Add LIT_MEM define to use more memory for a small deflate speedup.

A bug fix in zlib 1.2.12 resulted in a slight slowdown (1-2%) of
deflate. This commit provides the option to #define LIT_MEM, which
uses more memory to reverse most of that slowdown. The memory for
the pending buffer and symbol buffers is increased by 25%, which
increases the total memory usage with the default parameters by
about 6%.

deflate.c

@@ -493,7 +493,11 @@ int ZEXPORT deflateInit2_(z_streamp strm, int level, int method,
      * symbols from which it is being constructed.
      */
 
+#ifdef LIT_MEM
+    s->pending_buf = (uchf *) ZALLOC(strm, s->lit_bufsize, 5);
+#else
     s->pending_buf = (uchf *) ZALLOC(strm, s->lit_bufsize, 4);
+#endif
     s->pending_buf_size = (ulg)s->lit_bufsize * 4;
 
     if (s->window == Z_NULL || s->prev == Z_NULL || s->head == Z_NULL ||
@@ -503,8 +507,14 @@ int ZEXPORT deflateInit2_(z_streamp strm, int level, int method,
         deflateEnd (strm);
         return Z_MEM_ERROR;
     }
+#ifdef LIT_MEM
+    s->d_buf = (ushf *)(s->pending_buf + (s->lit_bufsize << 1));
+    s->l_buf = s->pending_buf + (s->lit_bufsize << 2);
+    s->sym_end = s->lit_bufsize - 1;
+#else
     s->sym_buf = s->pending_buf + s->lit_bufsize;
     s->sym_end = (s->lit_bufsize - 1) * 3;
+#endif
     /* We avoid equality with lit_bufsize*3 because of wraparound at 64K
      * on 16 bit machines and because stored blocks are restricted to
      * 64K-1 bytes.
@@ -720,9 +730,15 @@ int ZEXPORT deflatePrime(z_streamp strm, int bits, int value) {
 
     if (deflateStateCheck(strm)) return Z_STREAM_ERROR;
     s = strm->state;
+#ifdef LIT_MEM
+    if (bits < 0 || bits > 16 ||
+        (uchf *)s->d_buf < s->pending_out + ((Buf_size + 7) >> 3))
+        return Z_BUF_ERROR;
+#else
     if (bits < 0 || bits > 16 ||
         s->sym_buf < s->pending_out + ((Buf_size + 7) >> 3))
         return Z_BUF_ERROR;
+#endif
     do {
         put = Buf_size - s->bi_valid;
         if (put > bits)
@@ -1308,7 +1324,12 @@ int ZEXPORT deflateCopy(z_streamp dest, z_streamp source) {
     zmemcpy(ds->pending_buf, ss->pending_buf, (uInt)ds->pending_buf_size);
 
     ds->pending_out = ds->pending_buf + (ss->pending_out - ss->pending_buf);
+#ifdef LIT_MEM
+    ds->d_buf = (ushf *)(ds->pending_buf + (ds->lit_bufsize << 1));
+    ds->l_buf = ds->pending_buf + (ds->lit_bufsize << 2);
+#else
     ds->sym_buf = ds->pending_buf + ds->lit_bufsize;
+#endif
 
     ds->l_desc.dyn_tree = ds->dyn_ltree;
     ds->d_desc.dyn_tree = ds->dyn_dtree;

deflate.h

@@ -23,6 +23,10 @@
 #  define GZIP
 #endif
 
+/* define LIT_MEM to slightly increase the speed of deflate (order 1% to 2%) at
+   the cost of a larger memory footprint */
+/* #define LIT_MEM */
+
 /* ===========================================================================
  * Internal compression state.
  */
@@ -217,7 +221,12 @@ typedef struct internal_state {
     /* Depth of each subtree used as tie breaker for trees of equal frequency
      */
 
+#ifdef LIT_MEM
+    ushf *d_buf;          /* buffer for distances */
+    uchf *l_buf;          /* buffer for literals/lengths */
+#else
     uchf *sym_buf;        /* buffer for distances and literals/lengths */
+#endif
 
     uInt  lit_bufsize;
     /* Size of match buffer for literals/lengths.  There are 4 reasons for
@@ -239,7 +248,7 @@ typedef struct internal_state {
      *   - I can't count above 4
      */
 
