtr->base);
     ctxPtr->dictBase  = ctxPtr->base;
     ctxPtr->base = newBlock - ctxPtr->dictLimit;
     ctxPtr->end  = newBlock;
-    ctxPtr->nextToUpdate = ctxPtr->dictLimit;   /* match referencing will resume from there */
+    ctxPtr->nextToUpdate = ctxPtr->nextToUpdateBT = ctxPtr->dictLimit;   /* match referencing will resume from there */
 }
 
 static int LZ4_compressHC_continue_generic (LZ4_streamHC_t* LZ4_streamHCPtr,
@@ -621,6 +645,7 @@ int LZ4_saveDictHC (LZ4_streamHC_t* LZ4_streamHCPtr, char* safeBuffer, int dictS
         streamPtr->dictLimit = endIndex - dictSize;
         streamPtr->lowLimit = endIndex - dictSize;
         if (streamPtr->nextToUpdate < streamPtr->dictLimit) streamPtr->nextToUpdate = streamPtr->dictLimit;
+        if (streamPtr->nextToUpdateBT < streamPtr->dictLimit) streamPtr->nextToUpdateBT = streamPtr->dictLimit;
     }
     return dictSize;
 }
diff --git a/lib/lz4hc.h b/lib/lz4hc.h
index d8ac3c0..7ce5958 100644
--- a/lib/lz4hc.h
+++ b/lib/lz4hc.h
@@ -46,7 +46,8 @@ extern "C" {
 /* --- Useful constants --- */
 #define LZ4HC_MIN_CLEVEL        3
 #define LZ4HC_DEFAULT_CLEVEL    9
-#define LZ4HC_MAX_CLEVEL        16
+#define LZ4HC_MIN_CLEVEL_OPT    11
+#define LZ4HC_MAX_CLEVEL        12
 
 
 /*-************************************
@@ -58,7 +59,7 @@ extern "C" {
  * Compression is guaranteed to succeed if `dstCapacity >= LZ4_compressBound(srcSize)` (see "lz4.h")
  * Max supported `srcSize` value is LZ4_MAX_INPUT_SIZE (see "lz4.h")
  * `compressionLevel` : Recommended values are between 4 and 9, although any value between 1 and LZ4HC_MAX_CLEVEL will work.
- *                      Values >LZ4HC_MAX_CLEVEL behave the same as 16.
+ *                      Values >LZ4HC_MAX_CLEVEL behave the same as LZ4HC_MAX_CLEVEL.
  * @return : the number of bytes written into 'dst'
  *           or 0 if compression fails.
  */
@@ -130,11 +131,11 @@ LZ4LIB_API int LZ4_saveDictHC (LZ4_streamHC_t* streamHCPtr, char* safeBuffer, in
  * They are exposed to allow static allocation of `LZ4_streamHC_t`.
