From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Wed Jun 24 2015 - 11:16:34 CDT
On Wed, Jun 24, 2015 at 12:03 PM, Alexander Tzanov
<Alexander.Tzanov_at_csi.cuny.edu> wrote:
> Just to clarify a bit more:
>
> NAMD cannot use more than one core on a node. If I run with one core per
> node,
yes, NAMD *can*. the issue is that you are not using it in the right
way (but since you don't provide sufficient information about that, it
is difficult to give specific advice on that).
> the NAMD runs. If I ask for more than one core i.e. 4 cores, I got the
> error:
>
> --------------------------------------------------------------------------------------------------------------------------
> There are not enough slots available in the system to satisfy the 4 slots
> that were requested by the application:
> namd2
>
> Either request fewer slots for your application, or make more slots
> available
> for use.
> --------------------------------------------------------------------------
this looks like an error message coming from the MPI library and not NAMD.
this is usually very system/cluster/machine specific and also depends
on the MPI library.
you may need to set up a machine file or request the suitable number
of processors per node from the batch system.
you should compile a small MPI test problem and work things out. also,
this is best addressed by working with your local user support.
please note that NAMD has multiple ways to launch instances: you can
have independent individual processes and multiple threads per each
process. when you work with GPUs, they situation becomes incrementally
more complicated. also, you may need to pay attention to processor and
memory affinity.
again, this is all very specific to your local installation and thus
cannot be easily solved from remote.
axel.
>
> Thank you for your help.
>
> Alex
>
> On Jun 24, 2015, at 10:56 AM, Alexander Tzanov
> <Alexander.Tzanov_at_csi.cuny.edu> wrote:
>
> Dear all,
>
> I am trying to run NAMD 2.10 with CUDA support on single virtual node. My
> CUDA is 6.5.14. I am running
> on IB cluster which runs PBS pro ver. 12.1.0.131281. I want to run on a
> single virtual node with
> 4 CPU cores and 2 GPU (the CPU itself has 16 cores). I compiled NAMD with
> cuda support
> (mpi, smp CHARMM++ underneath). However when the NAMD cannot address more
> than one core.
>
> As a result I get error “ Not enough slots”. Does anyone see the problem and
> if so to share his/her experience.
>
> Thank you
>
> Alex
>
>
>
> ________________________________
> Celebrate the World of Peptide Chemistry and Biology. Register today for the
> Symposium in Honor of the Scientific Contributions of Dr. Fred Naider>
-- 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.
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:21:56 CST