Re: Jython performance

OK, I just made the change below [3] locally---I hope that was what Ian meant.

I've kicked off some performance runs at large and tight heaps (on the assumption that this change may have a memory footprint impact).  The data for the large heap is trickling in for 1st iteration [1] and 3rd iteration [2].  The data in those links will update every 20 mins or so.  The tight heap data will start appearing as a separate table at the bottom of each link once the large heap is finished.

Initial results for the large heap show a 3% win for VM_HashMap (13342HM) on first iteration and 4% on second iteration for jython, but results for other benchmarks are mixed.

Cheers,

--Steve

[1] http://cs.anu.edu.au/people/Steve.Blackburn/jikesrvm/vole-2007-08-19-Sun-161024-1.txt

[2] http://cs.anu.edu.au/people/Steve.Blackburn/jikesrvm/vole-2007-08-19-Sun-161024-3.txt

[3]

Index: libraryInterface/GNUClasspath/CPL/src/java/lang/VMString.java

===================================================================

--- libraryInterface/GNUClasspath/CPL/src/java/lang/VMString.java       (revision 13342)

+++ libraryInterface/GNUClasspath/CPL/src/java/lang/VMString.java       (working copy)

@@ -13,17 +13,16 @@

package java.lang;

import java.lang.ref.WeakReference;

-import java.util.WeakHashMap;

import org.jikesrvm.VM;

import org.jikesrvm.runtime.VM_Statics;

+import org.jikesrvm.util.VM_HashMap;

/**

  * Implementation of string interning for JikesRVM.

  */

final class VMString {

-  private static final WeakHashMap<String,WeakReference<String>> internedStrings =

-    new WeakHashMap<String,WeakReference<String>>();

+  private static final VM_HashMap<String,WeakReference<String>> internedStrings = new VM_HashMap<String,WeakReference<String>>();

   /**

    * Intern the argument string.