Yesterday I did some tests on Rice cluster BlueBiou. I had some issues when setting the option num_cores_per_sim_cu=2. If I do that, the bash script (something like radical_pilot_cu_launch_script-halUfC.sh) in the unit folder used to run gromacs simulations looks like:
!/bin/bash -l
cd /gpfs-biou/jp43/radical.pilot.sandbox/pilot-545fd4c7a208ed6788591a9d/unit-545fd576a208ed6788591ad8
by using this script, the file out.gro (output) will contain twice the number of configurations of the input .gro file and of course the files should be of the same length, because we run one trajectory per starting point (the out.gro should contain only the endpoints of the trajectories). Eventually, the GROMACS/ LSDMap workflow will crash because the .out file does not have the expected number of configs. However, the workflow works just fine when setting num_cores_per_sim_cu=1.
Yesterday I did some tests on Rice cluster BlueBiou. I had some issues when setting the option num_cores_per_sim_cu=2. If I do that, the bash script (something like radical_pilot_cu_launch_script-halUfC.sh) in the unit folder used to run gromacs simulations looks like:
!/bin/bash -l
cd /gpfs-biou/jp43/radical.pilot.sandbox/pilot-545fd4c7a208ed6788591a9d/unit-545fd576a208ed6788591ad8
/opt/apps/openmpi/1.4.3-gcc/bin/mpiexec -n 2 -host c18,c18, /bin/bash "-l" "-c" "export grompp_options=\"\" mdrun_options=\"\" ndxfile=\"\" && . run.sh grompp.mdp start.gro topol.top out.gro 2"
by using this script, the file out.gro (output) will contain twice the number of configurations of the input .gro file and of course the files should be of the same length, because we run one trajectory per starting point (the out.gro should contain only the endpoints of the trajectories). Eventually, the GROMACS/ LSDMap workflow will crash because the .out file does not have the expected number of configs. However, the workflow works just fine when setting num_cores_per_sim_cu=1.