[Mesa-users] Run MESA on the supercomputer

[Warrick Ball]

Hi Wenyu,

MESA is parallelized using OpenMP, which only parallelizes on a single node.  You should only run each instance of `./star` (or `./rn` or `./re`) on a single node.

If you're planning multiple MESA runs, you can set up all your data appropriately then submit a job to your supercomputer for each run.  This is, for example, how I and many others compute grids of models that involve multiple independent evolution tracks.

That MESA appeared to run on 1 node with 20 cores (or threads) is probably because that's how many cores were available on the main node that picked up the MESA job.  When you do run MESA on a cluster/supercomputer, you should make sure that your `OMP_NUM_THREADS` environment variable matches the number of threads you request from the scheduler.

If you do plan to run MESA on many threads, you may want to check that your runs use multiple threads efficiently.  How MESA's performance scales with threads depends a lot on what you want to do with MESA.

Cheers,
Warrick

___________

Warrick Ball
Postdoc, School of Physics and Astronomy
University of Birmingham, Edgbaston, Birmingham B15 2TT
W.H.Ball at bham.ac.uk
+44 (0)121 414 4552

On Wed, 18 Aug 2021, mesa-users at lists.mesastar.org wrote:

> 
> Dear users:
> 
> 
> I install mesa-r12778 on supercomputer and run a example in the test_suite.
> 
> I submitted the job with 10 nodes, which include 64 cores for each nodes, 
> 
> however, I find the job "star" only run in the first node and used about 20 cores. 
> 
> 
> I have set the environment variable as follow:
> 
> export OMP_NUM_THREADS=1000
> 
> 
> So, how can I run MESA on the supercomputer ? Is there any other modification
> 
> or setup necessary?
> 
> 
> Thanks,
> 
> Wenyu
> 
> 
>