Test using oasis3-mct master branch

[dougiesquire]

This PR is to test a prerelease build using the oasis3-mct master branch and demonstrate that this does not change answers.

[dougiesquire]

!test repro

[github-actions]

✅ The Bitwise Reproducibility Check Succeeded ✅

When comparing:

• preindustrial+concentrations-oasis3-master (checksums created using commit 5d216e9), against
• dev-preindustrial+concentrations (checksums in commit 9fa7b7f)

Further information

The experiment can be found on Gadi at /scratch/tm70/repro-ci/experiments/access-esm1.6-configs/5d216e901d34ba23ebf9a787be7b1433be3bcf12, and the test results at https://github.com/ACCESS-NRI/access-esm1.6-configs/runs/48484765395.

The checksums generated by this !test command are found in the testing/checksum directory of https://github.com/ACCESS-NRI/access-esm1.6-configs/actions/runs/17097195840/artifacts/3808177447.

The checksums compared against are found here https://github.com/ACCESS-NRI/access-esm1.6-configs/tree/9fa7b7f4545bc6764bb175dcdf4b1b835f0fe8cf/testing/checksum

Test summary:
✅ test_repro_historical
✅ test_repro_determinism

[dougiesquire]

!test repro

[github-actions]

❌ The Bitwise Reproducibility Check Failed ❌

When comparing:

• preindustrial+concentrations-oasis3-master (checksums created using commit 0fde150), against
• dev-preindustrial+concentrations (checksums in commit 9fa7b7f)

Further information

The experiment can be found on Gadi at /scratch/tm70/repro-ci/experiments/access-esm1.6-configs/0fde15019f4fd08670695f185862d03360655587, and the test results at https://github.com/ACCESS-NRI/access-esm1.6-configs/runs/48961712100.

The checksums generated by this !test command are found in the testing/checksum directory of https://github.com/ACCESS-NRI/access-esm1.6-configs/actions/runs/17253731746/artifacts/3859004845.

The checksums compared against are found here https://github.com/ACCESS-NRI/access-esm1.6-configs/tree/9fa7b7f4545bc6764bb175dcdf4b1b835f0fe8cf/testing/checksum

Test summary:
❌ test_repro_determinism
✅ test_repro_historical

[dougiesquire]

test_repro_determinism failed with:

CICE: ERROR failed to open input_ice.nml. Error code:  29  - file not found, un
 it 11, file /scratch/tm70/repro-ci/experiments/access-esm1.6-configs/0fde15019f
 4fd08670695f185862d03360655587/lab/work/exp_1d_runtime_repeat/ocean/input_ice.n
 ml

This is a known issue that occurs randomly. Will rerun.

[dougiesquire]

!test repro

[github-actions]

✅ The Bitwise Reproducibility Check Succeeded ✅

When comparing:

• preindustrial+concentrations-oasis3-master (checksums created using commit 0fde150), against
• dev-preindustrial+concentrations (checksums in commit 9fa7b7f)

Further information

The experiment can be found on Gadi at /scratch/tm70/repro-ci/experiments/access-esm1.6-configs/0fde15019f4fd08670695f185862d03360655587, and the test results at https://github.com/ACCESS-NRI/access-esm1.6-configs/runs/48962303635.

The checksums generated by this !test command are found in the testing/checksum directory of https://github.com/ACCESS-NRI/access-esm1.6-configs/actions/runs/17253951757/artifacts/3859073925.

The checksums compared against are found here https://github.com/ACCESS-NRI/access-esm1.6-configs/tree/9fa7b7f4545bc6764bb175dcdf4b1b835f0fe8cf/testing/checksum

Test summary:
✅ test_repro_historical
✅ test_repro_determinism

[dougiesquire]

I've done some basic testing of the performance impact of this change. I've run two experiments:

• Exp 1 uses the current dev-preindustrial+concentrations config as is (uses access-esm1p6/2025.07.000).
• Exp 2 uses the same configuration, but modified to use access-esm1p6/pr123-2. This prerelease uses oasis3-mct built from the master branch using the access-om2 version.

The experiments were run at the same time. Times below are the PAYU_WALLTIME from job.yaml for 2 x 1 year:

	Exp 1	Exp 2
year 1	1hr 15mins	1hr 15mins
year 2	1hr 15mins	1hr 16mins

So moving to using the master branch of oasis3-mct and switching from -xCORE-AVX512 to -axCORE-AVX2:

1. does not change answers - see above
2. has no obvious impact on performance

@manodeep, @harshula, @blimlim, @penguian, I think we agreed that with the above confirmed we are happy to use -axCORE-AVX2 in ACCESS-ESM1.6 as in ACCESS-OM2. I'll follow up with the change to the oasis3-mct SPR.