[MITgcm-support] scalability of high-resolution runs

[Maura Brunetti]

Dear Dimitris,

thank you for your answer. I know how to include blank tiles in the scaling estimates (it is well explained here: http://mitgcm.org/pipermail/mitgcm-devel/2009-May/003773.html) but how to include overlap regions?

Do you know if there are publications on speedup on the huge-resolution global llc set-ups that you suggested, on any machine (BGQ or others…)? Is the scaling the same for ocean and atmospheric runs? I am interested in trying to couple atmosphere and ocean in an efficient way, but before I would like to be sure that the scalability of  each component is near-linear.  I have performed tests on cubed-sphere192x28 on BlueGeneP and the results are not encouraging (see attached file).

Regards,

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
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20141024/ce6813d5/attachment-0001.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Efficiency_BlankTilesMonitorCorrections_cs192.eps
Type: image/eps
Size: 12416 bytes
Desc: Efficiency_BlankTilesMonitorCorrections_cs192.eps
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20141024/ce6813d5/attachment-0002.bin>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Speedup_BlanktilesMonitorCorrections_cs192.eps
Type: image/eps
Size: 12545 bytes
Desc: Speedup_BlanktilesMonitorCorrections_cs192.eps
URL: <http://mitgcm.org/pipermail/mitgcm-support/attachments/20141024/ce6813d5/attachment-0003.bin>