Hello,
Anyone know how to run simuation in parallel computing with workstation?
Its tried mpirun -np 128
but its showing node: imk****
Executable : LIGGGHTS
Your help is much needed.
Thank you
A project by DCS Computing and CFDEMresearch
This website uses cookies to ensure you get the best experience on our website. By continuing to browse the site you are agreeing to our use of cookies. This website uses cookies. By continuing to browse the site you are agreeing to our use of cookies. More info
mschramm | Thu, 07/25/2019 - 03:00
How many nodes
Hello,
Are you working on a single computer or do you have access to a cluster.
If it is a single computer, you should simply need to the following
mpirun -n 128 liggghts -in in.liggghts
If it is a cluster, the command gets longer...
astral1000 | Fri, 08/16/2019 - 17:01
Hello,
Hello,
Could you please specify what the command would be to run CFDEM on a cluster?
Tsimur
mschramm | Tue, 08/20/2019 - 19:03
Might not help
Running LIGGGHTS on a cluster is not as straight forward and you should contact the cluster admin on how to do it for the cluster.
if you are not using a scheduler, like SLURM (https://slurm.schedmd.com/documentation.html), and simply connected some computers together and are only using the mpi interface then you could try the following.
mpirun --mca plm_rsh_no_tree_spawn 1 --mca btl_tcp_if_include em1 --verbose --show-progress --display-map --display-topo -np 36 -hostfile HOSTFILE liggghts -in in.liggghts -partition 36x1
This is what I used as a proof of concept before we moved to a larger cluster. The above command spawned 36 different simulations at the same time. If you would like to split the domain into 36 pieces, simply omit -partition 36x1
em1 was the internet adapter that all computers used to "talk" to the master node. --mca plm_rsh_no_tree_spawn 1
turns off tree spawning so every machine does NOT need keys to all other machines. With this command, the slave only ever needs the master key.
HOSTFILE is a file that contains the ip address of all the machines used in this cluster
and the thing that took me longer to debug than I would like to admit...