Skip to content

Prereleases with Angus' MOM6 adaptive grid coordinate#140

Draft
dougiesquire wants to merge 4 commits intomainfrom
mom6-adaptive-grid
Draft

Prereleases with Angus' MOM6 adaptive grid coordinate#140
dougiesquire wants to merge 4 commits intomainfrom
mom6-adaptive-grid

Conversation

@dougiesquire
Copy link
Collaborator

@dougiesquire dougiesquire commented Aug 14, 2025

This PR is to produce prerelease deployments of ACCESS-OM3 with @angus-g's MOM6 adaptive grid (AG) coordinate.

The development of the AG coordinate has focused on MOM symmetric memory mode. However ACCESS-OM3 deployments/configurations use asymmetric memory. The plan at this point is to:

  1. deploy a version of ACCESS-OM3 that is identical to 2025.05.001 (as used in the 25km 1.0-beta release), except use MOM6 symmetric memory. ADDED: deployed as access-om3/pr140-2
  2. deploy a version of ACCESS-OM3 that is identical to 2025.05.001, except use MOM6 symmetric memory and include the MOM6 AG changes. ADDED: deployed as access-om3/pr140-5
  3. check whether the deployment from 1:
    a. is reproducible across restarts
    b. reproduces the existing 25km beta release control run

If 3a fails, we'll need to rethink.
If 3b:

  • passes: Angus can use the existing 25km beta control run for the z-star comparison
  • fails: Angus will need to rerun the 25km beta release config using 1. (or 2.) to create the required z-star data.

🚀 The latest prerelease access-om3/pr140-5 at e606bc4 is here: #140 (comment) 🚀


🚀 The latest prerelease access-om3/pr140-6 at f64842d is here: #140 (comment) 🚀

@dougiesquire dougiesquire self-assigned this Aug 14, 2025
@dougiesquire dougiesquire changed the title Prereleases for ACCESS-OM3 with Angus' MOM6 adaptive grid coordinate Prereleases with Angus' MOM6 adaptive grid coordinate Aug 14, 2025
@github-actions
Copy link

🚀 Attempted to deploy access-om3 Prerelease pr140-1 with commit ff6a68c

🖥️ Gadi Deployment ❌

@dougiesquire
Copy link
Collaborator Author

!redeploy

@github-actions
Copy link

🚀 Attempted to deploy access-om3 Prerelease pr140-2 with commit ff6a68c

🖥️ Gadi Deployment ✔️

Usage Instructions

access-om3, defined in ./spack.yaml, will be deployed to Gadi as:

  • 2025.05.001 as a Release (when merged).
  • pr140-2 as a Prerelease (during this PR).

This Prerelease is accessible on Gadi using:

module use /g/data/vk83/prerelease/modules
module load access-om3/pr140-2

When using the above modules, the binaries shall be on your $PATH.

For advanced users, this Prerelease is also accessible on Gadi via /g/data/vk83/prerelease/apps/spack/0.22/spack in the access-om3-pr140-2 environment.
Due to inode-saving measures, one will have to manually untar the environment metadata before environment activation with tar -xf .spack-env .spack-env.tar. It will require one to have write privileges.

Configuration Information

This Prerelease is deployed using:

If the above was not what was expected, commit changes to config/versions.json in this PR.

@github-actions
Copy link

🚀 Attempted to deploy access-om3 Prerelease pr140-3 with commit 7fb6ddd

🖥️ Gadi Deployment ❌

@dougiesquire
Copy link
Collaborator Author

!redeploy

@github-actions
Copy link

🚀 Attempted to deploy access-om3 Prerelease pr140-4 with commit 7fb6ddd

🖥️ Gadi Deployment ❌

@angus-g
Copy link
Collaborator

angus-g commented Aug 14, 2025

Aha, now we're getting somewhere. Looks like I'll have to make a couple of other changes, maybe add to the CMakeLists.txt?

@dougiesquire
Copy link
Collaborator Author

Yeah, looks like you might just have to add src/ALE/filter_utils.F90 to the list of target sources in CMakeLists.txt

@github-actions
Copy link

🚀 Attempted to deploy access-om3 Prerelease pr140-5 with commit e606bc4

🖥️ Gadi Deployment ✔️

Usage Instructions

access-om3, defined in ./spack.yaml, will be deployed to Gadi as:

  • 2025.05.001 as a Release (when merged).
  • pr140-5 as a Prerelease (during this PR).

