From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Sat Feb 06 2010 - 21:16:14 CST
On Sat, Feb 6, 2010 at 8:08 PM, snoze pa <snoze.pa_at_gmail.com> wrote:
> Try to use charmrun and define the number of nodes in a file nodeno.txt as
>
> ---------
> node1
> node2
> node3
> node4
>
> don't use dotted lines in your file, its for example. If each node
> have 2 cpu, then you can use +p 8. This means you are using 8 cpu's
please note that this advice is most useful for the -net versions downloaded
from the namd homepage. for self-compiled MPI versions, it is almost always
better to use mpirun with the local setup. particularly, node oversubscription
in the form as it is described here, will negate any effort from the sysadmin
to have the cpu affinity set up through batch/mpi. which in turn can have a
_significant_ impact on most current CPUs.
if you don't care about scalability, run only on a few nodes, or maxmimum
efficiency, then using the precompiled binaries is recommended (in fact,
quite often they may still be faster than locally optimized versions, when
those are "overoptimized").
axel.
>
> then use
>
> charmrun namd2 ++nodelist nodeno.txt +p 8 filename.namd > filename.log &
>
>
>
> On Sat, Feb 6, 2010 at 9:56 AM, #SHEPHERDSON IGNATIUS MARK#
> <SHEP0001_at_ntu.edu.sg> wrote:
>>
>>
>> Hi,
>>
>> I am trying to run parallel simulations on cluster computers. However, when
>> i specify 2 processors to use, I get 2 different processors running serially
>> twice on my log file, giving me 2 results instead of 1. Is there something
>> wrong with my script file?
>>
>>
>> #!/bin/sh
>>
>> #$ -V
>>
>> #$ -cwd
>>
>> #$ -N NAMD
>>
>> #$ -pe mpich-eth 2
>>
>> #/usr/voltaire/mpi/bin/mpirun_ssh -np $NSLOTS -hostfile $TMPDIR/machines
>> ./integral
>>
>> export
>>
>> ls -l $TMPDIR/machines
>>
>> cat $TMPDIR/machines
>>
>> MPIRUN=/usr/local/openmpi/intel/bin/mpirun
>>
>> ls -l $MPIRUN
>>
>> $MPIRUN -np 2 -machinefile $TMPDIR/machines
>> /home/students/FYP_Lim2/Ignatius/NAMD_2.7b1_Source/Linux-x86_64 g++/namd2
>> /home/students/FYP_Lim2/Ignatius/NAMD_2.7b1_Source/test1/z.conf>z.log
>>
>>
>>
>> Thank you.
>>
>> Ignatius
>>
>>
>
>
-- Dr. Axel Kohlmeyer akohlmey_at_gmail.com Institute for Computational Molecular Science College of Science and Technology Temple University, Philadelphia PA, USA.
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:55:24 CST