yade-dev team mailing list archive

[Bruno Chareyre <bruno.chareyre@xxxxxxxxxxx>]

CGAL needs 1.4sec for 100k spheres (regular triangulation, probably less 
for simple Delaunay).
In flow coupling, we need triangulations updated at run-time, which 
makes CPU time a major issue. I forget using Voro++ for this reason.
For post-processing applications like this, we don't need tesselation, 
but only triangulation : https://yade-dem.org/wiki/Triangulation.
I agree that connectivity data can, theoreticaly, replace iterators to 
cycle through tetrahedra, but looking at voro++ data, it would be really 
complicated, and probably very slow. I didn't find a simple way to get 
the 4 spheres incident to one voronoi vertice yet (do you know one?).
Very interesting in voro++ is the number of variants of boundaries 
geometry. I think 'll look at that in details one day, and see if/how 
similar constructs could be implemented with CGAL.


Bruno
> No, I didn't make direct compare, but on my laptop voro++ need 15 sec 
> to tesselate 112k particles+cylindrical wall. And it can to tesselate 
> only one sphere from all bulk if need. I use it for postprocessing now 
> in order to calc porocity. To learn it you need a pair hours, not 
> more. Yes, it does not have a iterators, but it can to track 
> neighbours of course, so iterator can be make, I think... (I had 
> boring with CGAL some years ago: it is very big and hard to learn, but 
> I not sure what it is slow although).
--

_______________
Bruno Chareyre
Associate Professor
ENSE³ - Grenoble INP
Lab. 3SR
BP 53 - 38041, Grenoble cedex 9 - France
Tél : +33 4 56 52 86 21
Fax : +33 4 76 82 70 43
________________