[Mitgcm-support] Re: a bug in kpp adjoint

[mitgcm-support at dev.mitgcm.org]

Hi Tony,

thanks for the warning.
I've checked the corresponding routines in ecco_c43_e13.
The problem you mentioned does not appear in this code.
But we'll definetely keep an eye on it.

Cheers
Patrick

Tony Lee wrote:
> 
> A potentially dangerous bug is found in ecco2 kpp adjoint.  It may
> have implication to ecco kpp adjoint too.
> For detail, see
> http://escher.JPL.NASA.GOV:2000/HyperNews/get/forums/assim2k1/31.html
> 
> Please be cautious about argument "mythid" in kpp adjoint subroutines
> (and other adjoint subroutines in general):
> e.g. kpp_calc uses mythid as argument, but adkpp_calc does not (TAF
> think it's passive), yet adkpp_calc calls
> statekpp (for forward recomputation) which uses mythid that's never
> set.
> 
> This is very wrong for thread environment. Even for MPI, it may cause
> probalem depending on whether mythid is actually used (to set array
> bound, etc.).  Furthermore, it precludes one to monitor the timing of
> kpp adjoint
> subroutines because start_timer('*',mythid) requires the value of
> "mythid".
> 
> Tony