4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
5 |
============================== |
============================== |
6 |
|
|
7 |
o pkg/seaice : |
o model/src: |
8 |
|
- start to implement clean wait to stop in initialization S/R (config_check.F, |
9 |
|
packages_check.F and pkg/obcs/obcs_check.F + pkg/mypackage) |
10 |
|
o folowing PRINT_MESSAGE changes: |
11 |
|
- remove reminder multiple print to the same file (STDOUT) when multi-threads. |
12 |
|
o pkg/bbl: |
13 |
|
- modification so it can work with z* |
14 |
|
- replace deltaT with dTtracerLev(k) for asynchronous time stepping |
15 |
|
- global_with_exf/input.yearly verification; changes output.yearly.txt |
16 |
|
o eesupp/src: |
17 |
|
- PRINT_MESSAGE: use OpenMP critical instruction (similar to PRINT_ERROR); |
18 |
|
writing to unit 0 (if unit=errorMessageUnit) extended to the case myThid=1 |
19 |
|
- CHECK_THREADS: specific code if using OpenMP threading: improve checking of |
20 |
|
# of threads across processors (e.g., different env OMP_NUM_THREADS setting) |
21 |
|
- improve special preliminary call to EESET_PARMS (if eedata is read 2 times). |
22 |
|
- when using MPI, always set usingMPI=T ; set default accordingly and stop |
23 |
|
if not consistent. |
24 |
|
- ALL_PROC_DIE: add a BARRIER after MPI_FINALIZE call ; skip this when thread |
25 |
|
argument is zero (in case called within single-thread section). |
26 |
|
o pkg/fizhi: |
27 |
|
clean-up turbulence cold-start switch: decided in fizhi_init_vars.F, stored |
28 |
|
in common bloc (fizhi_coms.h) and then passed as argument up to S/R TURBIO. |
29 |
|
o pkg/exch2: |
30 |
|
- add argument "updateCorners" to S/R exch2_get_uv_bounds (enable to |
31 |
|
activate EXCH_IGNORE_CORNERS in vector EXCH S/R). |
32 |
|
- switch 1rst EXCH call to IGNORE_CORNERS (instead of UPDATE_CORNERS) |
33 |
|
but keep the 2nd with UPDATE_CORNERS: this prevents overwriting good data |
34 |
|
with bad ones coming from a not-yet-updated halo. |
35 |
|
This should fix the problem when sNx <> sNy and GCD(sNx,sNy) < 2*Olx. |
36 |
|
o pkg/frazil: |
37 |
|
move FrzForcT diag from frazil_diagnostics_state.F to frazil_calc_rhs.F |
38 |
|
http://mitgcm.org/pipermail/mitgcm-devel/2012-March/005384.html |
39 |
|
o model/src/config_check.F: add a check for OLx>sNx/OLy>sNy |
40 |
|
o verification/lab_sea |
41 |
|
* #undef SEAICE_EXCLUDE_FOR_EXACT_AD_TESTING |
42 |
|
* #define ALLOW_SST_CONTROL & ALLOW_SST_CONTROL |
43 |
|
* nTimeSteps 3 -> 4 |
44 |
|
* changed reference output accordingly |
45 |
|
o pkg/ctrl: |
46 |
|
* fix ctrl_get_gen for xx_sst, xx_sss |
47 |
|
o pkg/fizhi: |
48 |
|
- fix 2 calls to S/R GRD2MSC (argument igrd) in update_earth_exports.F |
49 |
|
update results of exp. fizhi-cs-32x32x40 (fail with only 6 digits for cg2d) |
50 |
|
- by default, use new version of S/R GETPWHERE (#define TRY_NEW_GETPWHERE); |
51 |
|
fix few out-of-bounds indices in referencing vars (initialise igrd to 1); |
52 |
|
clean-up fizhi_turb.F (remove unused & uninitialised qdiaglsm array). |
53 |
|
o pkg/gmredi: |
54 |
|
- add diagnostic of Redi (main diagnonal) vertical diffusive flux of Temp |
55 |
|
o pkg/fizhi: |
56 |
|
- fix few initialisation and multi-tile Pb: aqualev & gridalt now pass the |
57 |
|
(old)aces open64 tuv test. |
58 |
|
o pkg/icefront: fix bug and reorder loop in icefront_init_varia.F |
59 |
|
o pkg/mom_vecinv: |
60 |
|
- fix 1 loop range in mom_vecinv.F to be like all the other loops which |
61 |
|
set/update gu,gv/Diss (fix Pb of uninitialised fVerUkp in admom_vecinv |
62 |
|
with non-lin free-surf). |
63 |
|
o model/src: move k-loop outside of i/j-loops when (re-)computing R_low |
64 |
|
in ini/update_masks_etc.F in order to avoid -O3 optimization problems with |
65 |
|
some compilers (gfortran 4.6.0 on MacOS) |
66 |
|
o momentum: |
67 |
|
- separate fVer?(:,:,kUp) & fVer?(:,:,kDown) in argument list of MOM_FLUXFORM |
