[petsc-users] Hypre + openMP

[Dave May]

Yes, coarsen the dmda until you hit your processor limitation, then use
preonly + hypre/GAMG as your smoother on the coarsest geometric multi grid
level.

Cheers,
  Dave

On Wednesday, 13 May 2015, Michele Rosso <mrosso at uci.edu> wrote:

> Hi Jed,
>
> I see what you mean. I am using geometric multigrid and therefore I have a
> limit on the number of processors I can use. I was looking for an
> alternative and found hypre, but I did not realize that since it's
> algebraic multigrid there is no such limitation.
> Anyway, let's say I want to keep the geometric multigrid: is there a way
> to work around the problem?
>
> Thanks,
> Michele
>  On May 12, 2015 10:20 PM, "Jed Brown" <jed at jedbrown.org
> <javascript:_e(%7B%7D,'cvml','jed at jedbrown.org');>> wrote:
>
>> Michele Rosso <mrosso at uci.edu
>> <javascript:_e(%7B%7D,'cvml','mrosso at uci.edu');>> writes:
>>
>> > Thanks Matt,
>> > Could you give me an example of not-native partitioning? I have a cubic
>> > domain and a 3D domain decomposition. I use dmda3d to create the
>> > partitioning.
>>
>> Are you talking about geometric multigrid (AMG coursening is typically
>> irregular)?  Anyway, this is just math: what does it mean to have less
>> than one degree of freedom on a level?  Geometric MG on DMDA does have
>> requirements, but if you're using Hypre, I don't see why you'd care
>> about that.
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150513/35a2ae9a/attachment-0001.html>