From: Archie Cobbs (archie@whistle.com)
Date: Tue Oct 13 1998 - 14:09:15 EDT
Godmar Back writes:
> Hi, while going through the recent changes, I noticed that class.final
> has disappeared. I believe we're taking a big hit by doing so.
> I know for sure that we lose bitterly in artificial gc benchmarks;
> I would suspect some impact under normal use as well.
>
> Basically, what needs to be done is to analyze whether a class has a
> non-empty finalizer (in most cases, classes will inherit the empty
> finalizer of java.lang.Object), and if so, set class.final and finalize.
> Whether a finalizer is empty or not should be simply inferred from the
> length of the code---or am I missing something here?
I think Tim said the new approach is to finalize every object; for
classes without finalizers, they just inherit the empty
java.lang.Object.finalize() which does nothing.
You're probably right in that this is an easy (and small in scope)
thing to optimize... I'd be interested in seeing a benchmark comparison
of before and after if you got one.
-Archie
___________________________________________________________________________
Archie Cobbs * Whistle Communications, Inc. * http://www.whistle.com
This archive was generated by hypermail 2b29 : Sat Sep 23 2000 - 19:57:00 EDT