> From: [EMAIL PROTECTED]
> Date: Mon, 24 Jul 2000 14:16:20 +0200
>
> > From: [EMAIL PROTECTED]
> > Date: Fri, 21 Jul 2000 16:45:22 +0200
> >
> > So I tried using JDK 1.3 (latest refresh beta) just to check out how
> > Sun was doing. Sure, now it did not leak at all. Stayed at about
> > 15MB of memory footprint. But the performace was a joke! Same
> > test gave me about 50 method invocations per second. This was without
> > JRMP 1.3 though, maybe that will make a difference.
>
> Just FYI, JRMP1.3 does not improve this benchmark. I am downloading
> the IBM JDK 1.3 at the moment, I will get back when I have comparison
> figures for that one.
Done so, and it's remarkably differently behaving. The JIT-engine
sure shows, as startup times are really feeling a lot longer. It also
seems to be adaptive, i.e. it seems optimize more often used code.
Now I am only speculating, but my benchmark results are pointing in
that direction. The first 500 invocations averages out at 130
invocation per second, but if I keep running for like tens of
thousands of runs, I get over 300 invocations per second. I.e. 6
times the Sun JDK 1.3, and 2.5 times the Sun JDK 1.2.2. Very
interesting.
Oh, this was only the JVM I tested, now will go on and see if the
compiler too alters the picture. The tests above were carried out
with classes compiled by the SUN JDK 1.3.
It'll also be interesting to see how well-behaved IBM's implementation
is when it comes to be run in Linux-emulating environments, like
OpenBSD's.
Niklas
--
--------------------------------------------------------------
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Problems?: [EMAIL PROTECTED]
