Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 16 Mar 2000 10:04:25 -0700 (MST)
From:      Nate Williams <nate@yogotech.com>
To:        Andrzej Bialecki <abial@webgiro.com>
Cc:        Nate Williams <nate@yogotech.com>, Craig Shaver <craig@ProGroup.COM>, FreeBSD ISP List <freebsd-isp@FreeBSD.ORG>
Subject:   Re: Java and FreeBSD
Message-ID:  <200003161704.KAA14557@nomad.yogotech.com>
In-Reply-To: <Pine.BSF.4.20.0003161754330.82593-100000@mx.webgiro.com>
References:  <200003161604.JAA14236@nomad.yogotech.com> <Pine.BSF.4.20.0003161754330.82593-100000@mx.webgiro.com>

next in thread | previous in thread | raw e-mail | index | archive | help
> > The reason for this is because the Blackdown team was given access to
> > the latest SUN JIT.  It has very little to do with JDK1.2 and everything
> > to do with using a well designed JIT.
> 
> > Again, this has little to do with it, and the biggest reason is because
> 
> Really? I thought that native kernel threads would help for IO bound
> application..

Sometimes it does, and sometimes it doesn't.  My experience with *LOTS*
of I/O networking threads is that the green threads model works better
than the native threads model, since select() can be used so that I/O
bound threads 'Do The Right Thing' and don't block.  Because of this,
the *MUCH* lower context switch time is a boon, and the system tends to
run much faster.

However, file and/or device I/O (I used serial code classes I bought)
cause the JVM to be almost unusable, but I think the latter was due to
poorly written native code for reading the serial device.
Unfortunately, I had neither the time nor the energy to re-write the
serial code.
> 
> > of the JIT.  For a much better performing JDK 1.1.8, try using one of
> > the JIT's in the ports area, either TYA or ShuJIT.  I've had good look
> > with the former...
> 
> My point was that even with one of those JITs the performance is less than
> breathtaking.

Sure, but it's not that much different than the performance using the
Blackdown JVM either.  Seriously!

> Don't take it personally, please - I know you spent
> tremendous efforts to make Java available on FreeBSD, and I admire you for
> that. But still, it's easy to make VolanoMark benchmark and see the
> results...

I'd love to see someone doing a real comparison using the Volano
benchmark of the native JDK + TYA against the Blackdown JIT.



Nate


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200003161704.KAA14557>