Opened 8 years ago

Closed 8 years ago

Last modified 4 years ago

#966 closed Bug (invalid)

GYRE using linear eos (nn_eos=1) very slow

Reported by: fabien.roquet Owned by: nemo
Priority: low Milestone:
Component: OCE Version: release-3.4
Severity: Keywords: EOS
Cc:

Description

When running the GYRE configuration (EXP00 experiment) with the linear equation of state (temperature only), i.e. nn_eos=1, the run is abnormally slow…

I'm not sure what could cause this slowing down, but I ran it for 10 mins, and the run was still not finished, while it takes only 3 mins with nn_eos=0 or nn_eos=2.

Commit History (0)

(No commits)

Change History (3)

comment:1 Changed 8 years ago by rblod

Salut Fabien
Not totally sure, but I have in mind that GYRE in 3.4 with linear EOS is blowing up after few hundred time step, It may be whas is happening in your case:

  • your barotropic solver iterating more and more (chack solver.stat)
  • depending on your computer, the model may also be computing and writing NaN
  • the run failed to stop properly after blowing up, and get stuck somewhere

You should check ocean.output if there is no error
Rachid

comment:2 Changed 8 years ago by gm

  • Resolution set to invalid
  • Status changed from new to closed

In fact in linear eos 1 ( rho=f(T) ) it is impossible to have a isoneutral operator. Indeed, iso-neutral surfaces = isotherms in this case ==⇒>> no lateral diffusivity in the model : it will blow up
That's logic, it is the user responsibility to properly set the choices in terms of physics and numerics….

comment:3 Changed 4 years ago by nicolasmartin

  • Keywords EOS added; eos removed
Note: See TracTickets for help on using tickets.