  * Using these definitions makes the code vulnerable to potential API break when upgrading LZ4
  **************************************/
-#define LZ4HC_DICTIONARY_LOGSIZE 16
+#define LZ4HC_DICTIONARY_LOGSIZE 17
 #define LZ4HC_MAXD (1<<LZ4HC_DICTIONARY_LOGSIZE)
 #define LZ4HC_MAXD_MASK (LZ4HC_MAXD - 1)
 
-#define LZ4HC_HASH_LOG (LZ4HC_DICTIONARY_LOGSIZE-1)
+#define LZ4HC_HASH_LOG 15
 #define LZ4HC_HASHTABLESIZE (1 << LZ4HC_HASH_LOG)
 #define LZ4HC_HASH_MASK (LZ4HC_HASHTABLESIZE - 1)
 
@@ -153,6 +154,8 @@ typedef struct
     uint32_t   dictLimit;        /* below that point, need extDict */
     uint32_t   lowLimit;         /* below that point, no more dict */
     uint32_t   nextToUpdate;     /* index from which to continue dictionary update */
+    uint32_t   nextToUpdateBT;   /* index from which to continue binary tree update */
+    uint32_t   searchNum;        /* only for optimal parser */
     uint32_t   compressionLevel;
 } LZ4HC_CCtx_internal;
 
@@ -169,12 +172,14 @@ typedef struct
     unsigned int   dictLimit;        /* below that point, need extDict */
     unsigned int   lowLimit;         /* below that point, no more dict */
     unsigned int   nextToUpdate;     /* index from which to continue dictionary update */
+    unsigned int   nextToUpdateBT;     /* index from which to continue binary tree update */
+    unsigned int   searchNum;        /* only for optimal parser */
     unsigned int   compressionLevel;
 } LZ4HC_CCtx_internal;
 
 #endif
 
-#define LZ4_STREAMHCSIZE        262192
+#define LZ4_STREAMHCSIZE       (4*LZ4HC_HASHTABLESIZE + 2*LZ4HC_MAXD + 56) /* 393268 */
 #define LZ4_STREAMHCSIZE_SIZET (LZ4_STREAMHCSIZE / sizeof(size_t))
 union LZ4_streamHC_u {
     size_t table[LZ4_STREAMHCSIZE_SIZET];
diff --git a/lib/lz4opt.h b/lib/lz4opt.h
new file mode 100644
index 0000000..f487edf
--- /dev/null
+++ b/lib/lz4opt.h
@@ -0,0 +1,395 @@
+/*
+    lz4opt.h - Optimal Mode of LZ4
+    Copyright (C) 2015-2016, Przemyslaw Skibinski <inikep@gmail.com>
+    Note : this file is intended to be included within lz4hc.c
+ 
+    BSD 2-Clause License (http://www.opensource.org/licenses/bsd-license.php)
+
+    Redistribution and use in source and binary forms, with or without
+    modification, are permitted provided that the following conditions are
+    met:
+
+    * Redistributions of source code must retain the above copyright
+    notice, this list of conditions and the following disclaimer.
+    * Redistributions in binary form must reproduce the above
+    copyright notice, this list of conditions and the following disclaimer
+    in the documentation and/or other materials provided with the
+    distribution.
+
+    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+    "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+    OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+    SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+    LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+    DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+    THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+    (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+    OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+    You can contact the author at :
+       - LZ4 source repository : https://github.com/lz4/lz4
+       - LZ4 public forum : https://groups.google.com/forum/#!forum/lz4c
+*/
+
