Hi YiLin, As Spencer has said, you need to compile with the NEKTAR_USE_MPI option set to ON. If haven’t done this and are trying to run a 16-core job on a cluster then I suspect you might actually just end up running 16 copies of same sequential job concurrently - this could explain why you’re getting an out of memory error. From the scripts you provided, I see that you’re running on a SLURM-based cluster. If you have already compiled Nektar++ with MPI support then it’s likely that you need to adjust the memory settings in your SLURM job submission script. The error suggests that a process has exceeded the available memory, although I’m not clear if it’s exceeded the available RAM on the node the process was running on or whether it’s just exceeded the amount of RAM you requested in the SLURM script and has been terminated for going outside the requested resources. In your job submission script you have requested 200GB memory - I believe this is a per node memory request, and since you’re requesting 16 tasks per node and requesting a single node, this would be the memory limit for the entire job. It depends, of course, on the size of your input mesh but it may be that more memory is required and you could try increasing this value. One final note, all parallelisation in Nektar++ is carried out via MPI so even though you’re running 16 processes on a single node, the processes will still communicate via MPI, hence the need to compile with the MPI option switched on, even when running multiple processes on a single node. Hope this information helps. Kind regards, Jeremy

On 22 Feb 2022, at 08:52, Sherwin, Spencer J <s.sherwin@imperial.ac.uk> wrote:

Hi YiLin

You need to set the NEKTAR_USE_MPI option to compile for multi-core processing.

Cheers, Spencer.

...

On 22 Feb 2022, at 03:00, Paul Liu <liuyilin.liu@gmail.com <mailto:liuyilin.liu@gmail.com>> wrote:

This email from liuyilin.liu@gmail.com <mailto:liuyilin.liu@gmail.com> originates from outside Imperial. Do not click on links and attachments unless you recognise the sender. If you trust the sender, add them to your safe senders list <https://spam.ic.ac.uk/SpamConsole/Senders.aspx> to disable email stamping for this address.

Hi Everyone,

I was just wondering if there are specific flags or other instructions that need to be set to compile nektar++ for multi-core processing. The main reason is that I am trying to run a 16 core job and I am often encountering a "c group out of memory" exception and I think it might be due to me not compiling things properly for mullti-core processing.

Thank you YiLin Liu 403-397-3457 liuyilin.liu@gmail.com <mailto:liuyilin.liu@gmail.com> MASc Student at CML <https://cml.mech.ubc.ca/> UBC _______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk <mailto:Nektar-users@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/nektar-users

_______________________________________________ Nektar-users mailing list Nektar-users@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/nektar-users