The ABINIT Users Mailing List ( CLOSED )

[Mauro Sgroi <maurosgroi@yahoo.it>]

Hi,
thanks a lo for the clarification. I will try to study
how MPI lib can handle this problem.
Thanks a lot and best regards,
Mauro.

--- Anglade Pierre-Matthieu <anglade@gmail.com> ha
scritto:

> Hi,
> 
> Not so much a fool question. Yet I think that
> nothing of this kind is
> already implemented into Abinit.
> I think you may feel free to do it however.
> On the other side doing parallel calculations over
> heterogeneous clusters is
> more of a computer scientis problems. It would,
> probably, be best solved on
> the MPI lib side. And it looks like some groups
> provides
> implementations/tricks and advices on how to run MPI
> calculations on
> heterogeneous systems. You may be interested in
> checking if any MPI flavor
> is able to solve your problem without changing
> anything in Abinit.
> 
> Best regards
> 
> PMA
> 
> On 7/9/07, maurosgroi@yahoo.it <maurosgroi@yahoo.it>
> wrote:
> >
> > Dear all,
> > maybe a fool question: is it possible to manually
> assign the number of
> > k-points to each processor during a parallel
> computation?
> > This could improve the performance of a cluster
> composed by different CPU
> > with different speeds.
> > Thanks a lot and best regards,
> > Mauro.
> >
> 
> 
> 
> -- 
> Pierre-Matthieu Anglade
> 





        
                
___________________________________ 
L'email della prossima generazione? Puoi averla con la nuova Yahoo! Mail: 
http://it.docs.yahoo.com/nowyoucan.html