4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
5 |
============================== |
============================== |
6 |
|
|
7 |
|
o exch2: |
8 |
|
- change index-bounds storage (moved from target to local tile, |
9 |
|
more intuitive this way) ; rename/remove some variables. |
10 |
|
- change utils/exch2/matlab-topology-generator accordingly. |
11 |
|
|
12 |
|
checkpoint61a |
13 |
|
o exch2/matlab-topology-generator: |
14 |
|
fix bug in index bounds for N-S internal exchange (hard to tell if this |
15 |
|
had any effect on the solution). |
16 |
|
o pkg/mnc: small fix in mnc_handle_err: make sure that length of msgbuf |
17 |
|
is exceeded |
18 |
|
o testreport: modify syntax of awk, so that it works also with |
19 |
|
awk, and not only with nawk and gawk |
20 |
|
|
21 |
|
checkpoint61 |
22 |
|
o autodiff |
23 |
|
Change default for TAF usage by removing argument '-nonew_arg' |
24 |
|
This changes TAF subroutine argument list (ordering of active var.) |
25 |
|
Omission of this flag leads to incompatibility w.r.t. TAMC |
26 |
|
To restore TAMC-compatibility, need following modifs: |
27 |
|
* use adoptfile tools/adjoint_options/adjoint_tamc_compatibility |
28 |
|
* use CPP option #define AUTODIFF_TAMC_COMPATIBILITY |
29 |
|
Tested TAF version is 1.9.22 |
30 |
|
N.B.: exch2 hand-written adjoint code currently not TAMC compatible |
31 |
|
|
32 |
|
checkpoint60 |
33 |
|
--->>> This is the last checkpoint for which generating the adjoint |
34 |
|
--->>> is done via the TAMC-compatibility TAF-flag "-nonewarg" |
35 |
|
--->>> by default. |
36 |
|
--->>> Future checkpoints will ommit this flag by default. |
37 |
|
o pkg/generic_advdiff, os7mp: simplify and avoid division by zero. |
38 |
|
o pkg/grdchk: Add gradient check to etan0 perturbations. |
39 |
|
o pkg/exf: read ice area from file (#ifdef ALLOW_ICE_AREAMASK). |
40 |
|
o pkg/autodiff: more flexibility in zeroadj. |
41 |
|
model/src: modifying do_oceanic_phys.F accordingly. |
42 |
|
o pkg/ctrl: Bounds for (kapgm,kapredi,eddypsi) parameter estimation. |
43 |
|
o model/src and pkg/gmredi: GMREDI_WITH_STABLE_ADJOINT CPP option |
44 |
|
special setup of gmredi for which the adjoint is stable |
45 |
|
enough for parameter optimization. |
46 |
|
o bridging the gap between eddy stress and GM. |
47 |
|
-> eddyTau is replaced with eddyPsi (eddyTau = f x rho0 x eddyPsi) |
48 |
|
along with a change in CPP option (now ALLOW_EDDYPSI). |
49 |
|
-> when using GM w/ GM_AdvForm: |
50 |
|
The total eddy streamfunction (Psi = eddyPsi + K x Slope) |
51 |
|
is applied either in the tracer Eq. or in momentum Eq. |
52 |
|
depending on data.gmredi (intro. GM_InMomAsStress). |
53 |
|
-> ALLOW_EDDYPSI_CONTROL for estimation purpose. |
54 |
|
The key modifications are in model/src/taueddy_external_forcing.F |
55 |
|
pkg/gmredi/gmredi_calc_*F pkg/gmredi/gmredi_*transport.F |
56 |
|
|
57 |
checkpoint59r |
checkpoint59r |
58 |
o new S/R (pkg/mdsio/mdsio_facef_read.F) reading of sNx+1 x sNy+1 array |
o new S/R (pkg/mdsio/mdsio_facef_read.F) reading of sNx+1 x sNy+1 array |
59 |
from per-face file (take S/R READSYMTILE_RS out of ini_curvilinear_grid.F); |
from per-face file (take S/R READSYMTILE_RS out of ini_curvilinear_grid.F); |