I see, but could the recompiling be avoided? Thank you! El 07/04/2014 20:01, <[email protected]> escribió:
> > You don't seem to be launching the job properly. See: > > http://slurm.schedmd.com/mpi_guide.html#mvapich2 > > > Quoting Sergio Iserte <[email protected]>: > > Hello everybody, >> >> I am using SLURM 2.6.2 with MVAPICH2 in a cluster (with the configuration >> attached). >> >> When I try to submit this command: >> >> srun --gres=gpu:1 $path_mpirun -np 1 $path_app/lmp_g++ -var x 2 -var y 2 >>> -var z 4 -sf cuda < $path_test/in.lj >>> >> >> I get the following error from srun: >> >> srun: error: Unable to create job step: Access/permission denied >>> >> >> Moreover, you will find attached the log of the controller and the compute >> node. >> >> Any idea about what happend? >> >> Regards! >> >> -- >> *Sergio Iserte Agut, research assistant,* >> *High Performance Computing & Architecture* >> *Jaume I University (Castellón, Spain)* >> > >
