Parent Directory | Revision Log | Revision Graph
Links to HEAD: | (view) (download) (annotate) |
Links to checkpoint48g_post: | (view) (download) (annotate) |
Sticky Tag: |
Yet more changes: o adgmredi_calc_tensor avoiding all recomputation of gmredi_slope_limit o adgmredi_x/y/rtransport added flag for excessive storing to avoid recomp. of u/v/rtans, dTdx/y/z -> this is not really necessary and very memory-consuming o adgmredi_slope_psi: consistency with gmredi_slope_limit in treatment of GM_slopeSqCutoff o gmredi_slope_limit re-activated full calculation of taperfct for case 'ac02'
fix few bugs and restore parameter value (e.g., Small_Number=1.D-12) and scheme (e.g., Large_SlopeSqr=1.D+48) of checkpoint47f_post
Here they are, before they get changed/lost/stolen. Mostly modif.'s to fix numerical sensitivities. Gradient checks OK for - GM_taper_scheme: * clipping * ac02 * linear * glw91 * dm95 * ldd97 - GMREDI_OPTIONS: * GM_VISBECK_VARIABLE_K * GM_NON_UNITY_DIAGONAL * GM_EXTRA_DIAGONAL * GM_BOLUS_ADVEC in conjunction with data.gmredi parameters to be checked in in a few minutes under verification/carbon/code/
Checing in one sort-of working adjoint GMRedi setup before I go ice climbing.
o * "clean" adjoint code (in terms of extensive recomputations) can now be obtained for all GMREDI options (i.e. for - GM_VISBECK_VARIABLE_K - GM_NON_UNITY_DIAGONAL - GM_EXTRA_DIAGONAL - GM_BOLUS_ADVEC ) * However, wrong gradient check problem remains unsolved. * New CPP options have been introduced for different tapering schemes
o cleaned up the use of rhoNil and rhoConst. - rhoNil should only appear in the LINEAR equation of state, everywhere else rhoNil is replaced by rhoConst, e.g. find_rho computes rho-rhoConst and the dynamical equations are all divided by rhoConst o introduced new parameter rhoConstFresh, a reference density of fresh water, to remove the fresh water flux's dependence on rhoNil. The default value is 999.8 kg/m^3 o cleanup up external_forcing.F and external_forcing_surf.F - can now be used by both OCEANIC and OCEANICP
Various modifications to generate more stable adjoint of GMRedi.
Modification to the GMREDI package : change units of tensor-K arrays, scale now like diffusivity initialise all common block arrays in S/R gmredi_init add option to use different isopycnal(Redi) & GM diffusivity add option to use the advective GM form or the skew-flux form (=default) bug in non_unity_diagonal part fixed.
Modifications related to GM-Redi tapering options: 1) avoid changing the slope sign in negatively stratified region 2) tapered GM-Redi tensor was not truly adiabatic: fixed. 3) add one (simple) tapering function.
Minor changes for adjoint.
CPP-Option "ALLOW_TIMEAVE" replaces "INCLUDE_DIAGNOSTICS_INTERFACE_CODE"
Made sure each .F and .h file had the CVS keywords Header and Name at its start. Most had header but very few currently have Name, so lots of changes!
Merged changes from branch "branch-atmos-merge" into MAIN (checkpoint34) - substantial modifications to algorithm sequence (dynamics.F) - packaged OBCS, Shapiro filter, Zonal filter, Atmospheric Physics
Corrected array dimensions. Temporary scalars introduced to avoid storing by TAMC.
Modifications to help TAMC's dependency analysis.
Packaged GM/Redi routines.
This form allows you to request diffs between any two revisions of this file. For each of the two "sides" of the diff, select a symbolic revision name using the selection box, or choose 'Use Text Field' and enter a numeric revision.
ViewVC Help | |
Powered by ViewVC 1.1.22 |