COSIMA TWG Meeting Minutes 2026

COSIMA TWG
, ,
13

17 June 2026

@jbisits, @dougiesquire, @aekiss, @AndyHoggANU, @cbull, @manodeep, @sofarrell, @anton, @ezhilsabareesh8

Numerical mixing diagnostic

Joey gave a presentation on his diagnostic for advection scheme variance production. Same presentation that he gave at COSIMA WG a few weeks ago. Joey has a PR to add the diagnostic into the ACCESS-NRI MOM6 fork. He asked about turning it on in a future ACCESS-OM3 run and everyone agreed this is a great idea.

Andy is interested in using the diagnostic with the adaptive grid.

Implementation in MOM6 is very close - see linked PR above. Just need advice from upstream on how to resolve a dimensionality test that is failing due to very small values.

Actions:

  • Waiting for advice from Bob Hallberg

Regional stencil naming convention

Ash/Helen/Dougie would like to add a branch to access-om3-configs that will act as a “stencil” for regional configurations.

This will act as a template, but also will serve as something we can test against when we make updates/test upstream changes.

We will want it to include BGC in the future, but regional-mom6 is not quite ready for that.

Suggestions for branch name:

  • regional-stencil
  • regional-relocatable
  • regional-template

People like regional-template

Actions:

  • Dougie/Ash/Helen will rename an existing branch of Helen’s and update.

8km configuration

Chris has 8km configuration originally set up by Angus running. He has been making changes in a branch in his personal fork of access-om3-configs.

Some discussion about whether this is ready to merge. Branch is behind other configs, inputs are spread across multiple projects, some input files still need changes…

Actions:

  • Chris will move the feature branch in his fork into the in the access-om3-configs repo and open a new PR.

OM3 optimization

Manodeep has been looking into why diagnostic density-level remapping is so slow.

Been testing in a idealised configuration. Has found a few places where changes can have a significant impact (see this MOM6 branch):

  • Reorder a few checks in remapping code
  • Force inline a number of routines
  • Added wrapper function for density deriv - leads to inlining and improved vectorization\

Also tested a number of other things, but these were the smallest changes with the biggest impact.

Actions:

  • Manodeep will create a PR with some suggested changes. Probably a good idea to ping Marshall early for thoughts, particularly on in-source compiler directives