[gmx-users] mpirun error

[Justin Lemkul]

On 11/19/12 12:09 PM, Parisa Rahmani wrote:
> Dear gmx users
>
> I have a problem with running parallel jobs on my Debian system(Openmpi
> installed on it),
> **Linux debian 3.2.0-1-amd64 #1 SMP , UTC 2012 x86_64 GNU/Linux**
> I am using gmx 3.3.3, because of the *lambda dynamics* method which is
> implemented in it.
>
> AS I know ,in gmx 3.x, the number of processors supplied for the subsequent
> mdrun needed to match the input file. but when i use **grompp -np 6 &
> mpirun -np 6 mdrun** the following error appears :
>
> ERROR : run input file md.tpr was made for 6 nodes,
> while mdrun expected it to be for 1 nodes.
>
> through search of mailing list i found similar problems, but non of
> the solutions worked for my case.
>
> wihtout -np option in grompp the error disappears, and then with each
> of these commands
>
> **
>
> 1)mpirun -np 6 mdrun -deffnm md
>
> 2)mpirun -np 6 mdrun -deffnm md -N 6
>
> 3)mpirun -np 6 mdrun -np 6 -deffnm md
>
> 4)mdrun -np 6 -s md -N 6
>
>
> **it uses 6 processors(each one at nearly 100%), but the simulation
> time is the same as for 1 processor.
>
> I have no problem with parallel jobs on our cluster(gmx 3.3.3 &
> openmpi), with following command:
>
> **
> grompp -np 6 -f ...
> mpiexec mdrun ...(number of processors is specified in the bash file)
> **
> ,
> but i can't run it on my 6 core system.
>
> Also, I have no problem with newer version of gmx (4.5.x), but i
> should use this version, and Hope someone can help me.
>

Have you properly compiled an MPI-enabled mdrun?  The default executable name 
should be mdrun_mpi.  It should be linked against libmpi, so running ldd on the 
mdrun executable should tell you.

-Justin

-- 
========================================

Justin A. Lemkul, Ph.D.
Research Scientist
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================