X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=include%2Fwimlib%2Flz_extend.h;h=2fb76bc921d47c5b1986c087b1e12e24cd7b62ac;hp=7d7ed8b8e88b02ceb6ebdd684ed1dddd33f92a08;hb=0ec3ead8ebd29703e12342c56b282ae37b188e6d;hpb=7e3f3761e0a9fc93341ed0b9c69f6056d9a97af9

diff --git a/include/wimlib/lz_extend.h b/include/wimlib/lz_extend.h
index 7d7ed8b8..2fb76bc9 100644
--- a/include/wimlib/lz_extend.h
+++ b/include/wimlib/lz_extend.h
@@ -1,73 +1,52 @@
 /*
- * lz_extend.h
+ * lz_extend.h - fast match extension for Lempel-Ziv matchfinding
  *
- * Fast match extension for Lempel-Ziv matchfinding.
+ * The following copying information applies to this specific source code file:
  *
- * The author dedicates this file to the public domain.
- * You can do whatever you want with this file.
+ * Written in 2014-2016 by Eric Biggers <ebiggers3@gmail.com>
+ *
+ * To the extent possible under law, the author(s) have dedicated all copyright
+ * and related and neighboring rights to this software to the public domain
+ * worldwide via the Creative Commons Zero 1.0 Universal Public Domain
+ * Dedication (the "CC0").
+ *
+ * This software is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE. See the CC0 for more details.
+ *
+ * You should have received a copy of the CC0 along with this software; if not
+ * see <http://creativecommons.org/publicdomain/zero/1.0/>.
  */
 
 #ifndef _WIMLIB_LZ_EXTEND_H
 #define _WIMLIB_LZ_EXTEND_H
 
-#include "wimlib/types.h"
-
-#if (defined(__x86_64__) || defined(__i386__)) && defined(__GNUC__)
-#  define HAVE_FAST_LZ_EXTEND 1
-#else
-#  define HAVE_FAST_LZ_EXTEND 0
-#endif
+#include "wimlib/bitops.h"
+#include "wimlib/unaligned.h"
 
-/* Return the number of bytes at @matchptr that match the bytes at @strptr, up
- * to a maximum of @max_len.  Initially, @start_len bytes are matched.  */
+/*
+ * Return the number of bytes at @matchptr that match the bytes at @strptr, up
+ * to a maximum of @max_len.  Initially, @len bytes are matched.
+ */
 static inline u32
 lz_extend(const u8 * const strptr, const u8 * const matchptr,
-	  const u32 start_len, const u32 max_len)
+	  u32 len, const u32 max_len)
 {
-	u32 len = start_len;
-
-#if HAVE_FAST_LZ_EXTEND
-
-	while (len + sizeof(unsigned long) <= max_len) {
-		unsigned long x;
-
-		x = *(const unsigned long *)&matchptr[len] ^
-		    *(const unsigned long *)&strptr[len];
-		if (x != 0)
-			return len + (__builtin_ctzl(x) >> 3);
-		len += sizeof(unsigned long);
-	}
-
-	if (sizeof(unsigned int) < sizeof(unsigned long) &&
-	    len + sizeof(unsigned int) <= max_len)
-	{
-		unsigned int x;
-
-		x = *(const unsigned int *)&matchptr[len] ^
-		    *(const unsigned int *)&strptr[len];
-		if (x != 0)
-			return len + (__builtin_ctz(x) >> 3);
-		len += sizeof(unsigned int);
-	}
-
-	if (sizeof(unsigned int) == 4) {
-		if (len < max_len && matchptr[len] == strptr[len]) {
-			len++;
-			if (len < max_len && matchptr[len] == strptr[len]) {
-				len++;
-				if (len < max_len && matchptr[len] == strptr[len]) {
-					len++;
-				}
-			}
+	while (UNALIGNED_ACCESS_IS_FAST && len + WORDSIZE <= max_len) {
+		machine_word_t v = load_word_unaligned(matchptr + len) ^
+				   load_word_unaligned(strptr + len);
+		if (v != 0) {
+			if (CPU_IS_LITTLE_ENDIAN)
+				len += ffsw(v) >> 3;
+			else
+				len += (8 * WORDSIZE - 1 - flsw(v)) >> 3;
+			return len;
 		}
-		return len;
+		len += WORDSIZE;
 	}
 
-#endif /* HAVE_FAST_LZ_EXTEND */
-
 	while (len < max_len && matchptr[len] == strptr[len])
 		len++;
-
 	return len;
 }