AMBER Archive (2007)

Subject: Re: AMBER: REMD error

From: Jason K (jmk9jmk_at_gmail.com)
Date: Tue May 22 2007 - 14:02:22 CDT

I have used the submission script for running regular MD in parallel, and
that ran with no problems.

Non-REMD multisander fails though, with the same error message (number of
groups greater than number of processors...)

What could cause that error?

Best wishes,
Jason

On 5/21/07, Carlos Simmerling <carlos.simmerling_at_gmail.com> wrote:
>
> have you run regular MD with more than 1 cpu using this code?
> also, have you run non-REMD multisander using this code?
>
>
> On 5/21/07, Jason K < jmk9jmk_at_gmail.com> wrote:
> >
> >
> > Dear AMBER,
> >
> > I have been trying to run a replica exchange simulation on our
> > departmental cluster (on which parallel jobs can be run on up to 12
> > four-processor nodes). The command that I use is - essentially - the
> > following
> >
> > mpirun -np 8 -machinefile {machine file} $AMBERHOME/exe/sander_mpi -ng 2
> > -groupfile groupfile
> >
> > the groupfile being the one exemplified in the test cases (I have
> > checked that it does not contain any errors)
> >
> > Nevertheless, when I submit the job I get error messages similar to
> > these:
> >
> >
> > Error: specified more groups ( 2 ) than the number of
> > processors (
> > 1 ) !
> > p0_2936: p4_error: : 1
> >
> > [0] MPI Abort by user Aborting program !
> > [0] Aborting program!
> >
> >
> > Can anyone suggest what may be the cause of this behaviour? Is this a
> > message generated by sander, or is it a failure with either MPI or the
> > queueing system?
> >
> > I can attach the actual files containing the standard output / error as
> > well as the input files if that can be any help.
> >
> > All help will be greatly appreciated.
> >
> > Jason
> >
> >
> >
>
>
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu