Dirk Reiners wrote:
>       Hi All,
> 
> Marcus Lindblom wrote:
>> I'm just saying it might be a nice feature to have, since it would allow 
>> you to save/send/clusterify the scene graph even after you have released 
>> the memory.
> 
> True, but it would make the whole system fairly complicated and brittle.

Yup. It's probably not worth focusing on at the moment, but I like to 
think about it. :)

>> The primary reason is to save memory during normal app run (to reduce 
>> process footprint & swapping). But, a save of the scene graph (due to 
>> debugging or whatever) will require some disc access anyway so 
>> temporarily increased memory usage might be tolerated in that context.
> 
> But I guess for debugging you could live with the increased memory footprint 
> too.

Indeed.

>> Me too. I would expect the user to manually restore data before saving.
> 
> The problem that I see is making sure that the frame has actually been 
> rendered. 
> Integrating that into a multi-threaded rendering pipeline gets messy.

Just for the discussion, I'd say that you would have the following logic:

If the GL-object is validated, it has been rendered, so therefore it can 
be released. It can be restored only if it has been released, and thus 
it must've been rendered too.

>> I don't know how big the typically clustered scenes are, but I'd expect 
>> this could be used to increase the amount of renderable geometry per 
>> node, especially if the computer does other things as well. (As we do, 
>> we have a lot of computation going on, so we can't reserve the entire 
>> 2gb of RAM to the graphics subsystem.)
> 
> The way I look at it it would make sense as something that's explicitly 
> triggered by the app, i.e. at some point it tells OpenSG: "BTW, I don't need 
> the 
> images in memory any more, let's get rid of them". At that point OpenSG would 
> just release all the images' memory and be optimistic that they really meant 
> it. 
> ;) After that point you couldn't save the scene any more or add cluster nodes.

That would definitely be helpful at a start.

Would this also apply to geometry that's been uploaded as a VBO or 
Displaylist?

> I could see that being possible, and there should be a way to make 
> clusterable, 
> too. Would that be useful?

Yes. The "restore" stuff was just a stray thought of mine that would 
make it less disruptive with saving and clustering.

Cheers,
/Marcus

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems?  Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Opensg-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/opensg-users

Reply via email to