3 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
4 |
============================== |
============================== |
5 |
|
|
6 |
|
checkpoint50 |
7 |
|
- 2nd step of merge with ecco-branch: |
8 |
|
merge changes in ecco-branch back into main trunk |
9 |
|
- bug fix in find_alpha |
10 |
|
- more storing in gmredi_calc_tensor.F, kpp_routines.F |
11 |
|
- avoid iterative loop in ini_pressure for ALLOW_AUTODIFF_TAMC |
12 |
|
- merged some sliced AD support routines |
13 |
|
|
14 |
|
checkpoint49 |
15 |
|
merged from ecco-branch: |
16 |
|
o exf: |
17 |
|
- Enable initialisation of forcing fields to constant |
18 |
|
(runtime) values. |
19 |
|
- in exf_getffields.F |
20 |
|
Reduce i-/j-loop to interior domain, discarding overlaps. |
21 |
|
That also fixes wrong TAF-key computations for key_1, key_2 |
22 |
|
with bulf formulae. |
23 |
|
- exf_init.F modify #ifdef for exf_init_evap |
24 |
|
- exf_getffieldrec.F, ctrl_getrec.F |
25 |
|
The following INT-usages are not safe: |
26 |
|
fldsecs = int(fldsecs/fldperiod)*fldperiod |
27 |
|
fldcount = int(fldsecs/fldperiod) + 1 |
28 |
|
and were modified. |
29 |
|
o autodiff: checkpoint_lev... |
30 |
|
- updated storing for bulk formulae (some arrays were missing) |
31 |
|
- added state for seaice package |
32 |
|
- synchronized with main branch |
33 |
|
o DIVA: |
34 |
|
Modif.s in |
35 |
|
- the_main_loop.F |
36 |
|
- the_model_main.F |
37 |
|
and added cost_final_restore.F cost_final_store.F |
38 |
|
o With these changes, it is possible to fully separate |
39 |
|
the first forward run to compute the cost function (mdthe_main_loop) |
40 |
|
and adjoint runs over individual DIVA intervals (adthe_main_loop) |
41 |
|
into separate model executions. |
42 |
|
This enables better tuning of DIVA to given queue sizes. |
43 |
|
o Loss of cost function value in consecutive DIVA runs is |
44 |
|
fixed by storing to 'costfinal' file. |
45 |
|
o Appropriate call of ctrl_unpack/_pack,grdchk_main at |
46 |
|
beginning/end of adjoint integration is accounted for. |
47 |
|
The current changes apply to TAF-1.4.26. |
48 |
|
TAF has one remaining bug, requiring one manual change |
49 |
|
(see MITgcm manual). |
50 |
|
o cal: synchronized between MAIN and ecco-branch |
51 |
|
|
52 |
|
checkpoint48i_post |
53 |
|
o Changes to restore differentiability of code w.r.t. previous tag |
54 |
|
(mostly adding new routines to make list and replacing |
55 |
|
pressure by totPhiHyd). |
56 |
|
o Updated and merged grdchk package |
57 |
|
- has both ADM and TLM checks |
58 |
|
- works for single- and multi-proc. |
59 |
|
- output cleaned |
60 |
|
- worked successfully for parallel DIVA |
61 |
|
o genmake: |
62 |
|
- Adding -platform=SP4 |
63 |
|
- Modif. for DEC Alpha (many don't have makedepend; use mkdep instead). |
64 |
|
o Modif. for DIVA: |
65 |
|
ctrl_unpack should not be called in intermediate DIVA interval. |
66 |
|
(similar thing would apply for active files such as |
67 |
|
tbar, sbar, psbar, which should not be reinitialised; |
68 |
|
only relevant for ecco-branch) |
69 |
|
|
70 |
|
checkpoint48h_post |
71 |
|
o improve "ini_p_ground" when using finite difference form for calc_phi_hyd |
72 |
|
o Added net flux to downward flux conversion to pkg/exf/exf_getffields.F |
73 |
|
o Added SEAICE_initialHEFF to pkg/seaice |
74 |
|
o ecco_the_model_main.F: SEAICE_WRITE_PICKUP must precede WRITE_CHECKPOINT |
75 |
|
|
76 |
checkpoint48g_post |
checkpoint48g_post |
77 |
--- EOS: |
--- EOS: |
78 |
o compute locally the pressure for use in EOS: UNESCO, JMD95P, MDJWF |
o compute locally the pressure for use in EOS: UNESCO, JMD95P, MDJWF |