>> Getting them to talk to each other is easy enough; you can create a >> new Parallel::Communicator from MPI_COMM_WORLD and then all our >> friendly helper functions should work with that. Getting the >> communication efficient will be a bear in any case, though; you'd have >> to construct your own send_list connecting completely independent >> partitionings. > > Hm. Let me elaborate some more. The two meshes are completely > independent, except that the solutions of both systems in the next > timestep depend on the current solution of both systems/meshes, not just > their own. So essentially all I need to communicate are vectors > containing all system dofs of a certain variable. (Think of it as two > separate but colliding elastic bodies. One of them is a sheet, the other > is volumetric. For collision handling, both need to know the current > position of the other.) Will that simplify the anticipated send_list effort?
"Communicate all dofs" is very good for program complexity, yes; not so good for computational complexity. --- Roy ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Libmesh-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/libmesh-users
