[Dune] Parallel grid partitioning

[Eike Mueller]

Dear Dune-list,

I wonder if someone could help me with these questions about 
partitioning my grids in parallel runs.

The first is about a unit cube realised with YaspGrid. If I let Dune do 
the partitioning and run on 8 cores, it will split all three space 
dimensions into two, so I end up with 8 small cubes, each with half the 
sidelength of the initial cube. However, I want it to keep vertical 
columns together, so when using eight cores the x-direction should be 
split into four bits and the y dimension into two bits and the 
z-direction should be completely unsplit. The constructor of YaspGrid 
seems to take a load balancer as an argument but I could not find any 
more information on this in the documentation. Do I have to write my own 
loadbalancer and what class would I have to derive it from?

Second, I'm setting up a spherical shell with ALUGrid (see [1]). Again I 
want to keep vertical columns together and not split into the radial 
direction, so the sphere should be covered with eight patches of equal size.
So far I only managed to get the serial ALUGridFactory to work, if I run 
in parallel, do I have to do the decomposition myself and call 
inserVertex() and insertElement() on the appropriate processor or can I 
do the balancing afterwards? Or do I have to insert all 
vertices/elements on the master node?

Thank you very much for any ideas,

Eike

[1] http://people.bath.ac.uk/em459/dune.html