ICB: extra possible cleanup to investigate
During the cleaning up task #220 (closed), I only focused myself on the cleaning of the extra halo variable with objectif to reproduced the same results of the main. During this I figured out some extra stuff that need to be checked and tested:
-
masking of utau_icb (not sure it is needed). Maybe even better to save wind straight from sbcblk. -
change hi_icb by hm_i change results => at the end, hm_i should be use (1 2d array less) -
sanity check on bounds is maybe no more working as array start from 1 and not 0 as for the icb extra halo array in the previous version => sanity check on index < 0 no more valid. -
limitation of the cfl to 0.4 can be ease to 0.9 but need changes on the sanity check (Ukii<1 and Vkij<1 I think) -
Can we use already define Nis0 and Nie0 instead of nicbdj, nicbej (used in icbclv.F90). -
Have a fresh look on the mpi of icb to see if it can be simplified. I have the feeling that ICB are send E/W then N/S to handle diagonal MPI communication. Is using mpinei(jpne)
able to do the job ?
This ticket is a master ticket and fixing it will be done by opening individual issues. This is mostly for reminder and tracking progress purposes.