Hi, I'm developing a new preconditioner in PETSc which I wish to test with all the great facilities dealii offers.
The problem is that such a preconditioner requires the PETSc parallel matrix type to be "MATIS", which is not of (neither derived from) the PETSc MPIAIJ type you use in the wrapper. Basically, it stores as many sequential matrices as the number of MPI procs, and you need to assemble local matrices on local elements only. The key is that the elements' distribution between MPI procs should be contiguous (connected set). Can you provide this new wrapper class in dealii? (I could help you with PETSc details) Or, alternatively, can you give me some quick hints to implement such a new class in my deal distribution? Thanks, -- Stefano
_______________________________________________ dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii
