Hi Matthias,
Hi Sebastian,
the sample code is called asynchronously from several compute threads.
When I make sure that only one thread at a time calls accept() everything is
rock solid (millions of calls). But it's running slowly and doesn't scale well
with the number of cores.
Without the lock the calls to accept() are asynchronous too and I get wrong
results every few hundred calls.
If the scenegraph isn't static I'd expect data races or even crashes if
elements are added/removed.
You can traverse the graph after update/traversal safely.
Take a look at osg::Operation. You can use it to add threaded operations
to the update phase and synchronize against the frame only instead of
interlocking your intersection traversal.
I'm using it to populate the scene with instanced vegetation by querying
the terrain height/color etc. and it is working fine.
I don't think the problem is asynchronously call accept on the visitor
but calling it when draw/update of the scenegraph are running. You could
grep the intersection code for non-reentrant code (use of static objects
for instance) to be sure.
Cheers
Sebastian
The compute threads themselves run independetly from the main thread with the
viewer. In fact this even happens in batch mode too without any visual
representation at all.
Thank you!
Cheers,
Matthias
------------------
Read this topic online here:
http://forum.openscenegraph.org/viewtopic.php?p=64596#64596
_______________________________________________
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
_______________________________________________
osg-users mailing list
osg-users@lists.openscenegraph.org
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
