[MITgcm-devel] lab_sea

[Martin Losch]

Hi,
once you define ALLOW_SEAICE_EVP you can switch between LSR and EVP  
solver with a run-time flag (SEAICEuseEVP), so I agree with Dimitris,  
test both.
My guess is, that EVP should be easier to adjoin.

M.

On 9 Dec 2006, at 02:34, Dimitris Menemenlis wrote:

> JM, in practice the choice between LSR/EVP should be simply one of  
> speed for particular grid configuration, etc., that is used.   
> Jinlun warns there may be some stability issues with the EVP solver  
> at certain resolutions due to yield curve, etc., but do not have  
> personal experience with any of these problems.
>
> Re adjoint, I am not sure if one is preferable than the other.  My  
> guess is that EVP might work better because LSR is an iterative  
> scheme?  Patrick, any words?
>
> So definitely both need to stay for time being.  If it's not too  
> much trouble, testing both using 2 tests with similar setup would  
> be ideal.  That way we make sure that neither gets broken as we  
> change the seaice/thsice code.
>
> D.
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel