[petsc-users] Diagnosing Poisson Solver Behavior

[Jed Brown]

"K. N. Ramachandran" <knram06 at gmail.com> writes:
> I am trying to simulate the movement of charged particles in a domain, and
> these particles become collimated along a particular direction. So one way
> to partition might be 1D slabs perpendicular to the collimated direction,
> so each rank can calculate the field and move the particles on each
> timestep. Of course communication across the slabs would be inevitable. But
> this way, the partitioning is based on the physics of the problem and the
> solver module just efficiently solves the system without having to know the
> boundary conditions, which makes sense to me overall.

I don't understand what you mean by "without having to know the boundary
conditions".  The boundary conditions are essential to solving
equations.  Of course if you're assembling a matrix, they need to be
part of that matrix.  (Usually the matrix will be singular "without"
boundary conditions.)

As for using a 1D partition, that may be very inefficient for the
solver, both due to limited scalability of the partition and more
interestingly if the directions of strong coupling do not lie entirely
within your planes and completely swamp the out-of-plane coupling
strength, then your solver can be expected to converge more slowly.

Note that DMDA can be instructed to produce a 1D partition if you like.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 818 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20151015/e3a41da6/attachment-0001.pgp>