+#define LZ4_LOG_PARSER(fmt, ...) //printf(fmt, __VA_ARGS__) 
+#define LZ4_LOG_PRICE(fmt, ...) //printf(fmt, __VA_ARGS__) 
+#define LZ4_LOG_ENCODE(fmt, ...) //printf(fmt, __VA_ARGS__) 
+
+
+#define LZ4_OPT_NUM   (1<<12)
+
+
+typedef struct
+{
+    int off;
+    int len;
+} LZ4HC_match_t;
+
+typedef struct
+{
+    int price;
+    int off;
+    int mlen;
+    int litlen;
+} LZ4HC_optimal_t;
+
+
+FORCE_INLINE size_t LZ4HC_GetLiteralsPrice(size_t litlen)
+{
+    size_t price = 8*litlen;
+    if (litlen >= (int)RUN_MASK) price+=8*(1+(litlen-RUN_MASK)/255);
+    return price;
+}
+
+
+FORCE_INLINE size_t LZ4HC_get_price(size_t litlen, size_t mlen)
+{
+    size_t price = 16 + 8; /* 16-bit offset + token */
+
+    price += 8*litlen;
+    if (litlen >= (int)RUN_MASK) price+=8*(1+(litlen-RUN_MASK)/255);
+
+    mlen -= MINMATCH;
+    if (mlen >= (int)ML_MASK) price+=8*(1+(mlen-ML_MASK)/255);
+
+    return price;
+}
+
+
+/*-*************************************
+*  Binary Tree search
+***************************************/
+FORCE_INLINE int LZ4HC_BinTree_InsertAndGetAllMatches (
+    LZ4HC_CCtx_internal* ctx,
+    const BYTE* const ip,
+    const BYTE* const iHighLimit,
+    size_t best_mlen,
+    LZ4HC_match_t* matches,
+    int* matchNum)
+{
+    U16* const chainTable = ctx->chainTable;
+    U32* const HashTable = ctx->hashTable;
+    const BYTE* const base = ctx->base;
+    const U32 dictLimit = ctx->dictLimit;
+    const U32 current = (U32)(ip - base);
+    const U32 lowLimit = (ctx->lowLimit + MAX_DISTANCE > current) ? ctx->lowLimit : current - (MAX_DISTANCE - 1);
+    const BYTE* const dictBase = ctx->dictBase;
+    const BYTE* match;
+    int nbAttempts = ctx->searchNum;
+    int mnum = 0;
+    U16 *ptr0, *ptr1, delta0, delta1;
+    U32 matchIndex;
+    size_t matchLength = 0;
+    U32* HashPos;
+    
+    if (ip + MINMATCH > iHighLimit) return 1;
+
+    /* HC4 match finder */
+    HashPos = &HashTable[LZ4HC_hashPtr(ip)];
+    matchIndex = *HashPos;
+    *HashPos = current;
+
+    ptr0 = &DELTANEXTMAXD(current*2+1);
+    ptr1 = &DELTANEXTMAXD(current*2);
+    delta0 = delta1 = (U16)(current - matchIndex);
+
+    while ((matchIndex < current) && (matchIndex>=lowLimit) && (nbAttempts)) {
+        nbAttempts--;
+        if (matchIndex >= dictLimit) {
+            match = base + matchIndex;
+            matchLength = LZ4_count(ip, match, iHighLimit);
+        } else {
+            const BYTE* vLimit = ip + (dictLimit - matchIndex);
+            match = dictBase + matchIndex;
+            if (vLimit > iHighLimit) vLimit = iHighLimit;
+            matchLength = LZ4_count(ip, match, vLimit);
+            if ((ip+matchLength == vLimit) && (vLimit < iHighLimit))
+                matchLength += LZ4_count(ip+matchLength, base+dictLimit, iHighLimit);
+        }
+
+        if (matchLength > best_mlen) {
+            best_mlen = matchLength;
+            if (matches) {
+                matches[mnum].off = (int)(ip - match);
+                matches[mnum].len = (int)matchLength;
+                mnum++;
+            }
+            if (best_mlen > LZ4_OPT_NUM) break;
+        }
+
+        if (ip+matchLength >= iHighLimit)   /* equal : no way to know if inf or sup */
+            break;   /* drop , to guarantee consistency ; miss a bit of compression, but other solutions can corrupt the tree */
+
+        if (*(ip+matchLength) < *(match+matchLength)) {
