Opened 2 years ago

Closed 19 months ago

#2028 closed Task (fixed)

VALID-07_odea-AMM

Reported by: deazer Owned by: deazer
Priority: high Milestone: 2018 release-4.0
Component: OCE Version: trunk
Severity: minor Keywords:
Cc: Review: failed
MP ready?: no
Progress: The AMM15 configuration was tested initially with revision 9356. The configuration ran ok for 6 months but had a problem with the boundary conditions using the zerograd option. With this removed the model has been run successfully for 2 years with sensible looking model fields. The model was also run in WAD successfully for 2 year integration, and some further modification to stop SBC fluxes in dry areas lead to sensible fields. It was not reproducible at this revision in terms of different processor decompositions. Currently testing revision before the code was reorganized as rose suites need considerable updating to work wit the reorganized code. Revision 9490 is being integrated (currently 4 months) as before. It is still not perfectly reproducible across different processor decompositions but the differences are much smaller then at revision 9356.

Description

Context

Test the AMM15 1.5km configuration in the new trunk

Implementation plan


Commit History (0)

(No commits)

Change History (5)

comment:1 Changed 21 months ago by nemo

  • Status changed from new to assigned

comment:2 Changed 20 months ago by deazer

The AMM15 configuration was tested initially with revision 9356.

The configuration ran ok for 6 months but had a problem with the boundary conditions using the zerograd option.
With this removed the model has been run successfully for 2 years with sensible looking model fields.

The model was also run in WAD successfully for 2 year integration, and some further modification to stop SBC fluxes in dry areas
lead to sensible fields.

It was not reproducible at this revision in terms of different processor decompositions.

Currently testing revision before the code was reorganized as rose suites need considerable updating to work wit the reorganized code.

Revision 9490 is being integrated (currently 4 months) as before. It is still not perfectly reproducible across different processor decompositions but the differences are much smaller then at revision 9356

comment:3 Changed 20 months ago by mathiot

The AMM15 configuration was tested initially with revision 9356.

The configuration ran ok for 6 months but had a problem with the boundary conditions using the zerograd option.
With this removed the model has been run successfully for 2 years with sensible looking model fields.

The model was also run in WAD successfully for 2 year integration, and some further modification to stop SBC fluxes in dry areas
lead to sensible fields.

It was not reproducible at this revision in terms of different processor decompositions.

Currently testing revision before the code was reorganized as rose suites need considerable updating to work wit the reorganized code.

Revision 9490 is being integrated (currently 4 months) as before. It is still not perfectly reproducible across different processor decompositions but the differences are much smaller then at revision 9356

comment:4 Changed 20 months ago by mathiot

  • Progress modified (diff)

comment:5 Changed 19 months ago by deazer

  • Resolution set to fixed
  • Status changed from assigned to closed

With additional zero init flags on the Cray, -hzero -ez

AMM15 is now reproducible across different domain decomposition.

AMM15 it self is ok, But may warrant further investigation to underlying init bug in NEMO

Note: See TracTickets for help on using tickets.