CGAL users discussion list

[Hermann Schneider <>]

Dear All,

I'm currently investigating to implement an out of core variant based on
CGAL's surface mesh simplification. During testing I encoutered a big
performance problem when edge_collapse is performed many times in a
incrementaly change surface.

After some profiling: it is evident that the several .push_backs of vectors
's inside of Edge_profile seem to be the issue.

On each construction of an Edge_profile in EdgeCollapse::create_profile;
the std::vectors mLink, mBorderEdges, mTriangles are newliy allocated on
each construction of Edge_profile and filled from scratch using push_back !

Are there plans to change/ improve this?

Ideas would be to create a factory for edge_profiles which return protoypes
with reserverd memory for std::vectors mLink, mBorderEdges, mTriangles?

Cheers,
Hermann Schneider



--
View this message in context: 
http://cgal-discuss.949826.n4.nabble.com/Performance-issues-in-Surface-Mesh-Simplification-tp4659098.html
Sent from the cgal-discuss mailing list archive at Nabble.com.