Re: [idvusers] Maximum Heap Size


I've decided to try to build IDV from the source code on a 64-bit computer to try and answer my own question. However, the only 64-bit computer that I have access to is a Windoze machine (running a 64-bit version of XP). I know that Ant is required to build the IDV from the source, and it looks like you can run Ant on windoze using Cygwin. I have some familiarity with Cygwin, but often seem to have problems with setting my paths, particularly when it comes to invoking java through Cygwin.

Could anyone help me with either A) "how to build IDV on a windoze box", or B) setting up my cygwin path?

Many thanks!!


Kevin Manross wrote:
Good thread!

It sounds as if the caching will at least allow me to load al my data in and it should be sufficient for data analysis. If we want to create long loops, we'll do that through ISL.

Getting back to what I understand about the Java heap, and what Valentijn has described, the limitation is dependent on:

Physical RAM
Architecture (32 bit vs. 64 bit)
Java Version(?)

However, even if I have a 64-bit machine with 4-8 Gb RAM running 64-bit version of WindozeXP (Or linux), the IDV comes with its own JVM. Will this limit my potential heap size? Would I be better off to install the 64-bit version of Java (Java3D?) and then download/build the IDV so it uses the computer's native JVM?



HansPeter Roesli wrote:
Hi all

This caching is what I was looking for since some time. Good to know it is there now. I am currently testing it with a long and heavy satellite image loop (keep fingers crossed).

I am running IDV on 2 notebooks, both with 2GB of RAM. One runs under XP and I cannot safely go beyond 1000m. On the other one under SuSE Linux 10.1 RAM is set to 1512m.

Cheers, HP

Valentijn Venus wrote:
Kevin. Jeff,

that new caching facility does sound promising!

As far as i know JAVA Heap spaces can grow unlimited on (OPen) Solaris systems, and probably also on Linux 64-bit with JAVA 64-bit but maybe someone else would be able to tell you...

Here are some of our experiences assigning space heap sizes all running IDV from a Java webstart jnlp:

-Linux 64-bit takes you up to 3.5 Gb
-Windows 32-bit varies (depending on what dll's are loaded at which memory address) roughly 1.5 Gb

see some of the jnlp's at at the "Download" section.

We'll have to wait for Dolphin JAVA virtual machine for large memory addressing to make it into JAVA webstart which by then will also support all the 64-bit goodies.

Jeff, would you be able to tell if a 64-bit version of java3d is in the making or will these two (32-bit java3d & 64-bit jvm) co-exist happily together?

Kind regards, Valentijn V. Venus, M.Sc. Researcher/Lecturer in RS/GIS for Food Security ITC
P.O.Box 6
7500 AA Enschede
The Netherlands
Tel: +31-53-4874549

=============================================================== De informatie opgenomen in dit bericht kan vertrouwelijk zijn en is uitsluitend bestemd voor de geadresseerde. Indien u dit bericht onterecht ontvangt wordt u verzocht de inhoud niet te gebruiken en de afzender direct te informeren. =============================================================== The information contained in this message may be confidential and is intended to be exclusively for the addressee. Should you receive this message unrightfully, please do not use the contents herein and notify the sender immediately by return e-mail. ===============================================================

-----Original Message-----
From: idvusers-bounces@xxxxxxxxxxxxxxxx on behalf of Jeff McWhirter
Sent: Wed 10/3/2007 23:18
To: Kevin L. Manross
Cc: IDV Users
Subject: Re: [idvusers] Maximum Heap Size
Hi Kevin,
I'm sure some folks will chime in with their experience with max memory size.
Specifically, I'm wanting to display relatively long animations of data displaying isosurfaces.
We have recently added a facility to cache the data to a local disk store. If you are running a recent build - for grids if you go to the properties dialog of the data source there is a "Always cache to disk" checkbox. Turning that on will result in the IDV reading in the data timestep by timestep right when each timestep is accessed, e.g., when the display is rendered. Once we read it we write it to disk and get rid of it from memory. Then, when the data is accessed again (say, when the display is re-rendered or the data is probed) we read it back from the local disk.

The "Delay" field is how long the IDV waits until it gets rid of the data from memory. For example, if you are probing the data you might want to increase the delay so the IDV doesn't keep going to disk every time the probe moves.

This is a user option because you do take a performance hit in the IO to local disk. But, it will dramatically reduce memory usage.

A caveat - this is fairly new the past 6 weeks or so but it seems to be working pretty well.


idvusers mailing list
For list information, to unsubscribe, visit:
idvusers mailing list
For list information, to unsubscribe, visit:

Kevin L. Manross           |  ** New Address **
CIMMS Research Associate   |     120 David L. Boren Bvd
NSSL : WRDD : SWAT         |     Rm 3923
<kevin.manross@xxxxxxxx>   |     405.325.6385 |
"My opinions are my own and not representative of
CIMMS, NSSL, NOAA or any affiliates"