[MITgcm-devel] thsice_step_fwd.F changes with adjoint

[Patrick Heimbach]

Hi Jean-Michel,

I think my last checkin fixes the problem
(and I've now set the default to #undef THSICE_OLD_STEP_FWD as desired).

My assessment is that TAF thinks that
re-initialization of three variables before call to S/R thsice_extend
is not necessary because they are computed within S/R thsice_extend,
and so ignores it in the TLM code.
Problem is that within thsice_extend they are only computed in the interior,
giving wrong values in overlap in absence of re-initialization.

Cheers
-Patrick

On Feb 8, 2012, at 10:01 AM, Jean-Michel Campin wrote:

> Hi Patrick,
> 
> I've check-in small modifications in thsice_step_fwd.F
> but have not remove yet the condition "IF ( frzmltMxL(i,j) > 0 )"
> so that results remain unchanged in FWD & AD (tried also FTL, 
> all tests with exp. global_ocean.cs32x15/input_ad.thsice/).
> 
> I would like to remove this condition (I put a CPP option,
> defined @ the top of the file: #define THSICE_OLD_STEP_FWD 
> to make the switch easier) which does not change the FWD
> results but changes both the AD & FTL results.
> 
> Here is what I a got with global_ocean.cs32x15/input_ad.thsice/
> FTL grdcheck summary now:
> grdchk output (g):   1     4.1977355234849E+00  4.1946984504646E+00 -7.2402654353398E-04
> grdchk output (g):   2     4.2504542521783E+00  4.2464803303452E+00 -9.3581543395294E-04
> grdchk output (g):   3     3.6019865030539E+00  3.6028105229743E+00  2.2871586366802E-04
> grdchk output (g):   4     3.1766752581461E+00  3.1766748195733E+00 -1.3806035892294E-07
> grdchk  summary  :  RMS of    4 ratios =  6.0255205740259E-04
> 
> And when I remove the condition (#undef THSICE_OLD_STEP_FWD):
> grdchk output (g):   1     4.1977355234849E+00  5.8302754912409E+00  2.8001077654198E-01
> grdchk output (g):   2     4.2504542521783E+00  6.0587180669255E+00  2.9845650429230E-01
> grdchk output (g):   3     3.6019865030539E+00  5.4493684625302E+00  3.3900845064503E-01
> grdchk output (g):   4     3.1766752581461E+00  4.9806219018770E+00  3.6219305124346E-01
> grdchk  summary  :  RMS of    4 ratios =  3.2155592667210E-01
> 
> If you have a chance to find time to look at this.
> Cheers,
> Jean-Michel
> 
> _______________________________________________
> MITgcm-devel mailing list
> MITgcm-devel at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-devel

---
Patrick Heimbach | heimbach at mit.edu | http://www.mit.edu/~heimbach
MIT | EAPS 54-1420 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach