Hi Wolgang

> In addition to the other suggestions, if you already have the points and
> how
> they are connected to cells, you could use the
> Triangulation::create_triangulation function without the detour of a mesh
> generator. See step-14.
>
> But in general: 83M points will make for some 83M cells. The rule of thumb
> is
> some 1000 bytes per cell in a typical FE simulation, so you'd need to have
> a
> machine with 80 GB of memory to do anything useful with a mesh of this
> size,
> and it'd most likely be very slow to solve anything on such a mesh unless
> you
> can parallelize a lot of operations. If I understand correctly, your mesh
> is
> essentially a box. You may be better off using a box with many fewer cells
> (e.g. the one generated by GridGenerator::hyper_rectangle), then refine it
> as
> necessary (maybe not to the same level of detail as you currently have or,
> if
> that is necessary using a parallel program) and determining material
> properties at individual quadrature points not based on the material id you
> read in with the mesh but by looking things up in an array that has your
> 83M
> points.
>
>

This is excellent. I understand now what Markus meant. This is indeed the
best approach as there's no chance of misalignments since I can just look up
values in an array like you indicated.

Many thanks to you and Markus

Ted
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii

Reply via email to