Hi Andreas, On 24 November 2017 at 13:44, Andreas Ekstrand <[email protected]> wrote: > Forgot to mention that the geometry before and after optimization is > identical in the different OSG versions, so the optimizer accomplishes the > same thing, it just takes about 10 times longer.
I am currently investigating the issue, currently waiting on valgrind --tool callgrind to complete to see what clues gives me. It's already been running quite a while so am getting on with other avenues of investigation while I wait. The changes to Optimizer.cpp that you sent to osg-submissions suggests that it's the removeChild() that is the performance bottleneck. The changes you've proposed don't handle the case where a group can have non Drawable children so can't be merged as is. It's still a possibility of merging it with mods to address this. I'd like to get down to route cause of the bottleneck though, my best guess right now is it's an issue of parts of the implementation triggering a O(N^2) behaviour, i.e. large number of children being merged. It might be this can be resolved with the approach you took, or it might be best adding better removeChildren() support that avoids the O(N^2) behaviour, such as passing a list of nodes to remove. Robert. _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
