-Xcompressedrefs performance

Jarkko Lietolahti jarkko.lietolahti at jab.fi
Wed Mar 11 12:47:56 PDT 2009


Hello,
 Well yes, this seems to be the case. -Xcompressedrefs is indeed interpreted
as -Xcomp which causes the slowdown. I was so excited about OpenJDK
supporting the compressedrefs I couldn't think straight anymore ;)

br,
 Jarkko

On Wed, Mar 11, 2009 at 9:40 PM, John Coomes <John.Coomes at sun.com> wrote:

> Jarkko Lietolahti (jarkko.lietolahti at jab.fi) wrote:
> > Hello,
> >  I originally posted this to hotspot-gc-use at openjdk.java.net.
> >  I just noticed that using -Xcompressedrefs causes bad performance
> issues.
> > This seems to be related to hotspot compiling. With -Xcompressedrefs my
> app
> > takes over 3 minutes to start and after removing only -Xcompressedrefs it
> > takes 1 minute and 23 seconds to start. So the issue seems to be related
> to
> > the flag -Xcompressedrefs.
> >
> > java version "1.6.0_0"
> > OpenJDK Runtime Environment (IcedTea6 1.4.1) (6b14-1.4.1-0ubuntu3)
> > OpenJDK 64-Bit Server VM (build 14.0-b08, mixed mode)
> >
> > Anyone else noticed something similar?
>
> AFAICT, -Xcompressedrefs is an option supported by the IBM virtual
> machine, not the Sun and OpenJDK VMs.  What I think is happening is
> that -Xcompressedrefs is being interpreted as -Xcomp, which is
> recognized by the Sun & OpenJDK VMs (we should be more strict in our
> argument parsing and reject -Xcompressedrefs).  -Xcomp means JIT
> compile every method the first time it is invoked.  The longer startup
> time you're seeing with -Xcomp is expected.
>
> When running in -Xcomp mode, the version string should say "compiled
> mode":
>
> $ java -Xcompextrajunkthatisignored -version
> java version "1.7.0-ea"
> Java(TM) SE Runtime Environment (build 1.7.0-ea-b48)
> Java HotSpot(TM) Server VM (build 15.0-b01, compiled mode)
>                                            ^^^^^^^^^^^^^
>
> -John
>
>


-- 
Jarkko Lietolahti
CTO
Hpj
+358 40 741 9973 jarkko.lietolahti at jab.fi
Jab Oy http://www.jab.fi/
Olympiastadion, Helsinki, Finland
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-dev/attachments/20090311/4c4d2438/attachment.html 


More information about the hotspot-dev mailing list