[MITgcm-devel] stripes in adjoint along tile edges

[Jean-Michel Campin]

Hi,

Just a comment regarding minimum ovelap size due to tracer advection scheme:
it's clearly (I think) set at the top of S/R GAD_INIT_FIXED for all 
tracer advection schemes: gad_init_fixed.F, lines 45-57.

And if using multi-dim advect with useCubedSphereExchange, Olx,Oly
have to be >= 2*OlMinSize ; also with GM_AdvForm, has to be >= OlMinSize+1.
I don't remember the condition for the combination of both, but I 
believe that the check is right (in gad_check.F, lines 98-118).

Cheers,
Jean-Michel

On Thu, Nov 20, 2014 at 08:22:14AM -0800, Dimitris Menemenlis wrote:
> Seems sufficient.  Not an expert.
> Hopefully JM and/or Patrick will chime in.
> 
> > On Nov 20, 2014, at 7:48 AM, Martin Losch <Martin.Losch at awi.de> wrote:
> > 
> > The overlap is 4 (as in MITgcm_contrib/arctic/cs_36km) and I am not using cubed sphere exchanges (in fact, not even exch2). 3 should be enough for advection schemes 30 or 33, shouldn’t it?
> > What do you suggest to use as overlap? 6?
> > 
> > M.
> 
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel