Re: [epp-dev] MAT in Java-based packages

[Doug Schaefer <dschaefer@xxxxxxxxxxxxxx>]

When indexing very large projects CDT produces a very large index. It’s a single file with a hand rolled database. We get around OOM errors by memory mapping chunks of the file and limiting how much we map at a given time. We had thought it would slow down access but it’s hardly noticeable.

Mind you with 64-bit now, in theory, you could memory map the whole thing. It doesn’t count against the Java heap, just the process address space. If MAT isn’t doing that, we should spend some time investing in it so that it does.

Doug.

From: <epp-dev-bounces@xxxxxxxxxxx> on behalf of Gunnar Wagenknecht <gunnar@xxxxxxxxxxxxxxx>
Reply-To: Eclipse Packaging Project <epp-dev@xxxxxxxxxxx>
Date: Wednesday, August 24, 2016 at 5:21 AM
To: Eclipse Packaging Project <epp-dev@xxxxxxxxxxx>
Subject: Re: [epp-dev] MAT in Java-based packages

On 24 Aug 2016, at 11:09, Mickael Istria <mistria@xxxxxxxxxx> wrote:

How does MAT handle huge dumps? Does it crash the application with an OOM or gently says it's sorry and not able to process such big files without more memory?

It does try to handle them but at some point OOM occurs and the standard Eclipse dialog pops up suggesting to close/restart the workbench.

-Gunnar