[CIG-LONG] Domain decomposition in SNAC
Eunseo Choi
echoi at ldeo.columbia.edu
Thu Jan 22 16:00:32 PST 2009
-------- Original Message --------
Hi, Mark
As you must have realized, this decomposition is way too far from optimum.
It appears to be happening because the decomposition is allowed only in
one dimension.
For 2- or 3-D domain decomposition, you need to change the value of
"decompDims" to 2 or 3.
This parameter is a part of the "mesh" structure and must be already in
the input file you're using.
Cheers,
Eunseo
Mark Fleharty wrote:
> Eunseo,
>
> I spoke too soon, I think I got it!
> 33 11 81 32 1 1
>
> Thanks!
>
> Mark
>
> On Thu, Jan 22, 2009 at 1:48 PM, Mark Fleharty <mark.fleharty at gmail.com> wrote:
>
>> Eunseu,
>>
>> I ran another simulation this time with 32 cores.
>> The grid size is 33, 11, 81
>> sim has: 1 10 80
>> snac2vtk 33 11 81 32 1 1
>> and
>> snac2vtk 33 11 81 1 1 32
>> but neither of these seem to work.
>>
>> Any ideas?
>>
>> Thanks,
>>
>> Mark
>>
More information about the CIG-LONG
mailing list