*** empty log message ***

3 files changed, 39 insertions, 15 deletions

diff --git a/Changes b/Changes
index db70df9..09cd72b 100644
--- a/Changes
+++ b/Changes
@@ -6,6 +6,8 @@
           and only a small win on amd cpus.
         - improve c++ compatibility of the code.
         - slightly improve compressor speed.
+        - halved memory requirements for compressor on 64 bit architectures,
+          which can improve the speed quite a bit on older cpus.
 
 3.5  Fri May  1 02:28:42 CEST 2009
 	- lzf_compress did sometimes write one octet past the given output
diff --git a/lzfP.h b/lzfP.h
index 7f0e02b..c54d087 100644
--- a/lzfP.h
+++ b/lzfP.h
@@ -121,27 +121,52 @@
 # define CHECK_INPUT 1
 #endif
 
+/*
+ * Whether to store pointers or offsets inside the hash table. On
+ * 64 bit architetcures, pointers take up twice as much space,
+ * and might also be slower. Default is to autodetect.
+ */
+/*#define LZF_USER_OFFSETS autodetect */
+
 /*****************************************************************************/
 /* nothing should be changed below */
 
 #ifdef __cplusplus
 # include <cstring>
+# include <climits>
 using namespace std;
 #else
 # include <string.h>
+# include <limits.h>
+#endif
+
+#ifndef LZF_USE_OFFSETS
+# if defined (WIN32)
+#  define LZF_USE_OFFSETS defined(_M_X64)
+# else
+#  ifdef __cplusplus
+#   include <cstdint>
+#  else
+#   include <stdint.h>
+#  endif
+#  define LZF_USE_OFFSETS (UINTPTR_MAX > 0xffffffffU)
+# endif
 #endif
 
 typedef unsigned char u8;
 
-typedef const u8 *LZF_STATE[1 << (HLOG)];
+#if LZF_USE_OFFSETS
+# define LZF_HSLOT_BIAS ((const u8 *)in_data)
+  typedef unsigned int LZF_HSLOT;
+#else
+# define LZF_HSLOT_BIAS 0
+  typedef const u8 *LZF_HSLOT;
+#endif
+
+typedef LZF_HSLOT LZF_STATE[1 << (HLOG)];
 
 #if !STRICT_ALIGN
 /* for unaligned accesses we need a 16 bit datatype. */
-# ifdef __cplusplus
-#  include <climits>
-# else
-#  include <limits.h>
-# endif
 # if USHRT_MAX == 65535
     typedef unsigned short u16;
 # elif UINT_MAX == 65535
diff --git a/lzf_c.c b/lzf_c.c
index 06afa63..e3b1108 100644
--- a/lzf_c.c
+++ b/lzf_c.c
@@ -106,7 +106,6 @@ lzf_compress (const void *const in_data, unsigned int in_len,
 #if !LZF_STATE_ARG
   LZF_STATE htab;
 #endif
-  const u8 **hslot;
   const u8 *ip = (const u8 *)in_data;
         u8 *op = (u8 *)out_data;
   const u8 *in_end  = ip + in_len;
@@ -133,10 +132,6 @@ lzf_compress (const void *const in_data, unsigned int in_len,
 
 #if INIT_HTAB
   memset (htab, 0, sizeof (htab));
-# if 0
-  for (hslot = htab; hslot < htab + HSIZE; hslot++)
-    *hslot++ = ip;
-# endif
 #endif
 
   lit = 0; op++; /* start run */
@@ -144,9 +139,11 @@ lzf_compress (const void *const in_data, unsigned int in_len,
   hval = FRST (ip);
   while (ip < in_end - 2)
     {
+      LZF_HSLOT *hslot;
+
       hval = NEXT (hval, ip);
       hslot = htab + IDX (hval);
-      ref = *hslot; *hslot = ip;
+      ref = *hslot + LZF_HSLOT_BIAS; *hslot = ip - LZF_HSLOT_BIAS;
 
       if (1
 #if INIT_HTAB
@@ -236,12 +233,12 @@ lzf_compress (const void *const in_data, unsigned int in_len,
           hval = FRST (ip);
 
           hval = NEXT (hval, ip);
-          htab[IDX (hval)] = ip;
+          htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
           ip++;
 
 # if VERY_FAST && !ULTRA_FAST
           hval = NEXT (hval, ip);
-          htab[IDX (hval)] = ip;
+          htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
           ip++;
 # endif
 #else
@@ -250,7 +247,7 @@ lzf_compress (const void *const in_data, unsigned int in_len,
           do
             {
               hval = NEXT (hval, ip);
-              htab[IDX (hval)] = ip;
+              htab[IDX (hval)] = ip - LZF_HSLOT_BIAS;
               ip++;
             }
           while (len--);