[MITgcm-devel] [MITgcm-cvs] MITgcm/pkg/streamice CVS Commit

[Jean-Michel Campin]

Hi Dan,

here is the new OpenAD testreport output (Cost,AD-grad,FWD-grad):
Y Y Y Y 16> 9<16 FAIL  halfpipe_streamice
and the one from last night:
Y Y Y Y 16>16<16 pass  halfpipe_streamice
so the differences are only in AD-grad and are relatively small (9 identical digits).

I will check-in the new output_oadm.txt file.

Cheers,
Jean-Michel

On Sat, Dec 13, 2014 at 06:26:52PM +0000, Daniel Goldberg wrote:
> I did not think that it would, actually. But now that I think about it, it could affect the adjoint gradients. Does the verification check for anything else?
> 
> I do not think I have openad installed on Baudelaire, so if you are able to do it quickly that would actually be easier, but I will check if I have it installed anyway..
> 
> Sent from my iPhone
> 
> > On Dec 13, 2014, at 6:20 PM, Jean-Michel Campin <jmc at ocean.mit.edu> wrote:
> > 
> > Hi Dan,
> > 
> > Looks like this modification changes the OpenAD output for experiment halfpipe_streamice.
> > 
> > Do you plan to update the output_oadm.txt file ?
> > If you prefer, I could do it (since I already run it on baudelaire).
> > 
> > Cheers,
> > Jean-Michel
> > 
> >> On Sat, Dec 13, 2014 at 09:13:11AM -0500, Daniel N. Goldberg, 54-1423, 6172532977,  wrote:
> >> Update of /u/gcmpack/MITgcm/pkg/streamice
> >> In directory forge:/tmp/cvs-serv27618
> >> 
> >> Modified Files:
> >>    adstreamice_cg_solve.F 
> >> Log Message:
> >> remove additional linear solve in adstreamice_cg_solve if ALLOW_OPENAD is defined
> >> 
> >> 
> >> _______________________________________________
> >> MITgcm-cvs mailing list
> >> MITgcm-cvs at mitgcm.org
> >> http://mitgcm.org/mailman/listinfo/mitgcm-cvs