[Pw_forum] parallel computation too slow

Elie M elie.moujaes at hotmail.co.uk
Mon Mar 19 16:05:16 CET 2012


Hello,
It is not a computer. It is a cluster and I am doing parallel computing on 3 nodes each with 8 processors which is a total of 24 processors; this is why I am surprised that the calculation is slow . I tried performing the calculations on different nodes but in vain. 
N.B: I did calculations on one server with only 8 processors for the Gamma point and the calculations finished in three weeks (though with some negative frequencies, this is why I decided to minimize the tr2_ph) But now that server is down. I thought on more nodes things would be faster; apparently not

Regards
Elie 

> Date: Mon, 19 Mar 2012 10:00:37 -0400
> Subject: Re: [Pw_forum] parallel computation too slow
> From: akohlmey at gmail.com
> To: elie.moujaes at hotmail.co.uk
> CC: pw_forum at pwscf.org
> 
> On Mon, Mar 19, 2012 at 9:45 AM, Elie M <elie.moujaes at hotmail.co.uk> wrote:
> > Professor Giannozzi,
> >
> > Thanks for the reply. The CPU time is 1 hr 26 min (2hrs 2 min normal time)
> > for the scf calculation on the nodes (3 of them). If I have 42 atoms then
> > this is equivalent to almost 8 days CPU time or 10 days normal time. Four
> > days have passed and I am still in representation 6 out of 126 which i guess
> > is not the normal situation...Can anything be done  (or checked) in this
> > case?
> 
> get a real computer! if you already lose a quarter of the time due to
> swapping for the scf calculation, you'll be seriously out of memory
> for anything more advanced. your laptop is obviously not suitable
> to carry out this kind of calculation. sorry.
> 
> axel.
> 
> >
> > Regards
> >
> >
> > Elie
> >
> >> CC: pw_forum at pwscf.org
> >> From: giannozz at democritos.it
> >> Subject: Re: [Pw_forum] parallel computation too slow
> >> Date: Fri, 16 Mar 2012 21:48:15 +0100
> >> To: elie.moujaes at hotmail.co.uk
> >
> >>
> >>
> >> On Mar 16, 2012, at 19:14 , Elie M wrote:
> >>
> >> > I am running phonon calculations for a supercell of 60 atoms at the
> >> > gamma point [...]
> >> > What could the possible causes of this slowness be? Is it something
> >> > to do with the
> >> > nodes or with phonon calculations in general? Is there a way to check?
> >>
> >> >
> >>
> >> a phonon calculation for N atoms requires a CPU time of the order of
> >> 3N the time of
> >> the correspoding scf calculation. Does yours? if it does, there is
> >> nothing that can be
> >> done.
> >>
> >> P.
> >> ---
> >> Paolo Giannozzi, Dept of Chemistry&Physics&Environment,
> >> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> >> Phone +39-0432-558216, fax +39-0432-558222
> >>
> >>
> >>
> >>
> >
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://www.democritos.it/mailman/listinfo/pw_forum
> >
> 
> 
> 
> -- 
> Dr. Axel Kohlmeyer
> akohlmey at gmail.com  http://goo.gl/1wk0
> 
> College of Science and Technology
> Temple University, Philadelphia PA, USA.
 		 	   		  
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20120319/f965267d/attachment.html>


More information about the users mailing list