[MITgcm-devel] gmredi K3D test case

[Jean-Michel Campin]

Hi Michael,

I've just added the 1rst test case experiment for the gmredi K3D case,
in MITgcm_contrib/verification_other.
It's almost exactly the set up you sent me few months ago,
with some simplifications (fewer pkgs to compile).

Right now, it will be tested every day but only once,
on acesgrid (gfortran with MPI), with "-fast" 1/3 days
and with "-devel" 2/3 days. I could add more tests later on.

It's not yet part of the standard verification experiment,
since there are few issues:
a) need lapack libs (but it's not too bad, have it on acesgrid and
  on baudelaire).
b) there are few "un-initialised variables" that prevent to run 
 most of the automated test we run over night (using "-devel").

I started to add more initialisation (in gmredi_k3d.F,
fCoriU, fCoriV, SlopeX, SlopeY, urms) but did not finish it,
so I did not checked-in these additions.
May be, when you have time, you can check (e.g., on baudelaire
or acesgrid with optfile linux_amd64_gfortran and genmake2 option "-devel")
and add the missing initialisations.

Also, regarding lapack, I added a "genmake_local" file in build dir
to check and add the needed libs. But may be we could replace
all the #ifdef use_lapack with #ifdef HAVE_LAPACK (which is set in
the Makefile) and remove "#define use_lapack" from GMREDI_OPTIONS.h ?

Cheers,
Jean-Michel

PS: I cc to MITgcm-devel so that the other users will know.