[petsc-users] Memory and Speed Issue of using MG as preconditioner

[Jed Brown]

"Alan Z. Wei" <zhenglun.wei at gmail.com> writes:
>        As you mentioned that the redundant LU for the coarse grid solver 
> primarily cause the large memory request for the 2-level case. How could 
> I change the coarse grid solver to reduce the memory requirement or 
> speed up the solver.

You should use more levels or coarsen faster.  You cannot have a
scalable method with a huge coarse level.  You could try switch to
algebraic multigrid on the coarse levels if geometric is not converging
well.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 835 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20131106/306530c5/attachment.pgp>