4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
5 |
============================== |
============================== |
6 |
|
|
7 |
|
o verification/cfc_example: |
8 |
|
- update file "data" with more standard choice (default rotationPeriod, |
9 |
|
exactConserv=T, reduce ivdc_kappa to 10, remove multiDimAdvection=F, |
10 |
|
set deltaTFreeSurf to deltaTtracer) and update the output. |
11 |
|
o pkg/layers: |
12 |
|
- fix bolus part so that it compiles with #undef GM_BOLUS_ADVEC; |
13 |
|
- ensure that "layers_bolus" is switched to False when useGMRedi=False. |
14 |
|
o verification/offline_exf_seaice: |
15 |
|
- set all {field}period to zero for constant-in-time forcing (not tested |
16 |
|
anywhere else). |
17 |
|
o verification/offline_exf_seaice: |
18 |
|
- dyn_lsr : use pkg/seaice S/R (instead of pkg/thsice) to advect sea-ice; |
19 |
|
- dyn_jfnk : use pkg/thsice S/R (instead of pkg/seaice) to advect sea-ice. |
20 |
|
o model/src: |
21 |
|
- implement Finite-Volume method for (hydrostatic) presure gradient |
22 |
|
from S.-J. Lin (QJRMS 1997), for atmosphere using sigma-coordinate. |
23 |
|
o pkg/seaice: implement EVP* by Lemieux et al. (2012). |
24 |
|
- In order to be compatible with the default EVP code, this requires changing |
25 |
|
the EVP code around a little; and because any small change immediately |
26 |
|
breaks the lab_sea.hb87 verification experiment, the reference output is |
27 |
|
updated two times (each time the remaining aggrement was only 2 digits): |
28 |
|
- after replacing 1/dt by 1*recip_dt with a precomputed recip_dt |
29 |
|
- after replacing 1/(1+0.5*dt/tau) by a/(a+1), with a=2*tau/dt |
30 |
|
The adjoint test lab_sea.evp also needed to be updated. |
31 |
|
- Turn on EVP* by setting SEAICE_deltaTevp as before, and setting |
32 |
|
SEAICEuseEVPstar=.true.; for EVP* SEAICE_deltaTevp does not need to be |
33 |
|
a factor of SEAICE_deltaTdyn, and you can set the number of EVP steps |
34 |
|
explicitly with SEAICEnEVPstarSteps (thereby overriding the default of |
35 |
|
INT(SEAICE_deltaTdyn/SEAICE_deltaTevp)). Controlling EVP* is deliberately |
36 |
|
kept obscure, because I need to test it a little more. |
37 |
|
o pkg/cheapaml: |
38 |
|
- fix few incorrect type (_RL/_RS) in subroutine calls. |
39 |
|
o pkg/seaice: |
40 |
|
- new S/R SEAICE_MAP_RS2VEC (similar to SEAICE_MAP2VEC) to map _RS arrays |
41 |
|
to _RL vector; this fixes JFNK code when compiling with "-use_r4" option. |
42 |
|
o model/src: |
43 |
|
- add a warning if using select_rStar=2 with nonlinFreeSurf <> 4; |
44 |
|
change (in input_ad*/data) select_rStar from 2 to 1 for AD verification |
45 |
|
experiments global_ocean.cs32x15 & tutorial_tracer_adjsens (+ update output) |
46 |
|
o pkg/cheapaml: |
47 |
|
- replace "273.1 _d 0" by celsius2K in pkg/cheapaml/cheapaml.F and |
48 |
|
update output of experiment cheapAML_box. |
49 |
|
- store in CHEAPAML.h coeffs ssq[0:2] for saturation specific humidity, |
50 |
|
and set them only in one place (cheapaml_readparms.F). |
51 |
|
- replace "273.16 _d 0" by celsius2K & replace "3.14159 _d 0" by constant "PI" |
52 |
|
- remove re-setting of main-model parameter "celsius2K" (in cheapaml_readparms) |
53 |
|
and re-setting of cheapAML params "lath" & "p0" (in cheapaml_init_varia). |
54 |
|
o model/src/the_main_loop.F: |
55 |
|
- fix revision 129 : need both ALLOW_ECCO and ALLOW_ECCO_EVOLUTION. |
