improve initial literals stats

diff --git a/lib/compress/zstd_compress_literals.c b/lib/compress/zstd_compress_literals.c
index 52b0a8059aba959ead63fbdf3033a72fc31f3831..438aee80bb81f3d62152a7b3352cf599a2598bf1 100644 (file)
--- a/lib/compress/zstd_compress_literals.c
+++ b/lib/compress/zstd_compress_literals.c
@@ -93,7 +93,6 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
         return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
 
     /* small ? don't even attempt compression (speed opt) */
-#   define COMPRESS_LITERALS_SIZE_MIN 63
     {   size_t const minLitSize = (prevHuf->repeatMode == HUF_repeat_valid) ? 6 : COMPRESS_LITERALS_SIZE_MIN;
         if (srcSize <= minLitSize) return ZSTD_noCompressLiterals(dst, dstCapacity, src, srcSize);
     }
@@ -136,7 +135,7 @@ size_t ZSTD_compressLiterals (ZSTD_hufCTables_t const* prevHuf,
     switch(lhSize)
     {
     case 3: /* 2 - 2 - 10 - 10 */
-        {   U32 const lhc = hType + ((!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
+        {   U32 const lhc = hType + (((U32)!singleStream) << 2) + ((U32)srcSize<<4) + ((U32)cLitSize<<14);
             MEM_writeLE24(ostart, lhc);
             break;
         }

diff --git a/lib/compress/zstd_compress_literals.h b/lib/compress/zstd_compress_literals.h
index 9775fb97cb70250058b1e0767fd1b7a8967fc44c..94e06d9e3ab688a63ee6fc7c1bbf197469c9ffc7 100644 (file)
--- a/lib/compress/zstd_compress_literals.h
+++ b/lib/compress/zstd_compress_literals.h
@@ -14,6 +14,11 @@
 #include "zstd_compress_internal.h" /* ZSTD_hufCTables_t, ZSTD_minGain() */
 
 
+/* Below this limit, the heuristic does not even attempt compression of literals,
+ * as the cost of the headers is expected to outweight the benefits.
+ * This limit is not applicable when re-using statistics from dictionary or previous block */
+#define COMPRESS_LITERALS_SIZE_MIN 63
+
 size_t ZSTD_noCompressLiterals (void* dst, size_t dstCapacity, const void* src, size_t srcSize);
 
 size_t ZSTD_compressRleLiteralsBlock (void* dst, size_t dstCapacity, const void* src, size_t srcSize);

diff --git a/lib/compress/zstd_opt.c b/lib/compress/zstd_opt.c
index 532edd5bc904991f7d1ace1f6b3870176a1ee998..6345f81d27f7d5a17531bc83b83ef366a42db152 100644 (file)
--- a/lib/compress/zstd_opt.c
+++ b/lib/compress/zstd_opt.c
@@ -1253,6 +1253,31 @@ size_t ZSTD_compressBlock_btopt(
 }
 
 
+#include "zstd_compress_literals.h"   /* COMPRESS_LITERALS_SIZE_MIN */
+
+static void transfer_LiteralStats(optState_t* opt, const seqStore_t* seqStore)
+{
+    size_t const nbLiterals = (size_t)(seqStore->lit - seqStore->litStart);
+    assert(seqStore->lit >= seqStore->litStart);
+
+    // test if literals are compressible
+    //char block[ZSTD_BLOCKSIZE_MAX];
+    //size_t const cLitSize = HUF_compress(block, sizeof(block), seqStore->litStart, nbLiterals);
+    //printf("nb literals = %zu  => %zu huf_compress\n", nbLiterals, cLitSize);
+
+    if (nbLiterals < COMPRESS_LITERALS_SIZE_MIN) {
+        /* literals won't be compressed anyway, give them a flat cost */
+        /* note : it would be better if it was also possible to extend this category
+         * to non-compressible literals which are more numerous than threshold */
+        unsigned u;
+        for (u=0; u<=MaxLit; u++) opt->litFreq[u]=2;
+    } else {
+        unsigned maxlit = MaxLit;
+        HIST_count_simple(opt->litFreq, &maxlit, seqStore->litStart, nbLiterals);
+    }
+    opt->litSum = ZSTD_downscaleStats(opt->litFreq, MaxLit, 0);  /* flatten stats, by providing at least 1 to every symbol */
+}
+
 
