CGAL users discussion list

[Frankie Li <>]

Hi Sebastien,

I'm terribly sorry for wasting your time.  I just realized that I had an 
extra line in the middle of the loop that copies the triangulation, left 
over from a test I was running and hence the slow down. ( Sleep 
deprivation => doing stupid things...)  As expected, getting adjacent 
vertices takes sub-second time.

Thanks,

Frankie

Sebastien Loriot (GeometryFactory) wrote:
> Frankie Li wrote:
> > Hi all,
> >
> > I'm trying to traverse all neighbors of each vertex in a regular 
> > triangulation, and naively, I've done something like this:
> >
> > // pseudo-code
> >
> > for( Finite_vertex_iterator it = tr.finite_vertices_begin();
> >       it != tr.finite_vertices_end(); ++ it )
> > {
> >
> >   vector<Vertex_handle> Ngb_vertices;
> >   tr.finite_adjacent_vertices( it, std::back_inserter( Ngb_vertices ));
> > }
> >
> > However, I found it to be surprisingly slow.  I have approximately 
> > 15000 vertices, and I'm using 
> > Exact_predicates_inexact_constructions_kernel. (It takes about 3000 
> > seconds to go through the loop above for 15000 vertices.  I'm running 
> > on a Core 2 Quad machine.)
> >
> >
> > Any advice is welcomed!  Thanks!
> >
> > Frankie
> Hello,
>
> what you report is quite strange, because this is a combinatoric 
> operation usually fast. I just make a quick test with 4500 weighted 
> points and without any optimisation the timing is less 0.03s.
>
> Can you test the program joint (it expects a file with line x y z radius 
> as first argument) and tell us about the running time.
>
>
> S.