Rev 3080: The numbers are a bit noisy... in http://bzr.arbash-meinel.com/branches/bzr/0.93-dev/patience_tuples

[John Arbash Meinel]

At http://bzr.arbash-meinel.com/branches/bzr/0.93-dev/patience_tuples

------------------------------------------------------------
revno: 3080
revision-id:john at arbash-meinel.com-20071205140005-aexhlajexv5r1w3l
parent: john at arbash-meinel.com-20071205134925-4s1wo9k920cbnavb
committer: John Arbash Meinel <john at arbash-meinel.com>
branch nick: patience_tuples
timestamp: Wed 2007-12-05 08:00:05 -0600
message:
  The numbers are a bit noisy...
  Using djb2 hash and memcmp/PyObject_Compare ranges from: 3.5s-4.1s
  Using original code ranges from: 3.9s-4.1s
modified:
  bzrlib/_patiencediff_c.c       _patiencediff_c.c-20070721205602-q3imkipwlgagp3cy-1
-------------- next part --------------
=== modified file 'bzrlib/_patiencediff_c.c'
--- a/bzrlib/_patiencediff_c.c	2007-12-05 13:49:25 +0000
+++ b/bzrlib/_patiencediff_c.c	2007-12-05 14:00:05 +0000
@@ -587,15 +587,16 @@
              */
             line->len = PyString_GET_SIZE(item);
             line->c_data = p = PyString_AS_STRING(item);
-            // /* 'djb2' hash. This gives us a nice compromise between fast hash
-            //     function and a hash with less collisions. The algorithm doesn't
-            //     use the hash for actual lookups, only for building the table
-            //     so a better hash function wouldn't bring us much benefit, but
-            //     would make this loading code slower. */
-            // hash = 5381;
-            // for (j = 0; j < line->len; j++)
-            //     hash = ((hash << 5) + hash) + *p++;
-            line->hash = PyObject_Hash(item);
+            /* 'djb2' hash. This gives us a nice compromise between fast hash
+                function and a hash with less collisions. The algorithm doesn't
+                use the hash for actual lookups, only for building the table
+                so a better hash function wouldn't bring us much benefit, but
+                would make this loading code slower. */
+            hash = 5381;
+            for (j = 0; j < line->len; j++)
+                hash = ((hash << 5) + hash) + *p++;
+            line->hash = hash;
+            // line->hash = PyObject_Hash(item);
         } else if (PyTuple_Check(item)) {
             total_len = 0;
             tuple_len = PyObject_Length(item);