[MITgcm-devel] question about verification checks

[Jean-Michel Campin]

Hi Ian,

I'am going to cc to devel list, since it's a kind of "natural" question
that might be of interest for others.

1) We run testreport sereral times. Sometime with each test in a different 
checked-out dir, e.g.: tools/example_scripts/csail/test_baudelaire (sepDir=1).
and sometime from the same MITgcm/verification dir., 
e.g. tools/example_scripts/csail/test_dickens (sepDir=0)
but in this case the built & output are gone (except for the last test).
Now, when you normally (=without any of those option: -nc , -q , -ro )
run testreport, there is a "makeclean" and "run_clean" step that
clean up respectively the build dir and the run dir.
I think this also answer your 3rd point.

2) the link from input* dir:
FWD main test (e.g.: lab_sea): link files from input dir.
FWD secondary test (e.g.: lab_sea.hb87): link files from input.hb87 ; then link
files from input dir.
AD main test (e.g.: lab_sea): link files from input_ad dir.
AD secondary test (e.g.: lab_sea.evp): link files from input_ad.evp ; then link
files from input_ad dir.

An other thing is that when a script "prepare_run" exists
in one of the input (FWD) or input_ad (AD) dir, then it is run
after the step described above. This allows to avoid double copy
of binary files (save disk space and downloading time).

Hope it's more clear now.

Cheers,
Jean-Michel

On Wed, Mar 02, 2011 at 08:19:24AM -0800, Fenty, Ian G (3248-Affiliate) wrote:
> Hi JM,
> Hope all is well with you.  I've got a few questions about testreport that I know you can help me with.
> 
> #1  When the verification experiments are checked nightly, is testreport invoked twice, once for the forward simulations and once for the adjoint ones?   I ask because I notice that the -ad flag is required with testreport to invoke taf, but that by using the -ad flag the the check on the regular forward simulation isn't done.
> 
> #2 Binary files: I see that in the lab_sea verification that there are binary files in the input/ directory that are not in input_ad/.  It seems that the binary files from input/ must be kept in run/ after the forward ./testreport so that they don't need to be copied over again from input_ad/ for the adjoint testreport.  Is is right that I only need to include binary files that are needed for the forward and adjoint runs in input/?   
> 
> #3 data.* files: I want some of the data.* files in input_ad/ to differ from those in input/.  However, ./testreport  softlinks the contents of input/ into run/, but if all the contents of input/ are kept in run/ after the forward run, how are the data.* files in input_ad/ brought into run while keeping the binary files from input/?
>  
> I am a bit confused and I'm sorry to say that the documentation doesn't cover these details!
> 
> Thanks,
> Ian
> 
> 
> 

On Wed, Mar 02, 2011 at 08:19:24AM -0800, Fenty, Ian G (3248-Affiliate) wrote:
> Hi JM,
> Hope all is well with you.  I've got a few questions about testreport that I know you can help me with.
> 
> #1  When the verification experiments are checked nightly, is testreport invoked twice, once for the forward simulations and once for the adjoint ones?   I ask because I notice that the -ad flag is required with testreport to invoke taf, but that by using the -ad flag the the check on the regular forward simulation isn't done.
> 
> #2 Binary files: I see that in the lab_sea verification that there are binary files in the input/ directory that are not in input_ad/.  It seems that the binary files from input/ must be kept in run/ after the forward ./testreport so that they don't need to be copied over again from input_ad/ for the adjoint testreport.  Is is right that I only need to include binary files that are needed for the forward and adjoint runs in input/?   
> 
> #3 data.* files: I want some of the data.* files in input_ad/ to differ from those in input/.  However, ./testreport  softlinks the contents of input/ into run/, but if all the contents of input/ are kept in run/ after the forward run, how are the data.* files in input_ad/ brought into run while keeping the binary files from input/?
>  
> I am a bit confused and I'm sorry to say that the documentation doesn't cover these details!
> 
> Thanks,
> Ian
> 
> 
>