Hello Carsten,

I start to realize that the usage of VisitSubTree cored nodes does have 
its own peculiarities. I have a lot of graph traversal calls that I have 
to check if they actually do still work properly.

Anyway I do not see an alternative to its usage. The problem I have is 
the amount of memory used for a basic node or core.

An OpenSG node counts on Windows 64 for 304 bytes (explicit cost). If I 
add the implicit cost for management (aspect,...) I do get 488 bytes.

This is, however, a best case value since multiple nodes do use more 
memory because of vector capacity issues and memory fragmentation. For 
instance, creating 1000000 nodes resulted in measured 596 bytes per node.

Extra memory is also needed for the draw tree and the GL backend 
(Currently I do not know, how much that is compared to the scene graph 
itself).

As a side note the MemoryConsumption's print function does not properly 
state the real cost of the elements.

For instance I got the following line from the function,
Node [49805088] Byte [48638] kByte [900096],

which leads to about 55 byte per node which differs by an order from the 
measured memory footprint. For that I did accumulate the actual sizes of 
the Node state. This gives me the 304 bytes mentioned above.

An additional route, I'm planing to go, is the creation of 'higher 
order' cores. For instance, in my models there is always a Transform 
node followed by an MultiSwich node. My idea is to create a 
'TransformingMultiSwitch' core resulting in a save of one node.

Maybe I'm not counting correctly or another way exists for working with 
huge models primarily containing small geometry per scene graph branch.

Any suggestion is appreciated.

Best,
Johannes





------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from 
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60135031&iu=/4140/ostg.clktrk
_______________________________________________
Opensg-users mailing list
Opensg-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensg-users

Reply via email to