Hi, I recently installed quantum espresso on a cluster machine equipped with openmpi and the OAR job submission system. Using the standard syntax: mpirun -np [n.o.cpus] /path/to/pw.x < inputfile.in > outputfile.out I found that the job would only use processors within a single node, such that there are several pw.x processes running on each CPU if the number of specified CPUs is larger than the number of CPUs per node, while other remain unused.
This problem is partially solved using the "-in/input" syntax: mpirun -np [n.o.cpus] /path/to/pw.x -in inputfile.in > outputfile.out (as in http://www.quantum-espresso.org/wiki/index.php/Running_on_parallel_machines#Trouble_with_input_files) That is: the job actually uses the number of requested CPUs, including when the number of requested CPUs is greater than the number of CPUs per node. However, a problem remains with the redirection of the output. Instead of writing the output in "ouputfile.out " pw.x writes it in the OAR output (OAR.jobID.stdout). In summary, the question is: how can I chose my output file when using the -in/input option ? Thanks for your help. Best regards. Sylvian -- ------------------------------------------------------------------------ *Dr. Sylvian CADARS* CEMHTI-CNRS 1D, Av. de la Recherche Scientifique 45071 Orl?ans Cedex 2, France ------------------------------------------------------------------------