68 |
|
& MOM_VECINV subroutines (to help TAF). This fixes Pb of uninitialised |
69 |
|
fVer?(kUp) in admom_fluxform with non-lin free-surf or bottom control. |
70 |
|
|
71 |
|
checkpoint63k (2012/03/17) |
72 |
|
o model/src: |
73 |
|
- selectAddFluid (ALLOW_ADDFLUID): add missing term in horizontal momentum |
74 |
|
flux-form (mom_fluxform.F) and vertical momentum (calc_gw.F) equations. |
75 |
|
- SOLVE_DIAGONAL: switch default to use adjoinable & vectorized version |
76 |
|
(#undef SOLVE_DIAGONAL_LOWMEMORY); update output of exp. front_relax.bvp. |
77 |
|
o model/src: solve_*diagonal.F |
78 |
|
- replace division with multiplication by inverse (similar to original version) |
79 |
|
update output of AD exp. global_ocean.90x40x15. |
80 |
|
- fix pkg/gmredi/gmredi_calc_psi_bvp.F & pkg/ggl90/ggl90_calc.F |
81 |
|
to enable the use of the vectorized & differentiable solve_tridiagonal.F |
82 |
|
- test exp. global_ocean.90x40x15/input_ad: switch to advection scheme 20 for |
83 |
|
vertical advection (to get solve_tridiagonal tested in AD mode). |
84 |
|
o pkg/seaice: ECCO-CLIVAR merge: |
85 |
|
1. seaice AREA relaxation |
86 |
|
* CPP option: SEAICE_ALLOW_AREA_RELAXATION |
87 |
|
* runtime flag: SEAICE_tauAreaObsRelax |
88 |
|
2. modified global mean cost imbalance code: CPP option |
89 |
|
ALLOW_SSH_GLOBMEAN_COST_CONTRIBUTION |
90 |
|
o another iteration on solve_*diagonal.F |
91 |
|
Make choices independent of AUTODIFF |
92 |
|
Default will be the adjoinable vectorized version. |
93 |
|
Alternative choices are: |
94 |
|
* non-adjoinable original version with low memory footprint: |
95 |
|
#define SOLVE_DIAGONAL_LOWMEMORY |
96 |
|
* still adjoinable but does not vectorize (k as inner-most loop): |
97 |
|
#define SOLVE_DIAGONAL_KINNER (keeping SOLVE_DIAGONAL_LOWMEMORY undef) |
98 |
|
o testreport |
99 |
|
- add option to use a home made wrapper around "make" |
100 |
|
- remove MITGCM_IEEE env. variable (no longer active) |
101 |
|
o pkg/shelfice |
102 |
|
- improve computation of friction velocity for velocity dependent |
103 |
|
exchange coefficients: include velocities into SHELFICEBoundaryLayer |
104 |
|
block to average velocities vertically |
105 |
|
- add TAF store directives for SHI_ALLOW_GAMMAFRICT |
106 |
|
o Modify diagonal solvers, now three versions (upon request): |
107 |
|
#ifndef SOLVE_DIAGONAL_AUTODIFF |
108 |
|
1. default version (dating back a while ago) |
109 |
|
#else /* def SOLVE_DIAGONAL_AUTODIFF */ |
110 |
|
# ifdef SOLVE_DIAGONAL_KINNER |
111 |
|
2. adjoinable version with k-innermost loop (G. Forget) |
112 |
|
# else |
113 |
|
3. adjoinable vectorized version of 2. |
114 |
|
# endif |
115 |
|
#endif /* SOLVE_DIAGONAL_AUTODIFF */ |
116 |
|
All flags are currently set within routines, but could/should be moved to CPP-options. |
117 |
|
Array bounds for 1. are as before, for 2.,3. cover the entire domain |
118 |
|
(could be changed for 1. too). |
119 |
|
o verification/lab_sea (fwd): |
120 |
|
- update the 4 fwd lab_sea test experiment: no longer SEAICE_GROWTH_LEGACY, |
121 |
|
use 7 categories (except hb87), new solve4temp params (except salt_plume), |
122 |
|
constant salinity (=4) or variable salinity (salt_plume) done through |
123 |
|
SItracer. |
124 |
|
o pkg/seaice: |
125 |
|
- rename parameters: SIsalFrac to SEAICE_saltFrac & SIsal0 to SEAICE_salt0; |
126 |
|
- remove old parameters ( SEAICE_freeze, SEAICEturbFluxFormula, SEAICE_gamma_t, |
127 |
|
SEAICE_gamma_t_frz, SEAICE_availHeatTaper & SEAICE_availHeatFracFrz) |
128 |
|
from SEAICE_PARMS.h ; still need to fix seaice_cost_sst.F (broken when |
129 |
|
SEAICE_freeze got retired) but already better than previously with |
130 |
|
SEAICE_freeze=123456.7 deg.C; |
131 |
|
- move setting of facOpenGrow/Melt from SEAICE_INIT_FIXED to SEAICE_READPARMS |
132 |
|
