CGAL users discussion list

["Sebastien Loriot (GeometryFactory)" <>]

On 01/24/2014 11:49 AM, Mel M wrote:
> Hi,
>
> I'm developing a program which uses CGAL - specifically the Polyhedron_3
> object and the remeshing and intersection algorithms. It also uses MPI. When
> in parallel, the domain and the polyhedral surfaces are split up spatially
> into cuboids based on the Cartesian co-ordinates. I then run various CGAL
> algorithms on each bit, and stitch them back together afterwards.
>
> My problem is with locality - obviously if the algorithms are not local my
> results will be dependent on how the domain is split up. Especially I'm
> having trouble with remeshing the polyhedral surface on each core and then
> stitching it back together afterwards, when there may have been vertices
> added or removed on one side.
>
> Has anyone done work using CGAL and MPI? How did you get the two to work
> together? (I've read Parallel Multi-Core Geometric Algorithms in CGAL, but
> it's not quite what I'm trying to do).
If I understand correctly, the stitching pb is not related to the
parallel runs but to the cuboid decomposition since the region
near the border of the cubes are not guarantee to match
(and won't in general).

Did I miss your question?

Sebastien.

> --
> View this message in context: 
> http://cgal-discuss.949826.n4.nabble.com/MPI-and-locality-tp4658707.html
> Sent from the cgal-discuss mailing list archive at Nabble.com.