[MITgcm-devel] tutorial_global_oce_biogeo OpenAD results

[Jean-Michel Campin]

Hi Oliver,

I only run the OpenAD test on baudelaire once a week, but it seems
that the Fwd Grad has changed a little:

I updated the reference output file (results/output_oadm.txt)
on Nov 5th (generated on baudelaire with gfortran) and checked
on Nov 6 (after the cost_final set of check-in) that the output has not 
changed (16 digits for cost Ad-Grad & Fwd-Grad).

And on Sun (Nov 10), the new results on baudelaire:
http://mitgcm.org/testing/results/2013_11/tr_baudelaire_20131110_10/summary.txt
show only 7 digits for the Fwd-Grad:
Y Y Y Y 16>16< 7 pass  tutorial_global_oce_biogeo

Can the changes you did on tracer_meanarea.F could explain
these little differences in the Fwd-Grad ?
If it's the case, Seems curious to me that it did not affect 
the TAF output.

Also it would be a good thing to know why the gfortran OpenAD 
test on acesgrid that you run everyday have a different Fwd-Grad
(for this experiment) than the one I run on baudelaire
using Patrick's OpenAD setting.
Are we using the same version of OpenAd on both side ?

Cheers,
Jean-Michel


On Fri, Nov 08, 2013 at 03:03:12PM -0500, Oliver Jahn wrote:
> Update of /u/gcmpack/MITgcm/pkg/dic
> In directory forge:/tmp/cvs-serv17881/pkg/dic
> 
> Modified Files:
> 	tracer_meanarea.F 
> Log Message:
> remove kLev argument and use global_sum_tile_rl in tracer_meanarea
> 
> 
> _______________________________________________
> MITgcm-cvs mailing list
> MITgcm-cvs at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-cvs