[Pw_forum] running PWSCF on IB-based cluster

高星 xinggao at impcas.ac.cn
Mon Sep 1 02:47:18 CEST 2014


Thanks Axel for your suggestions!

best,

Xing


> -----原始邮件-----
> 发件人: "Axel Kohlmeyer" <akohlmey at gmail.com>
> 发送时间: 2014年8月30日 星期六
> 收件人: "PWSCF Forum" <pw_forum at pwscf.org>
> 抄送: 
> 主题: Re: [Pw_forum] running PWSCF on IB-based cluster
> 
> On Sat, Aug 30, 2014 at 4:44 AM, 高星 <xinggao at impcas.ac.cn> wrote:
> > Dear Paolo,
> >
> > Thanks a lot for your quick response! We are using intel compiler and
> > openmpi. Actually, someone else configured our cluster, and I had to use
> > mpirun with option "bind-to-core" to run a job over more than one nodes. If
> 
> i seriously doubt that. rather you should discuss this with the person
> that configured the cluster and get an answer that is consistent with
> the existing configuration.
> 
> > I do not use "--bind-to-core" option, the job will be submitted to only one
> > node with np processors. There was no problem to run other program with
> > "--bind-to-core" option except QE, for which I got "processor affinity
> > failure". That's why I posted my question on pw_forum. Thanks again for your
> 
> this explanation doesn't make sense. the bind-to-core option does
> nothing like that. and it has definitely nothing to do with QE
> it sounds more like you do not provide a suitable hostfile. please
> spend some time going over the OpenMPI documentation, compile a simple
> MPI test program and experiment with that how to run properly on your
> cluster. if you have further problems, you should post to the OpenMPI
> mailing list, or bug your provider again.
> 
> axel.
> 
> > reply!
> >
> > Best,
> >
> > Xing
> >
> >
> >> -----原始邮件-----
> >> 发件人: "Paolo Giannozzi" <paolo.giannozzi at uniud.it>
> >> 发送时间: 2014年8月29日 星期五
> >> 收件人: "PWSCF Forum" <pw_forum at pwscf.org>
> >> 抄送:
> >> 主题: Re: [Pw_forum] running PWSCF on IB-based cluster
> >
> >>
> >> QE knows nothing about which processor is on which node.
> >> It must be related to the MPI libraries ("openmpi" and
> >> "intel mpi" are different, by the way, and one should
> >> be careful not to mix up the two) and to how MPI is
> >> started on your machine
> >>
> >> P.
> >>
> >> On Fri, 2014-08-29 at 17:53 +0800, 高星 wrote:
> >> >
> >> > Dear QE community,
> >> >
> >> > We have an infini-band based cluster with 8 computing nodes and each
> >> > nodes contains 16 cores. I compiled QE package on this cluster
> >> > successfully with compiler mpif90 (openmpi intel compiler) and there
> >> > was no problem to run  pw.x on one node with 16 processors. However,
> >> > when I tried to use two nodes (that is, -np 32), I met "processor
> >> > affinity failure". It says, "an attempt to set processor affinity has
> >> > failed - please check to ensure that your system supports such
> >> > functionality. If so, then this is probably something that should be
> >> > reported to the OMPI developers". Actually, there is no such problem
> >> > when I run other program. I am wondering if someone else met the
> >> > similar question before? And please give me any hint which may help me
> >> > on this problem! Thanks in advance!
> >> >
> >> > Best,
> >> >
> >> > Xing
> >> >
> >> >
> >> >
> >> >
> >> >
> >> >
> >> > _______________________________________________
> >> > Pw_forum mailing list
> >> > Pw_forum at pwscf.org
> >> > http://pwscf.org/mailman/listinfo/pw_forum
> >>
> >> --
> >>  Paolo Giannozzi, Dept. 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://pwscf.org/mailman/listinfo/pw_forum
> >
> >
> >
> >
> >
> >
> >
> > _______________________________________________
> > Pw_forum mailing list
> > Pw_forum at pwscf.org
> > http://pwscf.org/mailman/listinfo/pw_forum
> 
> 
> 
> -- 
> Dr. Axel Kohlmeyer  akohlmey at gmail.com  http://goo.gl/1wk0
> College of Science & Technology, Temple University, Philadelphia PA, USA
> International Centre for Theoretical Physics, Trieste. Italy.
> 
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum






-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20140901/58c45e29/attachment.html>


More information about the users mailing list