56 |
|
o pkg/seaice: |
57 |
|
- remove legacy branch code. |
58 |
|
- retire SEAICE_GROWTH_LEGACY and SEAICE_CAP_HEFF accordingly. |
59 |
|
o verification/*/code*/SEAICE_OPTIONS.h: |
60 |
|
- remove documentation for retired CPP flags (SEAICE_GROWTH_LEGACY, |
61 |
|
SEAICE_CAP_HEFF). |
62 |
|
o pkg/cheapaml: |
63 |
|
- use standard routine (S/R GET_PERIODIC_INTERVAL) to decide when it is time |
64 |
|
to load new forcing fields. |
65 |
|
- add missing "_d 0" where it's needed; + some editing (improve indentation). |
66 |
|
- remove switch to "-r8" compiler option (no longer needed) in verification |
67 |
|
cheapAML_box, keep same output (small trucation error differences). |
68 |
|
o verification/offline_exf_seaice: |
69 |
|
- add thermodynamics only test. |
70 |
|
o verification/cheapAML_box: |
71 |
|
- split the domain in 8 (2x4) tiles; add MPI and multi-threaded test; |
72 |
|
keep same output (same 14 digits for cg2d). |
73 |
|
- switch on "-r8" compiler option (some numerical values are missing |
74 |
|
the "_d 0") and update output (only 6 digits for cg2d were preserved). |
75 |
|
|
76 |
|
checkpoint64b (2012/12/24) |
77 |
|
o pkg/cheapaml: |
78 |
|
- remove un-used variables (Qnet,EmPmR) from pickup files. |
79 |
|
- store tendency of advected variables into cheapaml pickup file |
80 |
|
and fix the AB-2 restart. |
81 |
|
o verification/offline_exf_seaice and lab_sea adjoint: |
82 |
|
- switch to new code (#undef SEAICE_GROWTH_LEGACY). |
83 |
|
- switch to 4 time steps in offline, consistent with lab_sea. |
84 |
|
- match parameters from SEAICE_GROWTH_LEGACY branch in data.seaice. |
85 |
|
- update results after change of experiment seaice. |
86 |
|
o pkg/seaice: |
87 |
|
- remove option SEAICE_MULTICATEGORY; has been replaced with run-time |
88 |
|
parameter SEAICE_multDim. Issue a warning if SEAICE_MULTICATEGORY is |
89 |
|
still defined since SEAICE_multDim needs to be set to MULTDIM value (from |
90 |
|
SEAICE_SIZE.h) to recover previous #define SEAICE_MULTICATEGORY default. |
91 |
|
o model/src: |
92 |
|
- add run-time parameter to select CG2D solver minimum-residual solution |
93 |
|
(in case of a poor convergence). |
94 |
|
o pkg/seaice/seaice_lsr.F |
95 |
|
- move calculation of matrix coefficients into new |
96 |
|
S/R SEAICE_CALC_LSR_COEFFS; for now this routine remains in the same |
97 |
|
file; the main purpose of this change is to be able to reuse the code |
98 |
|
in S/R SEAICE_PRECONDITIONER |
99 |
|
- modify the computation of the residual so that the mass of sea-ice |
100 |
|
is not used twice; this simplifies the calculations, but since |
101 |
|
the residual is never used except when LSR_mixIniGuess.ge.4---and this is |
102 |
|
not tested in testreport---this does not affect the testreport results. |
103 |
|
o verification/seaice_obcs: |
104 |
|
- disable SEAICE_GROWTH_LEGACY; reset input files to match those of |
105 |
|
verification/lab_sea/input.salt_plume; changes results/output.txt* |
106 |
|
o model/src: |
107 |
|
- adapt quasi-hydrostatic formulation to work with p or p* coordinate. |
108 |
|
o pkg/seaice: |
109 |
|
- improve description of parameters in seaice_fgmres.F and start |
110 |
|
cleaning up the code |
111 |
|
- change name of SCALPROD to SEAICE_SCALPROD and modify meaning of |
112 |
|
second argument |
113 |
|
- add a metric based on grid cell area to SEAICE_SCALPROD; update |
