[news.eclipse.tools.hyades] Re: Notes of meeting of December 19 2002

[sluiman@xxxxxxxxxx (Harm Sluiman)]

Martin van den Bemt wrote:

> This is a multi-part message in MIME format.

> ------=_NextPart_000_0061_01C2BAF0.F7370250
> Content-Type: text/plain;
> 	charset="iso-8859-1"
> Content-Transfer-Encoding: quoted-printable

> ElementHave you ever considered combining the java profiling and =
> debugging information ? This way you can see if objects previously =
> created are eg actually released during debugging and doesn't need a =
> seperate application to do things twice.. Based on the elements =
> specified in the list below, i thought this approach may be of use for =
> hyades.

> Mvgr,
> Martin

Hello Martin.
Yes this is a very good point and we have looked at it. Currently running
under debug a JVM performance is affected greatly because of all the
things that might need to be done. This is via JVMDI, many profiling tools
use JVMPI to do profiling data collection, which in most cases can not be
done at the same time as using JVMDI. An alternate for profiling has been
to use byte code insertion which has it's own strength and weaknesses.
We will provide an JVMPI collection system in this project soon, and we
are also intending to provide a standard client that follows the new JVM
interfaces that will replace JVMPI and other interface to do much better
collection.
I guess another way to look at your suggestion is to extend debuggers with
heap trace and analysis support, and some could do that. It is something
we should look at as we go down this road. Even just snapshoting the Heap,
which is one of the things have code to do, while the debugger is attached
might be a help.
Thanks for the idea.