+            *ptr0 = delta0;
+            ptr0 = &DELTANEXTMAXD(matchIndex*2);
+            if (*ptr0 == (U16)-1) break;
+            delta0 = *ptr0;
+            delta1 += delta0;
+            matchIndex -= delta0;
+        } else {
+            *ptr1 = delta1;
+            ptr1 = &DELTANEXTMAXD(matchIndex*2+1);
+            if (*ptr1 == (U16)-1) break;
+            delta1 = *ptr1;
+            delta0 += delta1;
+            matchIndex -= delta1;
+        }
+    }
+
+    *ptr0 = (U16)-1;
+    *ptr1 = (U16)-1;
+    if (matchNum) *matchNum = mnum;
+  /*  if (best_mlen > 8) return best_mlen-8; */
+    if (!matchNum) return 1;
+    return 1; 
+}
+
+
+FORCE_INLINE void LZ4HC_updateBinTree(LZ4HC_CCtx_internal* ctx, const BYTE* const ip, const BYTE* const iHighLimit)
+{
+    const BYTE* const base = ctx->base;
+    const U32 target = (U32)(ip - base);
+    U32 idx = ctx->nextToUpdateBT;
+
+    while(idx < target)
+        idx += LZ4HC_BinTree_InsertAndGetAllMatches(ctx, base+idx, iHighLimit, 8, NULL, NULL);
+}
+
+
+/** Tree updater, providing best match */
+FORCE_INLINE int LZ4HC_BinTree_GetAllMatches (
+                        LZ4HC_CCtx_internal* ctx,
+                        const BYTE* const ip, const BYTE* const iHighLimit,
+                        size_t best_mlen, LZ4HC_match_t* matches, const int fullUpdate)
+{
+    int mnum = 0;
+    if (ip < ctx->base + ctx->nextToUpdateBT) return 0;   /* skipped area */
+    if (fullUpdate) LZ4HC_updateBinTree(ctx, ip, iHighLimit);
+    best_mlen = LZ4HC_BinTree_InsertAndGetAllMatches(ctx, ip, iHighLimit, best_mlen, matches, &mnum);
+    ctx->nextToUpdateBT = (U32)(ip - ctx->base + best_mlen);
+    return mnum;
+}
+
+
+#define SET_PRICE(pos, mlen, offset, litlen, price)    \
+{                                                      \
+    while (last_pos < pos)  { opt[last_pos+1].price = 1<<30; last_pos++; } \
+    opt[pos].mlen = (int)mlen;                         \
+    opt[pos].off = (int)offset;                        \
+    opt[pos].litlen = (int)litlen;                     \
+    opt[pos].price = (int)price;                       \
+    LZ4_LOG_PARSER("%d: SET price[%d/%d]=%d litlen=%d len=%d off=%d\n", (int)(inr-source), pos, last_pos, opt[pos].price, opt[pos].litlen, opt[pos].mlen, opt[pos].off); \
+}
+
+
+static int LZ4HC_compress_optimal (
+    LZ4HC_CCtx_internal* ctx,
+    const char* const source,
+    char* dest,
+    int inputSize,
+    int maxOutputSize,
+    limitedOutput_directive limit,
+    const size_t sufficient_len,
+    const int fullUpdate
+    )
+{
+    LZ4HC_optimal_t opt[LZ4_OPT_NUM + 1];
+    LZ4HC_match_t matches[LZ4_OPT_NUM + 1];
+    const BYTE *inr;
+    size_t res, cur, cur2;
+    size_t i, llen, litlen, mlen, best_mlen, price, offset, best_off, match_num, last_pos;
+
+    const BYTE* ip = (const BYTE*) source;
+    const BYTE* anchor = ip;
+    const BYTE* const iend = ip + inputSize;
+    const BYTE* const mflimit = iend - MFLIMIT;
+    const BYTE* const matchlimit = (iend - LASTLITERALS);
+    BYTE* op = (BYTE*) dest;
+    BYTE* const oend = op + maxOutputSize;
+
+    /* init */
+    ctx->end += inputSize;
+    ip++;
+
+    /* Main Loop */
+    while (ip < mflimit) {
+        memset(opt, 0, sizeof(LZ4HC_optimal_t));
+        last_pos = 0;
+        llen = ip - anchor;
+        match_num = LZ4HC_BinTree_GetAllMatches(ctx, ip, matchlimit, MINMATCH-1, matches, fullUpdate);
