[aspect-devel] PRACE

Wolfgang Bangerth bangerth at math.tamu.edu
Fri Jan 20 08:58:45 PST 2012

On 01/20/2012 07:39 AM, Thomas Geenen wrote:
> at some point we would like to run models with several hundred million
> degrees of freedom or more.
> for that our local cluster will not be large enough.
> i was offered support from PRACE from our national supercomputer center
> to port dealii to one (or several) of the available petaflop machines in
> europe.
> would that be something useful?

Most definitely!

> i am thinking mainly about issues with
> compilers (what is the status of the latest xlc compiler?) and system
> specific issues like IO etc
> i think it will also help when we apply for cycles on those machines
> that is has been officially ported and profiled.

Yes, having had someone help with porting is definitely a plus on any 
application for cycles.

I don't know about the kind of systems PRACE provides. As for IBM xlC, 
that's sort of a sore point since the IBM compiler has never been able 
to compile deal.II completely. I must have submitted about two dozen bug 
reports to them over the past 6 or 7 years (I know the people who write 
xlC) but they are very slow addressing them and they never seem to get 
to the point where everything actually works.

So the safe side would be to stay with x86-systems and using either GCC 
or the Intel ICC compiler. I would expect that anything on which GCC 
runs would also be ok (which may include IBM Power systems) but I have 
less experience with that.


Wolfgang Bangerth               email:            bangerth at math.tamu.edu
                                 www: http://www.math.tamu.edu/~bangerth/

More information about the Aspect-devel mailing list