Re: AS/400 and Java

["David Morris" <David.Morris@xxxxxxxxxxxxx>]

Mark,

On our system there are definitely cases where static compiles make
sense. In
my testing, individual class files are almost always better off being
compiled and
this is on V5R1 and V5R2. One thing the JIT can't do today is
anticipate how you
intend to use a piece of software. With Sun's JIT, they support
switches that
allow you to tune the JIT process. Changing to server or adjusting the
hit count
to compile makes a huge difference in overall and perceived
performance.

David Morris

>>> MarkP@softlanding.com 11/04/02 08:21AM >>>
This is a multipart message in MIME format.
--
[ Picked text/plain from multipart/alternative ]
Mike,

Actually most of the IBM'ers I have heard speak about this issue say
that
the JIT compiler is the direction of the future.  It already produces
code
that is comparable to optimization level 40 and will only get better.
Application servers like WebSphere and Tomcat are unable to take
advantage
of the output of CRTJVAPGM so a JIT became necessary.  Much to the
surprise of the members of the VM team that thought static compilation
would be the best result, the JIT produces amazing results.  This is
due
to a lot of incredible patented technologies that IBM has developed in
their research labs.

I use the JIT exclusively on a V4R5 system with great results and it
is
supposed to be much much faster at V5R1 and V5R2.

I suspect that CRTJVAPGM will be phased out in future releases unless
they
can come up with some ways to make it relevant again.  It doesn't
really
make sense to waste engineering effort on it, however, when a JIT can
produce such great results.

Mark