[MITgcm-devel] OpenAD verification test

[Patrick Heimbach]

Hi Jean,

I think it's expected
(although the change ocurred on 11/19 according to CVS).
I assume you are trying OpenAD/code_oad/ ?

In
model/src/timestep.F
some summation orders have changed when using the CD code:
http://mitgcm.org/viewvc/MITgcm/MITgcm/model/src/timestep.F? 
r1=1.47&r2=1.48&sortby=date
Even with using the backward compatibility flags
this results in truncation errors.

-Patrick



On Nov 21, 2009, at 2:54 AM, Jean Utke wrote:

> Hi Jean-Michel,
>
> In my own regression testing I am comparing some outputs directly  
> without
> using the reference output in the repository and it seems that between
> 11/17
> (or 11/18) and now the cost function  output of a simple setup changed
> somewhat
> as you see below:   (old<  vs new >)
> <   local fc =  0.804679274767541D+06
> <  global fc =  0.804679274767541D+06
> ---
>>   local fc =  0.804687744759401D+06
>>  global fc =  0.804687744759401D+06
>
> This running the plain mitgcm code (no AD involved).
> I wanted to verify with you that this is an expected change,,,
> I am asking this because the value of the cost function in this  
> test has
> not
> changed for a long time...
> Please let me know either way so I update my regression tests
>
> Jean
>
>
> Jean-Michel Campin wrote:
>> Hi Jean and Patrick,
>>
>> Yesterday, when I wanted to add this CD_CODE_OPTIONS.h to the OpenAD
>> verification test, I was a little bit confused about where it should
>> go since there are many "code*" dir for this experiment.
>> For instance, there are 5 CPP_OPTIONS.h files arround, but we only
>> do 2 tests (1 with TAF and 1 with OpenAD).
>> Apparently, I picked the right ones this time since last night test
>> went well. But when I did change seawater.F a couple of weeks ago,
>> I missed the local version in code_oad_all.
>>
>> It could be a good idea to:
>> - remove dir & files that are not used (e.g., code_ad_openad ?  
>> code ?)
>> - update the README file in verification/OpenAD
>> - I would be tempted also to clean-up results_ad dir.
>>   (move what we want to keep to results, but I guess some
>>    of the files in results_ad are not up to date ?)
>> What do you think ?
>>
>> Cheers,
>> Jean-Michel
>>
>> _______________________________________________
>> MITgcm-devel mailing list
>> MITgcm-devel at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-devel
>>
>
>
> -- 
> Jean Utke
> Argonne National Lab. / MCS (240-1152)
> 9700 S Cass Ave., Argonne, IL, 60439, USA
> phone +1 630 252 4552 / mobile  +1 630 363 5753
>
>
> _______________________________________________
> 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 54-1518 | 77 Massachusetts Ave | Cambridge MA 02139 USA
FON +1-617-253-5259 | FAX +1-617-253-4464 | SKYPE patrick.heimbach