This is the mail archive of the
java@gcc.gnu.org
mailing list for the Java project.
Re: AccessController speedup
- From: Andrew Haley <aph at redhat dot com>
- To: "Scott Gilbertson" <scottg at mantatest dot com>
- Cc: "Gary Benson" <gbenson at redhat dot com>, <kgallowa at redhat dot com>, <java at gcc dot gnu dot org>
- Date: Tue, 22 Aug 2006 16:58:17 +0100
- Subject: Re: AccessController speedup
- References: <20060814142608.GA10501@redhat.com> <005f01c6c603$21fa6f60$3c16a8c0@scott>
Scott Gilbertson writes:
> From: "Gary Benson"
> > Classpath's VMAccessController uses ThreadLocal objects to store per-
> > thread state information. Both Classpath and GCJ have a pure Java
> > ThreadLocal implementation which results in a lot of extra GC activity
> > with a security manager. This commit changes VMAccessController to
> > use an instance variable in Thread to store its state. On my Tomcat
> > benchmark this improves performace from 2400 to 2700 requests per
> > second.
>
> I'm getting a segmentation fault that could be related to the
> recent unwinder changes. I have the core dump and binary that
> generated the backtrace below. The problem happens with both
> static and dynamic executables.
>
> Is there any information I can provide (memory dumps or whatever)
> that would help to determine what's happening?
>
> Do you suppose "--disable-tls" or "--enable-sjlj-exceptions" has
> anything to do with it?
I suspect so. Don't use sjlj on Linux.
Andrew.