-    uInt sym_next;      /* running index in sym_buf */
+    uInt sym_next;      /* running index in symbol buffer */
     uInt sym_end;       /* symbol table full when sym_next reaches this */
 
     ulg opt_len;        /* bit length of current block with optimal trees */
@@ -318,6 +327,25 @@ void ZLIB_INTERNAL _tr_stored_block(deflate_state *s, charf *buf,
   extern const uch ZLIB_INTERNAL _dist_code[];
 #endif
 
+#ifdef LIT_MEM
+# define _tr_tally_lit(s, c, flush) \
+  { uch cc = (c); \
+    s->d_buf[s->sym_next] = 0; \
+    s->l_buf[s->sym_next++] = cc; \
+    s->dyn_ltree[cc].Freq++; \
+    flush = (s->sym_next == s->sym_end); \
+   }
+# define _tr_tally_dist(s, distance, length, flush) \
+  { uch len = (uch)(length); \
+    ush dist = (ush)(distance); \
+    s->d_buf[s->sym_next] = dist; \
+    s->l_buf[s->sym_next++] = len; \
+    dist--; \
+    s->dyn_ltree[_length_code[len]+LITERALS+1].Freq++; \
+    s->dyn_dtree[d_code(dist)].Freq++; \
+    flush = (s->sym_next == s->sym_end); \
+  }
+#else
 # define _tr_tally_lit(s, c, flush) \
   { uch cc = (c); \
     s->sym_buf[s->sym_next++] = 0; \
@@ -337,6 +365,7 @@ void ZLIB_INTERNAL _tr_stored_block(deflate_state *s, charf *buf,
     s->dyn_dtree[d_code(dist)].Freq++; \
     flush = (s->sym_next == s->sym_end); \
   }
+#endif
 #else
 # define _tr_tally_lit(s, c, flush) flush = _tr_tally(s, 0, c)
 # define _tr_tally_dist(s, distance, length, flush) \

trees.c

@@ -899,14 +899,19 @@ local void compress_block(deflate_state *s, const ct_data *ltree,
                           const ct_data *dtree) {
     unsigned dist;      /* distance of matched string */
     int lc;             /* match length or unmatched char (if dist == 0) */
-    unsigned sx = 0;    /* running index in sym_buf */
+    unsigned sx = 0;    /* running index in symbol buffers */
     unsigned code;      /* the code to send */
     int extra;          /* number of extra bits to send */
 
     if (s->sym_next != 0) do {
+#ifdef LIT_MEM
+        dist = s->d_buf[sx];
+        lc = s->l_buf[sx++];
+#else
         dist = s->sym_buf[sx++] & 0xff;
         dist += (unsigned)(s->sym_buf[sx++] & 0xff) << 8;
         lc = s->sym_buf[sx++];
+#endif
         if (dist == 0) {
             send_code(s, lc, ltree); /* send a literal byte */
             Tracecv(isgraph(lc), (stderr," '%c' ", lc));
@@ -931,8 +936,12 @@ local void compress_block(deflate_state *s, const ct_data *ltree,
             }
         } /* literal or match pair ? */
 
-        /* Check that the overlay between pending_buf and sym_buf is ok: */
+        /* Check for no overlay of pending_buf on needed symbols */
+#ifdef LIT_MEM
+        Assert(s->pending < (s->lit_bufsize << 1) + sx, "pendingBuf overflow");
+#else
         Assert(s->pending < s->lit_bufsize + sx, "pendingBuf overflow");
+#endif
 
     } while (sx < s->sym_next);
 
@@ -1082,9 +1091,14 @@ void ZLIB_INTERNAL _tr_flush_block(deflate_state *s, charf *buf,
  * the current block must be flushed.
  */
 int ZLIB_INTERNAL _tr_tally(deflate_state *s, unsigned dist, unsigned lc) {
+#ifdef LIT_MEM
+    s->d_buf[s->sym_next] = (ush)dist;
+    s->l_buf[s->sym_next++] = (uch)lc;
+#else
     s->sym_buf[s->sym_next++] = (uch)dist;
     s->sym_buf[s->sym_next++] = (uch)(dist >> 8);
     s->sym_buf[s->sym_next++] = (uch)lc;
+#endif
     if (dist == 0) {
         /* lc is the unmatched char */
         s->dyn_ltree[lc].Freq++;