modifies array size to NCellsSize

[alicebarthel]

Modifies array sizes and loop bounds to NCellsSize (fixes #367)
Checklist

• Linting
  • Pre-commit run locally
• Building
  • CMake build does not produce any new warnings from changes in this PR
• Testing
  • CTest unit tests pass on pm-cpu_gnu Debug
  • The Polaris omega_pr test suite: job 50825116 passes all but the del4 convergence test

[alicebarthel]

@mwarusz Would that solve your issue?
If so, I'll do a similar update to the vertMix Arrays.

[mwarusz]

@alicebarthel

Thanks ! Looks good, but I don't think it was necessary to modify the loop bounds. Could you revert those changes ?

[alicebarthel]

@mwarusz Is there ever a case where we might read in values in the arrays of size NCellsSize that were not computed (given that the loop bounds are NCellsAll) and still have their default value? Wouldn't we want to avoid this?
Clearly I am still not understanding the subtleties correctly...

[alicebarthel]

It now fails the IO_STREAM ctest... I'll see what I can find.

[mwarusz]

@alicebarthel

Just checking that you know about #356.

[alicebarthel]

Thanks @mwarusz! yes, it sounded familiar so I looked it up and am rerunning ctests in a clean folder now... And passes! 🎉

[alicebarthel]

Update on testing: 1 test fails but it's the del4 convergence

Polaris omega_pr suite

• Baseline workdir: /global/homes/a/abarthel/pscratch/polaris-scratch/baseline_omega_pr
• Baseline build: /global/homes/a/abarthel/pscratch/polaris-scratch/baseline_omega_pr/build
• PR build: /global/homes/a/abarthel/pscratch/omega_scratch/pr-fix-arraysize-cpugnu-v2
• PR workdir: /pscratch/sd/a/abarthel/polaris-scratch/pr-fix-arraysize-cpu-v2
• Machine: pm-cpu
• Compiler: gnu
• Build type: Debug
• Log: /pscratch/sd/a/abarthel/polaris-scratch/pr-fix-arraysize-cpu-v2/polaris_omega_pr.o50825116
• Result:
  • Diffs (1 of 8):
    • ocean/planar/manufactured_solution/convergence_both/del4

Is that something that we expect (I've heard del4 failures are common) and that we expect to be resolved by a fix recently discusses?

[mwarusz]

Is that something that we expect (I've heard del4 failures are common) and that we expect to be resolved by a fix recently discusses?

@alicebarthel

There is a known issue with this test (#376) that has a fix (#377). However, this issue should only affect you if your pr run used a different number of MPI tasks than your baseline run. I don't have access to your baseline workdir, so I couldn't check if that was the case.

[mark-petersen]

Ran a quick check, and this passes all cTests on PM CPU.