Yes - I tried whatever. According to openMPI, mpirun should get PBS
environment and all assigned CPUs. I tried to submit by machine list
option, although it is not everything well supported. And it was no
success, also. I've tried to set default MPI 's communication protocol
to ssh or rsh in PBS script (export OMPI_MCA_pls_rsh_agent=ssh) - no luck.
Should OMPI libraries to be compiled with PSB support in order to use
PBS ? However, compiling OMPI I kept with the installation protocol,
which came with OMPI sources...
Arturas
David LeBard wrote:
> Sorry, I see you commented this out. Please forget my last email.
>
> On Mon, Jun 16, 2008 at 11:07 AM, David LeBard <david.lebard.asu.edu
> <mailto:david.lebard.asu.edu>> wrote:
>
> Hi Arturas,
>
> Have you tried to explicitly tell mpirun how many processors to
> use like this?
>
> mpirun -np $N $AMBERHOME/exe/sander.MPI ...
>
> Just a thought,
>
> David
>
>
> On Mon, Jun 16, 2008 at 10:56 AM, Arturas Ziemys
> <arturas.ziemys.uth.tmc.edu <mailto:arturas.ziemys.uth.tmc.edu>>
> wrote:
>
>
> Here is my PBS file:
> **************************************************************
> #! /bin/sh
> #PBS -N P_min_prl5
> #PBS -l cput=20:00:01
> #PBS -l walltime=22:00:00
> #PBS -l nodes=3:ppn=2
> #PBS -j oe
> #PBS -V
>
> echo
> echo -------------
> echo PBS_NODEFILE:
> cat $PBS_NODEFILE
> echo -------------
> echo
>
> N=`wc -l < $PBS_NODEFILE`
> echo this job has allocated $N nodes
>
> cd /home/aziemys/Works/Pep/AMBER/P/min
>
>
> #mpirun -np 6 $AMBERHOME/exe/sander.MPI -O -i min.in
> <http://min.in> -o P_min_prl5.out -c BRCT.inpcrd -p
> BRCT.prmtop -r BRST_min_prl5.rst
>
> mpirun $AMBERHOME/exe/sander.MPI -O -i min.in <http://min.in>
> -o P_min_prl5.out -c BRCT.inpcrd -p BRCT.prmtop -r
> BRST_min_prl5.rst
>
> **************************************************************
>
>
> Carlos Simmerling wrote:
>
> can you show us your qsub script? what flags are you giving
> to mpirun?
>
> On Mon, Jun 16, 2008 at 1:00 PM, Arturas Ziemys
> <arturas.ziemys.uth.tmc.edu
> <mailto:arturas.ziemys.uth.tmc.edu>> wrote:
>
>
> Please, help me to setup AMBER 9 parallel (AMBER and
> openMPI compiled with
> g95,gcc4.3.x,redhat,xeon cpu). AMBER parallel test
> runs fine. Also, I run
> AMBER job fine from a command prompt. But if I submit
> job with 'qsub', AMBER
> runs just with a single CPU using sandes.MPI. What is
> wrong ?
>
> Arturas
>
> --
>
> Arturas Ziemys, PhD
> School of Health Information Sciences
> University of Texas Health Science Center at Houston
> 7000 Fannin, Suit 880
> Houston, TX 77030
> Phone: (713) 500-3975
> Fax: (713) 500-3929
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> <mailto:amber.scripps.edu>
> To unsubscribe, send "unsubscribe amber" (in the
> *body* of the email)
> to majordomo.scripps.edu <mailto:majordomo.scripps.edu>
>
>
>
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> <mailto:amber.scripps.edu>
> To unsubscribe, send "unsubscribe amber" (in the *body* of
> the email)
> to majordomo.scripps.edu <mailto:majordomo.scripps.edu>
>
>
>
>
>
> --
>
> Arturas Ziemys, PhD
> School of Health Information Sciences
> University of Texas Health Science Center at Houston
> 7000 Fannin, Suit 880
> Houston, TX 77030
> Phone: (713) 500-3975
> Fax: (713) 500-3929
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu <mailto:amber.scripps.edu>
> To unsubscribe, send "unsubscribe amber" (in the *body* of the
> email)
> to majordomo.scripps.edu <mailto:majordomo.scripps.edu>
>
>
>
--
Arturas Ziemys, PhD
School of Health Information Sciences
University of Texas Health Science Center at Houston
7000 Fannin, Suit 880
Houston, TX 77030
Phone: (713) 500-3975
Fax: (713) 500-3929
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" (in the *body* of the email)
to majordomo.scripps.edu
Received on Wed Jun 18 2008 - 06:07:32 PDT
