[Mitgcm-support] Re: exch_xz, exch_yz

[mitgcm-support at dev.mitgcm.org]

Hi Chris,

I agree, it's only for the special case (which we have in mind)
of obcs along lon/lat.
Discussed already with Alistair that the solution
is to exch after applying obcs, i.e. the same solution
that you suggest, I guess.

Patrick

Chris Hill wrote:
> 
> Hi Patrick,
> 
>  Would it be very hard to impose an exchange after
> OBC application rule? Looked at _xz, _yz. You did
> a good job but I see two probs.
> 
>  1. How does this help for general OBCS, which might
>     not orient with grid.
> 
>  2. For cube there are mapping YZ <-> XZ, YZ <-> -XZ
>     etc... for certain connections. Enabling this
>     would require further effort.
> 
>  Both 1. and 2. are fixed by imposing a post-OBC exchange.
>  Plus it makes the merge easier! And it means we can
>  survive without extending the exch() operator for a
>  while longer.
> 
>  Alistair what do you think?
> 
> Chris