FV3PETS Calculation

[DavidHuber-NOAA]

Description
Is the FV3PETS calculation correct in config.resources#L535? It seems like nthreads_fv3 should not be used in the calculation of PETS.

Machines affected

All

Context

kjet was failing with out of memory errors for a C384 gdasfcst run, so I attempted to increase the number of threads 2, which quadrupled the number of cores needed, rather than doubled as expected.

Possible Implementation

Remove nthreads_fv3 from the calculation. Perhaps this would require nthreads_fv3 in another location.

[WalterKolczynski-NOAA]

When we moved to ESMF-managed threading (#1371), the calculation changed to include the threads because ESMF sorts it all out.

Hera has also seen some memory issues lately, but it seems to be from the more recent UFS update (#1509) that added write quilting (see #1572).

[DavidHuber-NOAA]

@WalterKolczynski-NOAA Ah, that makes sense. This is my first encounter playing with the threading after the ESMF managed threading and I tried editing config.ufs and the .xml myself instead of rerunning setup_xml.py. I'll keep this in mind. Thanks!