[Pw_forum] problems with running Quantum Espresso in parallel

Lorenzo Paulatto lorenzo.paulatto at impmc.upmc.fr
Tue Apr 22 12:20:40 CEST 2014

Check that the output of:
which mpirun
which mpif90
are in the same directory: if they are not, you are in troubles.

I would recommend you ask whoever set up the cluster for help, it will 
definitely save both of you time.


On 04/22/2014 12:02 PM, torstein.fjermestad at unito.it wrote:
> Dear Axel Kohlmeyer,
> thank you for your fast reply.
> Unfortunately, I am not really sure if I have understood all the
> details of your suggestions.
> The version of mpirun is 1.4.5. This information I got from the output
> of "mpirun -V":
> mpirun (Open MPI) 1.4.5
> I am configuring and compiling pw.x on the same (virtual) machine as
> Open MPI is installed. I do this in the following way:
> ./configure
> make pw
> I doubt that there are several versions of Open MPI installed on the
> same machine, but in case there are, is the version of Open MPI written
> somewhere in the output of "make pw"?
> I understand that this problem is not a pure Quantum Espresso problem,
> and I have therefore also described the problem to the mailing list of
> StarCluster.
> Thanks in advance for your help.
> Yours sincerely,
> Torstein Fjermestad
> On 17.04.2014 13:03, Axel Kohlmeyer wrote:
>> On Thu, Apr 17, 2014 at 6:41 AM,  <torstein.fjermestad at unito.it>
>> wrote:
>>> Dear all,
>>> I recently installed QE on a virtual cluster configured with
>>> StarCluster.
>>> QE configures and compiles without errors. However, when I submit a
>>> parallel calculation on 16 processors, the following is written near
>>> the
>>> start of the output file:
>>> Parallel version (MPI), running on     1 processors
>>> The line is repeated 16 times in the output. To me it seems like I
>>> am
>>> actually running 16 single processor calculations that all write to
>>> the
>>> same output file (stdout in this case).
>> this indicates that you are using an mpirun command that "belongs" to
>> a different MPI library than the one the pw.x you are using was
>> compiled with. please have a close look at the version output, to see
>> if it is really the pw.x you expect to be using. also re-check the
>> cluster documentation that you include the correct mpirun matching
>> your MPI.
>>> The way I submit this calculation, is the following:
>>> I write the following submit script (submit.sh):
>>> cp /path/to/executable/pw.x
>>> mpirun -np 16 pw.x -in input.inp
>> here is a problem: unlike on windows, the current directory is not
>> part of the search path, so you would have to use './pw.x' instead of
>> 'pw.x'
>> to use the pw.x directory in your current working directory (unless
>> you have changed your profile to have '.' included in your $PATH
>> variable, which is a very, very bad idea).
>>> Then I submit the job with the following command:
>>> qsub -cwd -pe orte 16 ./submit.sh
>>> The queuing system of StarCluster is Open Grid Scheduler.
>>> For the line in the submit script, I have also tried several
>>> alternatives such as:
>>> mpirun pw.x -in input.inp
>>> mpirun pw.x -inp input.inp
>>> mpirun -np 16 pw.x -inp input.inp
>>> mpirun -np 16 pw.x < input.inp
>>> In the archives of this mailing list I have seen some similar
>>> problems,
>>> but in spite of this I was still not able to solve my problem.
>>> I would appreciate very much if someone could give me suggestions on
>>> how to solve the problem.
>>> Thanks in advance.
>>> Yours sincerely,
>>> Torstein Fjermestad
>>> University of Turin,
>>> Italy
>>> _______________________________________________
>>> Pw_forum mailing list
>>> Pw_forum at pwscf.org
>>> http://pwscf.org/mailman/listinfo/pw_forum
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum

Dr. Lorenzo Paulatto
IdR @ IMPMC -- CNRS & Université Paris 6
+33 (0)1 44 275 084 / skype: paulatz
23-24/4é16 Boîte courrier 115, 4 place Jussieu 75252 Paris Cédex 05

More information about the users mailing list