+        if (!match_num) { ip++; continue; }
+        LZ4_LOG_PARSER("%d: match_num=%d last_pos=%d\n", (int)(ip-source), match_num, last_pos);
+
+        if ((size_t)matches[match_num-1].len > sufficient_len) {
+            best_mlen = matches[match_num-1].len;
+            best_off = matches[match_num-1].off;
+            cur = 0;
+            last_pos = 1;
+            goto encode;
+        }
+
+        /* set prices using matches at position = 0 */
+        for (i = 0; i < match_num; i++) {
+           mlen = (i>0) ? (size_t)matches[i-1].len+1 : MINMATCH;
+           best_mlen = (matches[i].len < LZ4_OPT_NUM) ? matches[i].len : LZ4_OPT_NUM;
+           LZ4_LOG_PARSER("%d: start Found mlen=%d off=%d best_mlen=%d last_pos=%d\n", (int)(ip-source), matches[i].len, matches[i].off, best_mlen, last_pos);
+           while (mlen <= best_mlen) {
+                litlen = 0;
+                price = LZ4HC_get_price(llen + litlen, mlen) - LZ4HC_GetLiteralsPrice(llen);
+                SET_PRICE(mlen, mlen, matches[i].off, litlen, price);
+                mlen++;
+           }
+        }
+
+        if (last_pos < MINMATCH) { ip++; continue; }
+
+        /* check further positions */
+        opt[0].mlen = opt[1].mlen = 1;
+        for (cur = 1; cur <= last_pos; cur++) {
+            inr = ip + cur;
+
+            if (opt[cur-1].mlen == 1) {
+                litlen = opt[cur-1].litlen + 1;
+                if (cur != litlen) {
+                    price = opt[cur - litlen].price + LZ4HC_GetLiteralsPrice(litlen);
+                    LZ4_LOG_PRICE("%d: TRY1 opt[%d].price=%d price=%d cur=%d litlen=%d\n", (int)(inr-source), cur - litlen, opt[cur - litlen].price, price, cur, litlen);
+                } else {
+                    price = LZ4HC_GetLiteralsPrice(llen + litlen) - LZ4HC_GetLiteralsPrice(llen);
+                    LZ4_LOG_PRICE("%d: TRY2 price=%d cur=%d litlen=%d llen=%d\n", (int)(inr-source), price, cur, litlen, llen);
+                }
+            } else {
+                litlen = 1;
+                price = opt[cur - 1].price + LZ4HC_GetLiteralsPrice(litlen);                  
+                LZ4_LOG_PRICE("%d: TRY3 price=%d cur=%d litlen=%d litonly=%d\n", (int)(inr-source), price, cur, litlen, LZ4HC_GetLiteralsPrice(litlen));
+            }
+           
+            mlen = 1;
+            best_mlen = 0;
+            LZ4_LOG_PARSER("%d: TRY price=%d opt[%d].price=%d\n", (int)(inr-source), price, cur, opt[cur].price);
+            if (cur > last_pos || price < (size_t)opt[cur].price)
+                SET_PRICE(cur, mlen, best_mlen, litlen, price);
+
+            if (cur == last_pos || inr >= mflimit) break;
+            LZ4_LOG_PARSER("%d: CURRENT price[%d/%d]=%d off=%d mlen=%d litlen=%d\n", (int)(inr-source), cur, last_pos, opt[cur].price, opt[cur].off, opt[cur].mlen, opt[cur].litlen); 
+
+            match_num = LZ4HC_BinTree_GetAllMatches(ctx, inr, matchlimit, MINMATCH-1, matches, fullUpdate);
+            LZ4_LOG_PARSER("%d: LZ4HC_BinTree_GetAllMatches match_num=%d\n", (int)(inr-source), match_num);
+
+            if (match_num > 0 && (size_t)matches[match_num-1].len > sufficient_len) {
+                best_mlen = matches[match_num-1].len;
+                best_off = matches[match_num-1].off;
+                last_pos = cur + 1;
+                goto encode;