114 |
|
results of offline_exf_seaice.dyn_jfnk |
115 |
|
- compute initial residual norm in seaice_jfnk based on SEAICE_SCALPROD |
116 |
|
o pkg/aim_v23: |
117 |
|
- fix an other bug in dissipation heating (grid-cell area was missing). |
118 |
|
o model/src + pkg/seaice: |
119 |
|
- add global (multi-tile) linear solver for pair of tri-diagonal system |
120 |
|
along X and Y lines, respectively for U and V component. |
121 |
|
Note: 1) MPI and cube-exchange not yet coded. |
122 |
|
2) probably not accurate for poorly conditioned / large size problem. |
123 |
|
- add option for SEAICE_LSR to use global (multi-tile) tri-diagonal solver; |
124 |
|
default: not compiled and not used. |
125 |
|
o pkg/seaice: |
126 |
|
- first attempt at a multi-threaded fgmres solver for the JFNK solver: |
127 |
|
introduce bi/bj loops and use global_sum_tile_rl instead of |
128 |
|
mpi_allreduce in seaice_fgmres.F |
129 |
|
- improve diagnostics output for fgmres solver (seaice_fgmres.F) |
130 |
|
- replace a few unnecessary goto statements by if-blocks (seaice_fgmres.F) |
131 |
|
- add a verification experiment for JFNK in offline_exf_seaice |
132 |
|
o verification: |
133 |
|
- change (for now, only fwd test) offline_exf_seaice experiment: |
134 |
|
use a re-entrant channel with idealized coastline (trapezoidal shape) |
135 |
|
at southern boundary to explore seaice dynamics and thermodynamics |
136 |
|
with simple forcing. Main set-up has dynamics+thermo and SST relaxation; |
137 |
|
secondary test (input.thsice) use thermodynamics only with SST relaxation. |
138 |
|
- add seaice-dynamics only secondary test using LSR and pkg/thsice advection. |
139 |
|
o pkg/exf: |
140 |
|
- disable "sstExtrapol" when Nr=1 |
141 |
|
o verification: |
142 |
|
- add new experiment "cheapAML_box" (simple 1 layer set-up from Bruno Deremble) |
143 |
|
to test pkg/cheapaml. |
144 |
|
o pkg/aim_v23 & pkg/shap_filt: |
145 |
|
- fix dissipation heating (erroneous gravity multiplication factor). |
146 |
|
o pkg/seaice: |
147 |
|
- remove unused field pressPre from jfnk solver |
148 |
|
- rename two subroutines in seaice_preconditioner.F |
149 |
|
o eesupp/src: |
150 |
|
- S/R MDSFINDUNIT avoids reserved unit numbers (scrUnit1, eeDataUnit ...) |
151 |
|
o pkg/seaice/seaice_preconditioner.F: fix preconditioner once again |
152 |
|
- remove wrong parts of the rhs and the lhs (were in there |
153 |
|
because I did not understand the concept of this preconditioner) |
154 |
|
- add an exchange between u- and v-sweep. This increases communication |
155 |
|
by a factor of two but improves the convergence of jfnk. Ideally, |
156 |
|
one would like the u- and v-sweep to be independent. So far within |
157 |
|
undocumented ifdefs |
158 |
|
- reorganize the extra computations for printing the residuals |
159 |
|
o tools/genmake2: |
160 |
|
- comment out the "mpif.h" include in NetCDF test programm (for MPI built) |
161 |
|
since it's not necessary (and was preventing to use MNC in some cases). |
162 |
|
|
163 |
|
checkpoint64a (2012/11/16) |
164 |
o pkg/autodiff: |
o pkg/autodiff: |
165 |
- improve pkg-internal logical parameter switch (inAd/inFwd mode) so that |
- improve pkg-internal logical parameter switch (inAd/inFwd mode) so that |
166 |
autodiff_readparms.F no longer depends on calling order (within |
autodiff_readparms.F no longer depends on calling order (within |