This Prerelease is accessible on Gadi using:

module use /g/data/vk83/prerelease/modules
module load access-om3/pr140-5

When using the above modules, the binaries shall be on your $PATH.

For advanced users, this Prerelease is also accessible on Gadi via /g/data/vk83/prerelease/apps/spack/0.22/spack in the access-om3-pr140-5 environment.
Due to inode-saving measures, one will have to manually untar the environment metadata before environment activation with tar -xf .spack-env .spack-env.tar. It will require one to have write privileges.

Configuration Information

This Prerelease is deployed using:

If the above was not what was expected, commit changes to config/versions.json in this PR.

@dougiesquire
Copy link
Collaborator Author

  1. check whether the deployment from 1:
    a. is reproducible across restarts
    b. reproduces the existing 25km beta release control run

This is tested in ACCESS-NRI/access-om3-configs#721, which shows that access-om3/pr140-2:

  • is reproducible across restarts, but
  • does not reproduce the existing 25km beta release control run

So @angus-g, unfortunately I don't think you'll be able to compare to the existing 25km beta control run for the z-star comparison. You'll need to do your own z-star run.

To reiterate what's above, access-om3/pr140-5 is identical to the build used in the 25km 1.0-beta release, except that it uses MOM6 symmetric memory and includes your MOM6 AG changes. Let me know if you'd like any help setting up you experiment configurations.

@angus-g
Copy link
Collaborator

angus-g commented Aug 21, 2025

Thanks @dougiesquire. I'm just coming back to this after thinking about the symmetric/asymmetric reproducibility thing. Maybe it's a new compiler thing, but the tc2.b unit test which is using AG is indeed failing on the PR at ACCESS-NRI/MOM6#25 for the Linux runner. However, on the Mac runner and when I build and run the test myself, it seems to pass the unit tests. Perhaps this means that I can get away with using asymmetric memory and being able to use the existing beta release control run?

@dougiesquire
Copy link
Collaborator Author

dougiesquire commented Aug 25, 2025

Sorry for the delay @angus-g. Your call I think. I'll add another prerelease to this PR that is identical to the 25km 1.0-beta release, except includes your MOM6 AG changes.

@angus-g
Copy link
Collaborator

angus-g commented Aug 25, 2025

Sorry for the delay @angus-g. Your call I think. I'll add another prerelease to this PR that is identical to the 25km 1.0-beta release, except includes your MOM6 AG changes.

Thanks! I'll see how it goes.

@github-actions
Copy link

🚀 Attempted to deploy access-om3 Prerelease pr140-6 with commit f64842d

🖥️ Gadi Deployment ✔️

Usage Instructions

access-om3, defined in ./spack.yaml, will be deployed to Gadi as:

  • 2025.05.001 as a Release (when merged).
  • pr140-6 as a Prerelease (during this PR).

This Prerelease is accessible on Gadi using:

module use /g/data/vk83/prerelease/modules
module load access-om3/pr140-6

When using the above modules, the binaries shall be on your $PATH.

For advanced users, this Prerelease is also accessible on Gadi via /g/data/vk83/prerelease/apps/spack/0.22/spack in the access-om3-pr140-6 environment.
Due to inode-saving measures, one will have to manually untar the environment metadata before environment activation with tar -xf .spack-env .spack-env.tar. It will require one to have write privileges.

Configuration Information

This Prerelease is deployed using:

If the above was not what was expected, commit changes to config/versions.json in this PR.

@dougiesquire
Copy link
Collaborator Author

Just confirming that access-om3/pr140-6 is reproducible across restarts and does reproduce the existing 25km beta release control run - tested here.

@angus-g
Copy link
Collaborator

angus-g commented Aug 26, 2025

Perfect, thanks!

@dougiesquire
Copy link
Collaborator Author

@angus-g are you still using these prerelease deployments? If not, I'll close the PR.

@angus-g
Copy link
Collaborator

angus-g commented Jan 27, 2026

I have been, we have 20 years run using the 140-6 deployment. Not sure if we would need more though. @AndyHoggANU?

@dougiesquire
Copy link
Collaborator Author

All good - we can leave it open :)

@AndyHoggANU
Copy link

My view is that we don't need a longer run, and that it passes the minimal test of being functional.
I want to finish the write-up (at least the figures) and run them past a few others to be certain we are happy, so maybe keep it open for just a little while longer?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants