[MITgcm-support] scalability of high-resolution runs

Dimitris Menemenlis dmenemenlis at gmail.com
Thu Oct 23 13:52:49 EDT 2014


Dear Maura, there is some specific CS50 set-ups available here:
http://mitgcm.org/viewvc/MITgcm/MITgcm_contrib/high_res_cube/code-mods/

I am not aware of publications that describe scalability/speed-up analysis for CS510.
It's really machine dependent.  But CS510 is a relatively small configuration for
modern hardware and scalability is typically near-linear (for high-end, low-latency
hardware) provided you include the overlap regions (and blank tiles, if any)
in your scaling estimates.

If you wanted to aim for even higher-resolution global MITgcm set-ups, there is
a bunch of lat-lon-cap configurations available here:
http://mitgcm.org/viewvc/MITgcm/MITgcm_contrib/llc_hires/

Regards

Dimitris Menemenlis

> On Oct 23, 2014, at 9:54 AM, Maura Brunetti <Maura.Brunetti at unige.ch> wrote:
> 
> Dear all, 
> 
> do you know any publication on the core number needed for running high-resolution configurations  (i.e. cs510 or ecco2 simulations) or any link to scalability/speedup analysis for this kind of configurations?   
> Any help appreciated!
> Thanks 
> 
> Maura
> 
> 
> --
> Dr. Maura Brunetti
> Institute for Environmental Sciences (ISE)
> University of Geneva
> Site de Battelle / D, 7 route de Drize
> CH-1227 Carouge / GE -- Switzerland
> 
> 
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list