 #include "zstd_lazy.h"
 /* ZSTD_initStats_greedy():
@@ -1275,16 +1300,16 @@ ZSTD_initStats_greedy(ZSTD_matchState_t* ms,
     assert(ms->window.dictLimit == ms->window.lowLimit);   /* no dictionary */
     assert(ms->window.dictLimit - ms->nextToUpdate <= 1);  /* no prefix (note: intentional overflow, defined as 2-complement) */
 
-    ZSTD_compressBlock_greedy(ms, seqStore, tmpRep, src, srcSize);   /* generate stats into seqstore */
-
-    /* transfer stats into ms-opt */
-    /* literals stats */
-    {   unsigned maxlit = MaxLit;
-        assert(seqStore->lit >= seqStore->litStart);
-        HIST_count_simple(ms->opt.litFreq, &maxlit, seqStore->litStart, (size_t)(seqStore->lit - seqStore->litStart));
-        ms->opt.litSum = ZSTD_downscaleStats(ms->opt.litFreq, MaxLit, 0);  /* flatten stats, by providing at least 1 to every symbol */
+    {   size_t const lastLits = ZSTD_compressBlock_greedy(ms, seqStore, tmpRep, src, srcSize);   /* generate stats into seqstore */
+        /* add last lits into literals buffers for proper accounting */
+        assert(lastLits <= srcSize);
+        ZSTD_memcpy(seqStore->lit , (const char*)src + srcSize - lastLits, lastLits);
+        seqStore->lit += lastLits;
     }
 
+    /* transfer stats from seqStore into ms-opt */
+    transfer_LiteralStats(&ms->opt, seqStore);
+
     /* seqStats */
     assert(seqStore->sequences >= seqStore->sequencesStart);
     {   U32 const nbSeq = (U32)(seqStore->sequences - seqStore->sequencesStart);
@@ -1367,6 +1392,7 @@ ZSTD_initStats_ultra(ZSTD_matchState_t* ms,
                      U32 rep[ZSTD_REP_NUM],
                const void* src, size_t srcSize)
 {
+    size_t lastLits;
     U32 tmpRep[ZSTD_REP_NUM];  /* updated rep codes will sink here */
     ZSTD_memcpy(tmpRep, rep, sizeof(tmpRep));
 
@@ -1378,13 +1404,19 @@ ZSTD_initStats_ultra(ZSTD_matchState_t* ms,
 
     if (srcSize <= 16 KB) {
         /* raw btultra, initialized by default starting stats */
-        ZSTD_compressBlock_opt_generic(ms, seqStore, tmpRep, src, srcSize, 2 /*optLevel*/, ZSTD_noDict);  /* generate stats into ms->opt*/
+        lastLits = ZSTD_compressBlock_opt_generic(ms, seqStore, tmpRep, src, srcSize, 2 /*optLevel*/, ZSTD_noDict);  /* generate stats into ms->opt*/
     } else {
         /* in this mode, btultra is initialized greedy;
          * measured better for larger blocks, but not for small ones */
-        ZSTD_compressBlock_btultra(ms, seqStore, tmpRep, src, srcSize);  /* generate stats into ms->opt*/
+        lastLits = ZSTD_compressBlock_btultra(ms, seqStore, tmpRep, src, srcSize);  /* generate stats into ms->opt*/
     }
 
+    /* transfer stats from seqStore into ms-opt */
+    assert(lastLits <= srcSize);
+    ZSTD_memcpy(seqStore->lit , (const char*)src + srcSize - lastLits, lastLits);
+    seqStore->lit += lastLits;
+    transfer_LiteralStats(&ms->opt, seqStore);
+
     /* invalidate first scan from history */
     ZSTD_resetSeqStore(seqStore);
     ms->window.base -= srcSize;