K,
Just run the code with mpiexec -np 8 (or whatever) as Roy suggested. The
"greedy algorithm" is sequential, but the finite element solves are
parallel and so is the "sampling" at each greedy iteration. (The
training set is divided up among the processors.)
David
On 08/16/2012 10:04 PM, Ky
Dear Roy,
Thank you for explaining the necessary steps. I'm not quite sure reduced
basis training can work in parallel since it is more like a sequential
process. I'll ask to David to get a better understanding.
Regards,
K. Lee.
On Fri, Aug 17, 2012 at 9:07 AM, Roy Stogner wrote:
>
> On Fri,
On Fri, 17 Aug 2012, Kyunghoon Lee wrote:
> I noticed that my workstation uses only one core for reduced basis training
> despite having eight cores, so I guess I may use more CPU power by
> utilizing parallel. I checked two examples, introduction_ex4.C and
> transient_ex1.C, to see if how it wo
Hi all,
I noticed that my workstation uses only one core for reduced basis training
despite having eight cores, so I guess I may use more CPU power by
utilizing parallel. I checked two examples, introduction_ex4.C and
transient_ex1.C, to see if how it works, but I couldn't tell which part of
code
