Update hash table for first byte at start of LZS compression loop

We're always going to need to do it, so do it while we have everything
to hand. Roughly 19% speed improvement on the random-biased corpus;
perhaps a little less if there are more matches and we end up having
to loop anyway.

Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

lzs.c

@@ -218,6 +218,9 @@ int lzs_compress(unsigned char *dst, int dstlen, const unsigned char *src, int s
 		hash = HASH(src + inpos);
 		hofs = hash_table[hash];
 
+		hash_chain[inpos & (MAX_HISTORY - 1)] = hofs;
+		hash_table[hash] = inpos;
+
 		longest_match_len = 0;
 
 		while (hofs != INVALID_OFS && hofs + MAX_HISTORY > inpos) {
@@ -230,8 +233,7 @@ int lzs_compress(unsigned char *dst, int dstlen, const unsigned char *src, int s
 		}
 		if (!longest_match_len) {
 			PUT_BITS(9, src[inpos]);
-			hash_chain[inpos & (MAX_HISTORY - 1)] = hash_table[hash];
-			hash_table[hash] = inpos++;
+			inpos++;
 		} else {
 			/* Output offset, as 7-bit or 11-bit as appropriate */
 			int offset = inpos - longest_match_ofs;
@@ -259,13 +261,15 @@ int lzs_compress(unsigned char *dst, int dstlen, const unsigned char *src, int s
 					PUT_BITS(4, length);
 			}
 
-			/* Add byte(s) to the hash tables unless we're done */
+			/* If we're already done, don't bother updating the hash tables. */
 			if (inpos + longest_match_len >= srclen - 1) {
 				inpos += longest_match_len;
 				break;
 			}
 
-			while (longest_match_len--) {
+			/* We already added the first byte to the hash tables. Add the rest. */
+			inpos++;
+			while (--longest_match_len) {
 				hash = HASH(src + inpos);
 				hash_chain[inpos & (MAX_HISTORY - 1)] = hash_table[hash];
 				hash_table[hash] = inpos++;