+            }
+
+            /* set prices using matches at position = cur */
+            for (i = 0; i < match_num; i++) {
+                mlen = (i>0) ? (size_t)matches[i-1].len+1 : MINMATCH;
+                cur2 = cur;
+                best_mlen = (cur2 + matches[i].len < LZ4_OPT_NUM) ? (size_t)matches[i].len : LZ4_OPT_NUM - cur2;
+                LZ4_LOG_PARSER("%d: Found1 cur=%d cur2=%d mlen=%d off=%d best_mlen=%d last_pos=%d\n", (int)(inr-source), cur, cur2, matches[i].len, matches[i].off, best_mlen, last_pos);
+
+                while (mlen <= best_mlen) {
+                    if (opt[cur2].mlen == 1) {
+                        litlen = opt[cur2].litlen;
+
+                        if (cur2 != litlen)
+                            price = opt[cur2 - litlen].price + LZ4HC_get_price(litlen, mlen);
+                        else
+                            price = LZ4HC_get_price(llen + litlen, mlen) - LZ4HC_GetLiteralsPrice(llen);
+                    } else {
+                        litlen = 0;
+                        price = opt[cur2].price + LZ4HC_get_price(litlen, mlen);
+                    }
+
+                    LZ4_LOG_PARSER("%d: Found2 mlen=%d best_mlen=%d off=%d price=%d litlen=%d price[%d]=%d\n", (int)(inr-source), mlen, best_mlen, matches[i].off, price, litlen, cur - litlen, opt[cur - litlen].price);
+                    if (cur2 + mlen > last_pos || price < (size_t)opt[cur2 + mlen].price) { // || (((int)price == opt[cur2 + mlen].price) && (opt[cur2 + mlen-1].mlen == 1))) {
+                        SET_PRICE(cur2 + mlen, mlen, matches[i].off, litlen, price);
+                    }
+                    mlen++;
+                }
+            }
+        } /* for (cur = 1; cur <= last_pos; cur++) */
+
+        best_mlen = opt[last_pos].mlen;
+        best_off = opt[last_pos].off;
+        cur = last_pos - best_mlen;
+
+encode: /* cur, last_pos, best_mlen, best_off have to be set */
+        for (i = 1; i <= last_pos; i++) {
+            LZ4_LOG_PARSER("%d: price[%d/%d]=%d off=%d mlen=%d litlen=%d\n", (int)(ip-source+i), i, last_pos, opt[i].price, opt[i].off, opt[i].mlen, opt[i].litlen); 
+        }
+
+        LZ4_LOG_PARSER("%d: cur=%d/%d best_mlen=%d best_off=%d\n", (int)(ip-source+cur), cur, last_pos, best_mlen, best_off); 
+
+        opt[0].mlen = 1;
+        while (1) {
+            mlen = opt[cur].mlen;
+            offset = opt[cur].off;
+            opt[cur].mlen = (int)best_mlen; 
+            opt[cur].off = (int)best_off;
+            best_mlen = mlen;
+            best_off = offset;
+            if (mlen > cur) break;
+            cur -= mlen;
+        }
+          
+        for (i = 0; i <= last_pos;) {
+            LZ4_LOG_PARSER("%d: price2[%d/%d]=%d off=%d mlen=%d litlen=%d\n", (int)(ip-source+i), i, last_pos, opt[i].price, opt[i].off, opt[i].mlen, opt[i].litlen); 
+            i += opt[i].mlen;
+        }
+
+        cur = 0;
+        while (cur < last_pos) {
+            LZ4_LOG_PARSER("%d: price3[%d/%d]=%d off=%d mlen=%d litlen=%d\n", (int)(ip-source+cur), cur, last_pos, opt[cur].price, opt[cur].off, opt[cur].mlen, opt[cur].litlen); 
+            mlen = opt[cur].mlen;
+            if (mlen == 1) { ip++; cur++; continue; }
+            offset = opt[cur].off;
+            cur += mlen;
+
+            LZ4_LOG_ENCODE("%d: ENCODE literals=%d off=%d mlen=%d ", (int)(ip-(const BYTE*)source), (int)(ip-anchor), (int)(offset), (int)mlen);
