4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
5 |
============================== |
============================== |
6 |
|
|
7 |
|
o testreport (AD-tests): |
8 |
|
no longer link files from dir "../input" (1rst link files from |
9 |
|
"../input_ad.[xx]" if secondary test "xx" and then from "../input_ad") |
10 |
|
|
11 |
|
checkpoint62f (2010/04/29) |
12 |
|
o model/src/the_main_loop.F: |
13 |
|
adjust store directives for seaice evp and multicategory according to |
14 |
|
pkg/ecco/the_main_loop.F |
15 |
|
o adjoint: |
16 |
|
global_sum_ad was incomplete (noted many times by many people ;) |
17 |
|
nice description by Jean U. as follows: |
18 |
|
* allreduce(var) in global_sum.F can be interpreted as |
19 |
|
reduce(sum) followed by broadcast of sum. |
20 |
|
* The adjoint of that is |
21 |
|
sum(adjoint) followed by broadcast(summed adjoint value) |
22 |
|
or in other words, allreduce(sum, adjoint vars) |
23 |
|
We were missing the sum(adjoint)! |
24 |
|
Benign, except(!) for "global" scalars appearing in tile loops (bi,bj) |
25 |
|
(noted by J.M. Campin, e.g., in cost_ssh_mean, variable offset). |
26 |
|
The fix should produce 16-digit accuracy between |
27 |
|
single-CPU multi-tile (e.g. nSx=2, nPx=1) and |
28 |
|
multi-CPU single-tile (e.g. nSx=1, nPx-2) |
29 |
|
We'll probably change this yet again shortly |
30 |
|
o eesupp/src: |
31 |
|
- fix global_sum_r4 (wrong from previous modif in Jun 2009; ever used ?) |
32 |
|
o pkg/icefront |
33 |
|
- ALLOW_SUBGLACIAL_RUNOFF option adds capability for time-evolving |
34 |
|
specification of addmass array based on a 2D field, which is added |
35 |
|
at bottommost wet level at each horizontal location. |
36 |
|
o model/src: |
37 |
|
horizontal grid initialisation (other than curvilinear): add code for Exch2 IO |
38 |
|
layout; (not always compatible with delX,delY setting; commented out for now). |
39 |
|
o pkg/exch2: |
40 |
|
change tile to processor setting (case nSy*nPy > 1): allows default model mapping |
41 |
|
(myX and myYGlobalLow) to be used for simple (1 facet, no blank tile) config. |
42 |
|
|
43 |
|
checkpoint62e (2010/04/15) |
44 |
|
o pkg/obcs: |
45 |
|
- fix wrong index in OBCS_SPONGE_V for Eastern-OB (reported by David Hebert). |
46 |
|
o pkg/dic: |
47 |
|
-fix bug in fe_chem.F for multi-tile (nSx*nSy > 1) set-up; |
48 |
|
(note: not presently tested in any forward verification exp.) |
49 |
|
update tutorial_global_oce_biogeo AD output file. |
50 |
|
o pkg/exf: |
51 |
|
- account for the variation of albedo as a function of zenith angle |
52 |
|
by defining ALLOW_ZENITHANGLE in EXF_OPTIONS.h |
53 |
|
o build_options: |
54 |
|
- rename "linux_ia64_ifort+mpi_ice_nas" to "linux_amd64_ifort+mpi_ice_nas" |
55 |
|
o pkg/dic: |
56 |
|
change DIC atmos-CO2 box (method 3, dic_int1=3) initialisation & restart: |
57 |
|
- use binary pickup file for restart; |
58 |
|
- initialise atmos-CO2 box from dic_pCO2 parameter; |
59 |
|
- fix multi-threaded. |
60 |
|
o pkg/rbcs: |
61 |
|
- clean-up RBCS code: add RBCS_OPTIONS.h file ; fix multi-threaded ; |
62 |
|
change (+ fix IO calls) type of array RBC_mask from _RL to _RS. |
63 |
|
- add RBCS to exp4 passive tracer 01 (rbcs was previously untested) |
64 |
|
and update output.txt |
65 |
|
o pkg/kpp: |
66 |
|
- add diagnostics for KPP non-local flux of Temp, Salinity and pTracers; |
67 |
|
- change diagnostic of KPP non-local term: output average of KPPghat*KPPdiffKzS |
68 |
|
(= ratio of KPP non-local (salt) flux relative to surface-flux) when |
69 |
|
using pkg/timeave & pkg/diagnostics (diag 'KPPghatK' replaces 'KPPghat '); |
70 |
|
- change pkg/offline to read-in average KPPghat*KPPdiffKzS from file |
71 |
|
KPP_ghatKFile (instead of the average of KPPghat from file KPP_ghatFile). |
72 |
|
o pkg/offline: |
73 |
|
fix bug in S/R OFFLINE_FIELDS_LOAD regarding surfaceForcing conversion factor |
74 |
|
(bug was added in version 1.10 on 2007-08-23). |
75 |
|
o pkg/offline: |
76 |
|
- no need to specify periodicExternalForcing, etc ... to use offline pack |
77 |
|
- offline input files no longer just single prec --> parameter offlineLoadPrec |
78 |
|
o verification/lab_sea: |
79 |
|
change AD test to multi-tile setup and update all 4 AD output. |
80 |
|
NOTE: while cost function values remain very similar, |
81 |
|
gradients are very different, although gradient checks are self-consistent |
82 |
|
Should try to get to the bottom of this (cost function summation order?) |
83 |
|
o pkg/dic: |
84 |
|
in carbon_chem.F, replace log10 (AUTODIFF) with explicit form to avoid |
85 |
|
a weakness of TAF that drops part of double precision in log10 derivative. |
86 |
|
<= changes removed since problem has been fixed in current TAF version 1.9.71 |
87 |
|
update AD output of exp. tutorial_dic_adjoffline & tutorial_global_oce_biogeo. |
88 |
o pkg/ecco: |
o pkg/ecco: |
89 |
add missing "_d 0" in cost_readers.F, cost_readtopex.F & cost_readtopexmean.F |
add missing "_d 0" in cost_readers.F, cost_readtopex.F & cost_readtopexmean.F |
90 |
update all 4 lab_sea AD output. |
update all 4 lab_sea AD output. |