Re: [Libmesh-devel] Recursive partitioning

[John P. <jwp...@gm...>]

On Tue, Jul 22, 2008 at 10:56 AM, Benjamin Kirk <ben...@na...> wrote:
> Ultimately I would like to generalize the partitioner interface to work on
> an input iterator range.
>
> You could then do something similar to this...
>
> // partition into NNodes
> Partition(mesh.active_elements_begin(),
>          mesh.active_elements_end(),
>          n_nodes);
>
> // partition each subdomain into ProcsPerNode
> Partition(mesh.active_local_elements_begin(),
>          mesh.active_local_elements_end(),
>          ppn,
>          pid*ppn);
>
> But I think for testing purposes I will create some manual partitioning
> maps...

Interesting.  I'm seeing that slide by Saeed Iqbal with the picture of
recursive partitioning.

I'm sure it won't be too hard to handle corner cases like, what to do
if a "node" as more cores available than elements which need to be
partitioned.  Not unlikely on smaller grids or when Intel starts
releasing 100 core CPUs :-)

-- 
John