Skip to content

Mixed precision fixes for mass conservation #2

@FlorianDeconinck

Description

@FlorianDeconinck

Mass conservation at f32: GMAO did the work of keeping mass conservation for f32 calculation.

Unfortunately those changes will invalidate 8.1.3 version of the FV3GFS we are using for translate test. We need a new strategy and/or data set.

Another task is to move D_SW fluxes (dpx, x/yflux, cx/y in original Fortran) to always be f64 for the GEOS model. And ask NOAA if they also want this change. Original code is in @fvdycore/model/sw_core.F90:d_sw

Parent: GEOS-ESM/SMT-Nebulae#55


  • We have a set of data on f32 generated from GEOS 11.5.2
  • D_SW mixed precision is ported
  • Mass conservation fixes for f32 are ported
  • Dynamics dpx fix to mass is implemented

Metadata

Metadata

Labels

No labels
No labels

Type

No type
No fields configured for issues without a type.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions