[gmx-users] Gromacs on Stampede

[Christopher Neale]

Why not put it in a slurm script and submit that script as a (probably single-node) job. It is not generally 
acceptable to use a large fraction of the head node of a shared resource for a substantial amount of 
time.

If your problem is different and of a gromacs nature, you may need to describe it better. (i.e., if you're really just saying that you can't use MPI with g_hbond then show us what you did, what happened, and likely somebody will be able to answer you. Personally, I don't think any of the analysis tools are MPI-enabled, but I could be wrong).

If you problem is really more about using stampede, you can get help directly by submitting an xsede help 
ticket (portal.xsede.org).

Chris.

-- original message --

Hello,

I have a question about running gromacs utilities on Stampede and hopefully someone can point me in the right direction. I compiled gromacs using instructions in this thread and mdrun works fine. Also, some utilities like g_energy, g_analyze (single - core utilities, I believe) seem to be working fine. 

I am interested in computing life time of hydrogen bonds and this calculation is  quite expensive. Is there a way to submit this as a job using 32 or higher cores? When I run g_hbond on my workstation (16 cores) it runs on 16 threads by default. However, I am not sure if it is a good idea to run it on Stampede without submitting it as a job. 

I noticed that g_hbond works on OpenMP, while gromacs was compiled for Mpi according to these instructions. Just curious, if that would be the reason and if there is a suitable workaround for this problem.

As always, help is greatly appreciated. 
Thanks,