[Mitgcm-support] RE: -D now

mitgcm-support at dev.mitgcm.org mitgcm-support at dev.mitgcm.org
Wed Jul 9 15:54:58 EDT 2003


Hi Chris,

it's Lars, I just visited Patrick. He suggests the following.
Let's branch it from c44d_post, i.e. it has all my
adj. related changes, and is just before the start
of the ptracer changes, shap. filt. changes etc.

So I'll branch from c44d_post and include the GM-related changes
to release_1 final.

I'll also check in the changes into the Head so that
they're up to date AD-wise.

Alistair, my suspicion is that the FAILED for global_ocean
is related to the change of the call of packages_init_variables
(the new verif/glo... uses ptracer and this is also initialised
from packages_init_variables).

Lars

Quoting Chris Hill <cnh at plume.mit.edu>:

> 
> Hi Patrick,
> 
>  Great, although its not quite I had told
> Lars we were planning to do. I'm not sure what this means 
> for the "release1_final" branch? Does this mean it will need to 
> be branched as the latest code, or can it still be
> branched at c44b_post + adj correct and gm correct (the setup
> on your blackboard).
> What do you think is best? I think Lars et. al.
> would prefer it to be based on c44b_post
> since this is likely to be more stable
> than HEAD. Does that make sense?
> 
> Chris
> 
> -----Original Message-----
> From: heimbach at MIT.EDU [mailto:heimbach at MIT.EDU]
> Sent: Saturday, March 23, 2002 9:54 AM
> To: support at mitgcm.org
> Subject: -D now
> 
> 
> 
> So I've fixed the adjoint for the latest in the main branch.
> There were a few minor bugs (missing stuff),
> an intermediate problem (new storing in gad_advection),
> and one serious problem:
> in initialise_varia, the call of packages_init_variables
> had been shifted. This affects the control flow, since
> ctrl_init, ctrl_map_... are called from there.
> packages_init_variables has to be called after 
> convective_adjustment_ini.
> (I'll have to look exactly what's going on,
> but this problem had already appeared in the past).
> Not tested for the adjoint are the ptracer package
> and the nonlin. free surface.
> 
> With my changes I ran the testscript and got the following:
> Y Y Y Y 16 16 16 16  0 16 16 16 16 16 16 16 16 16 22  0  0 pass 
> adjustment.128x64x1
> Y Y Y Y 16 16 16 16  0 16 16 16 16 16 16  0  0 16 16  0  0 pass 
> adjustment.cs-32x32x1
> Y Y Y Y 16 16 16 16  0 16 16 16 16 16 16 22  0 16 16 22  0 pass 
> adjust_nlfs.cs-32x32x1
> Y Y Y Y 16 16 16 16 13 16 16 16 16 16 16 16 16 16 16 16 16 pass 
> advect_cs
> Y Y Y Y 16 22 16 16 16 16 16 16 13 16 16 16 16 16 16 16 16 pass 
> advect_xy
> Y Y Y Y 16  8  9  8  8 16 16 16 16 16 16 16 16 16 16 16 16 pass 
> advect_xz
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O  
> aim.5l_cs
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O  
> aim.5l_Equatorial_Channel
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O  
> aim.5l_LatLon
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 12 13 16 16 13 16 pass  exp0
> Y Y Y Y 16 16 16 16 16 16 16 16 22 16 16 16 16 16 16 22 16 pass  exp1
> Y Y Y Y 13 13 16 16 16 16 16 16 16 13 13 16 16 13 13 12 13 pass  exp2
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O   exp4
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O   exp5
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 pass 
> front_relax
> Y Y Y Y 11 16 13 13 13 12 16 16 13 10 12 11 12 10 10 11 11 FAIL 
> global_ocean.90x40x15
> Y Y Y Y 14 16 16 16 16 16 16 16 16 11 13 22 13 16 16  9 16 pass 
> hs94.128x64x5
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 22 16 pass 
> hs94.1x64x5
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 22 16 16 16 22 16 pass 
> hs94.cs-32x32x5
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O  
> internal_wave
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 pass 
> natl_box
> Y Y N N -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- N/O  
> plume_on_slope
> Y Y Y Y 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 16 pass 
> solid-body.cs-32x32x1              
> 
> global_ocean failed, but it looks more like an accuracy problem.
> 
> P.
> 
> -- 
> _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/
> 
> Patrick Heimbach       Massachusetts Institute of Technology
> FON: +1/617/253-5259                      EAPS, Room 54-1518
> FAX: +1/617/253-4464                 77 Massachusetts Avenue
> mailto:heimbach at mit.edu                   Cambridge MA 02139
> http://www.mit.edu/~heimbach/                         U.S.A.
> 
> 




-------------------------------------------------------------------------
This e-mail message was sent using MIT's WebMail System, a service of
Information Systems. For more information, see http://web.mit.edu/webmail



More information about the MITgcm-support mailing list