Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

Minimum decomposed computational patch size

This post was from a previous version of the WRF&MPAS-A Support Forum. New replies have been disabled and if you have follow up questions related to this post, then please start a new thread from the forum home page.

RCarpenter

Member
In v4.0, in share/module_check_a_mundo.f90, a check has been added that prevents REAL and WRF from running if a patch is too small (< 10 cells in either dimension). This prevents for a modest set of grids (below) from running if the number of procs is below about 300 (both grids 1 and 2 fail the check). In this case, we might want to suffer some over decomposition on the outer 2 grids in order to achieve better performance on the innermost grid. Can (or should) this check be relaxed?

e_we = 181, 199, 259
e_sn = 181, 199, 325
 
Hi,

I believe the check should be the other way around, so that you cannot use too many processors for a small grid. So for the first domain, if you used 36 processors, the decomposition would be 6x6, and 181/6 is ~30 grid cells, which would be perfectly fine, as it's greater than 10.
 
Top