[Gmsh] upgrading GmshClient to onelab

Nigel Nunn nnnunnn at gmail.com
Fri Apr 25 21:34:14 CEST 2014


On 09 Jul 2013, at 10:34, Christophe Geuzaine <cgeuzaine at ulg.ac.be>
wrote:

> Hi Nigel - Have a look at utils/solvers/c++ and utils/solvers/python in
SVN.

Dear Christophe,
Thanks for the hint!  I used /utils/solvers/c++/pend.cpp as an example for
interacting with ONELAB, added some code to handle MPI launch commands, and
everything is working even better than before  :-)

For our MPI+GPU solvers, it's important to be able to [Pause] and inspect a
simulation, so I added a [Pause] button.  This uses the process id returned
from MPI proc[0] (during /Action = "initialize") to pause that particular
process, which just happens to pause all the MPI procs at the next MPI
barrier.

Since my code to make this work (on Win x64) is a bit of a hack, I wonder
if any of your getDP or ONELAB collaborators have set up a more
ONELAB-standard [Pause] mechanism?

thanks for all the splendid work!

Nigel


On 07 May 2013, at 00:55, Nigel Nunn <nnnunnn at gmail.com> wrote:

> Hi Gmsh,
>
> Sometime before the onelab update, I used GmshClient  (from old
> GmshSocket.h) to drive our family of mpi gpu accelerated solvers
> (discontinuous Galerkin).  I hooked up a pause button to manage the
> mpi fleet, and a few parameters for adjusting runtime results passed
> back to Gmsh.
>
> After looking at the classes in [gmsh/contrib/onelab] and
> [gmsh/projects/onelab, and the way getdp uses onelab, I am not sure
> how to begin upgrading from GmshSocket to onelab.  Can someone please
> suggest which of the classes derived from onelab client would do?
> First stage is to reproduce what we had with GmshClient :  get Gmsh to
> query our solver (mpi proc [0]) for parameters, then launch, pause and
> stop a simulation using 4 gpus on a single node (dual Xeon).
>
> runGmshClient() from [gmsh/Common/onelabUtils.cpp] sounds promising,
> but which client?
>

Hi Nigel - Have a look at utils/solvers/c++ and utils/solvers/python in SVN.


> thanks for any help!
> Nigel
>
> _______________________________________________
> gmsh mailing list
> gmsh at geuz.org
> http://www.geuz.org/mailman/listinfo/gmsh

-- 
Prof. Christophe Geuzaine
University of Liege, Electrical Engineering and Computer Science
http://www.montefiore.ulg.ac.be/~geuzaine
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20140426/0b46bdc5/attachment.html>