+            res = LZ4HC_encodeSequence(&ip, &op, &anchor, (int)mlen, ip - offset, limit, oend);
+            LZ4_LOG_ENCODE("out=%d\n", (int)((char*)op - dest));
+
+            if (res) return 0; 
+
+            LZ4_LOG_PARSER("%d: offset=%d\n", (int)(ip-source), offset);
+        }
+    }
+
+    /* Encode Last Literals */
+    {   int lastRun = (int)(iend - anchor);
+        if ((limit) && (((char*)op - dest) + lastRun + 1 + ((lastRun+255-RUN_MASK)/255) > (U32)maxOutputSize)) return 0;  /* Check output limit */
+        if (lastRun>=(int)RUN_MASK) { *op++=(RUN_MASK<<ML_BITS); lastRun-=RUN_MASK; for(; lastRun > 254 ; lastRun-=255) *op++ = 255; *op++ = (BYTE) lastRun; }
+        else *op++ = (BYTE)(lastRun<<ML_BITS);
+        LZ4_LOG_ENCODE("%d: ENCODE_LAST literals=%d out=%d\n", (int)(ip-(const BYTE*)source), (int)(iend-anchor), (int)((char*)op-dest));
+        memcpy(op, anchor, iend - anchor);
+        op += iend-anchor;
+    }
+
+    /* End */
+    return (int) ((char*)op-dest);
+}
diff --git a/programs/lz4io.c b/programs/lz4io.c
index bc2ba95..46daaf7 100644
--- a/programs/lz4io.c
+++ b/programs/lz4io.c
@@ -591,12 +591,15 @@ int LZ4IO_compressMultipleFilenames(const char** inFileNamesTable, int ifntSize,
     char* dstFileName = (char*)malloc(FNSPACE);
     size_t ofnSize = FNSPACE;
     const size_t suffixSize = strlen(suffix);
-    cRess_t const ress = LZ4IO_createCResources();
+    cRess_t ress;
+
+    if (dstFileName == NULL) return ifntSize;   /* not enough memory */
+    ress = LZ4IO_createCResources();
 
     /* loop on each file */
     for (i=0; i<ifntSize; i++) {
         size_t const ifnSize = strlen(inFileNamesTable[i]);
-        if (ofnSize <= ifnSize+suffixSize+1) { free(dstFileName); ofnSize = ifnSize + 20; dstFileName = (char*)malloc(ofnSize); }
+        if (ofnSize <= ifnSize+suffixSize+1) { free(dstFileName); ofnSize = ifnSize + 20; dstFileName = (char*)malloc(ofnSize); if (dstFileName==NULL) { LZ4IO_freeCResources(ress); return ifntSize; } }
         strcpy(dstFileName, inFileNamesTable[i]);
         strcat(dstFileName, suffix);
 
@@ -1022,7 +1025,7 @@ int LZ4IO_decompressMultipleFilenames(const char** inFileNamesTable, int ifntSiz
     size_t const suffixSize = strlen(suffix);
     dRess_t ress = LZ4IO_createDResources();
 
-    if (outFileName==NULL) exit(1);   /* not enough memory */
+    if (outFileName==NULL) return ifntSize;   /* not enough memory */
     ress.dstFile = LZ4IO_openDstFile(stdoutmark);
 
     for (i=0; i<ifntSize; i++) {
@@ -1032,7 +1035,7 @@ int LZ4IO_decompressMultipleFilenames(const char** inFileNamesTable, int ifntSiz
             missingFiles += LZ4IO_decompressSrcFile(ress, inFileNamesTable[i], stdoutmark);
             continue;
         }
-        if (ofnSize <= ifnSize-suffixSize+1) { free(outFileName); ofnSize = ifnSize + 20; outFileName = (char*)malloc(ofnSize); if (outFileName==NULL) exit(1); }
+        if (ofnSize <= ifnSize-suffixSize+1) { free(outFileName); ofnSize = ifnSize + 20; outFileName = (char*)malloc(ofnSize); if (outFileName==NULL) return ifntSize; }
         if (ifnSize <= suffixSize  ||  strcmp(suffixPtr, suffix) != 0) {
             DISPLAYLEVEL(1, "File extension doesn't match expected LZ4_EXTENSION (%4s); will not process file: %s\n", suffix, inFileNamesTable[i]);
             skippedFiles++;