Thanks, Roy.
For the last question, since I need to use one GPU card, that means I need
to use the matrix with SEQAIJ type.
To my understanding, if I use more than 2 CPU cores in MPI parallel, the
matrix type will become MPIAIJ.

Best,
Yujie

On Tue, Feb 21, 2012 at 9:53 PM, Roy Stogner <[email protected]>wrote:

>
> On Tue, 21 Feb 2012, recrusader wrote:
>
>  Now, does libmesh support the multithread?
>>
>
> Yes.
>
>
>  Do you have any examples about how to use it?
>>
>
> Search through the library code itself for usage of the Threads::
> namespace functions.
>
>
>  I am doing tests to compare GPU and CPU.
>> Sometimes, I need to use a huge mesh. However, I usually use 6 CPU
>> cores in one CPU chip. In MPI-based parallel, the memory is not
>> enough to save 6 copies of the mesh.
>>
>
> Threading is a good way to get around that problem without having to
> switch to a distributed ParallelMesh.
>
>
>  In addition, When I use one GPU card, I have to use one CPU core,
>> which keeps the mesh is not partitioned.
>>
>
> Not sure what this means exactly...
> ---
> Roy
>
------------------------------------------------------------------------------
Virtualization & Cloud Management Using Capacity Planning
Cloud computing makes use of virtualization - but cloud computing 
also focuses on allowing computing to be delivered as a service.
http://www.accelacomm.com/jaw/sfnl/114/51521223/
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to