[Nek5000-users] Issues running on Mira

[nek5000-users at lists.mcs.anl.gov]

Hi Jefferson,


What are lx1 and left set to in your SIZE file?


It's possible that you're running out of memory.


If you have, say, 16000 elements total, then you would only need lelt=2

to run on 8192 MPI ranks.


Note that we generally run with -c32 mode, 2 ranks per core, as this typically

makes better use of the same number of node hours.


Please let me know if this helps.


Thanks,

Paul


________________________________
From: Nek5000-users <nek5000-users-bounces at lists.mcs.anl.gov> on behalf of nek5000-users at lists.mcs.anl.gov <nek5000-users at lists.mcs.anl.gov>
Sent: Monday, October 9, 2017 2:37:57 PM
To: nek5000-users at lists.mcs.anl.gov
Subject: [Nek5000-users] Issues running on Mira


I'm currently attempting to run Nek5000 on the Argonne supercomputer (Mira) using 512 nodes, but if using all available cores (16 per node) the code seems to hang after some point. All attempts have been made with a run time of 1 hour. When running on 512 nodes with a single core on each node, the code runs fine. Is it possible that start-up for ~8000 cores takes more than an hour or is there something else that could be causing this issue? Note that we definitely have a sufficient number of elements to run on ~8000 nodes.

Thanks,

Jefferson Davis

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/nek5000-users/attachments/20171010/429115fb/attachment.html>