(safer multi-threaded setting); always set SEAICEuseEVP; |
133 |
|
- setting of ocean-ice turb. flux coeff: moved from SEAICE_CHECK to |
134 |
|
SEAICE_READPARMS & SEAICE_INIT_FIXED (this fixes wrong summary report); |
135 |
|
stop if multiple specifications for the same coeff; make sure default |
136 |
|
SEAICE_mcPheePiston is compatible with drF(1) & deltaT. |
137 |
|
o pkg/obcs: |
138 |
|
- allow to switch to upwind 1rst order advection scheme for the advective |
139 |
|
flux computation at the open-boundary; disabled for adjoint built. |
140 |
|
o pkg/ptracers: |
141 |
|
- rename PTRACERS_RESTART.h -> PTRACERS_START.h |
142 |
|
- rename ptracers_fields_blocking_exchanges.F -> ptracers_fields_blocking_exch.F |
143 |
|
- add parameter to decide when to start a tracer; use internal flag to |
144 |
|
know which tracer to step-forward; switch this internal flag at the top |
145 |
|
of forward_step.F (new S/R PTRACERS_SWITCH_ONOFF); for now, not enabled |
146 |
|
for adjoint built. |
147 |
|
o pkg/seaice: |
148 |
|
- add SEAICE_SIZE.h anywhere in the code where SEAICE.h is included ; |
149 |
|
- move MULTDIM declaration and setting from SEAICE.h to SEAICE_SIZE.h ; |
150 |
|
- keep experiment offline_exf_seaice truly single category (set MULTDIM=1). |
151 |
|
o pkg/ecco: |
152 |
|
- fix writing of output file DiagnosCost_ClimSalt/Theta |
153 |
|
and DiagnosCost_DriftSalt/Theta (if multi-tiles per proc). |
154 |
|
o pkg/seaice: |
155 |
|
- add missing initialisation of COR_ICE array in dynsolver.F |
156 |
|
- merging SEAICE_MULTICATEGORY define and undef cases. Now the same code is |
157 |
|
compiled either way, and is tested in all exps (multidim was only tested |
158 |
|
in lab_sea ad before). |
159 |
|
- cosmetic change to prepare for future addition of ITD. no change to results. |
160 |
|
no defaults change. |
161 |
|
- added run time param SEAICE_multDim (7 for SEAICE_MULTICATEGORY, 1 otherwise). |
162 |
|
- by lack of adequate verification experiments, I further used custom versions |
163 |
|
of global_ocean.cs32x15 to test multicat or not, with pickup or not. All |
164 |
|
should be correct. |
165 |
|
- avoid recomputations in ad. |
166 |
|
o array initialisation in adjoint mode: |
167 |
|
1) pkg/mom_common/mom_calc_ke.F: |
168 |
|
add full array initialisation of KE inside mom_calc_ke.F (ALLOW_AUTODIFF_TAMC): |
169 |
|
The S/R which calls MOM_CALC_KE initialises KE but in some recomputation |
170 |
|
part this initialisation is gone, resulting in Floating Point Exception |
171 |
|
(caught by open64 compiler with debug option). |
172 |
|
2) model/src/convective_weights.F |
173 |
|
set output convectCount over the full array range (instead of doing |
174 |
|
an update): this fix a similar issue when called within recomputation (adjoint) |
175 |
|
which could by-pass the initialisation (normally done in the caller S/R). |
176 |
|
o pkg/seaice: |
177 |
- final version of turbulent flux params, as agreed at MIT ice meet: |
- final version of turbulent flux params, as agreed at MIT ice meet: |
178 |
SEAICE_mcPheePiston (default is ~ 0.000875) |
SEAICE_mcPheePiston (default is ~ 0.000875) |
179 |
SEAICE_mcPheeTaper (default is 0.) |
SEAICE_mcPheeTaper (default is 0.) |
180 |
SEAICE_frazilFrac (defualt is 1.) |
SEAICE_frazilFrac (defualt is 1.) |
181 |
SEAICE_mcPheeStepFunc (default is false) |
SEAICE_mcPheeStepFunc (default is false) |
182 |
- changes results at trunction error level. |
- changes results at trunction error level |
183 |
o added pkg/frazil - see frazil_description.tex for details |
- ifdef SEAICE_EXTERNAL_FLUXES, disallow SEAICE_waterAlbedo |
184 |
|
o added pkg/frazil: |
185 |
|
- see frazil_description.tex for details |
186 |
|
- verification experiment is global_with_exf/input.yearly; changes output.yearly.txt |
187 |
o pkg/shelfice/shelfice_thermodynamics.F |
o pkg/shelfice/shelfice_thermodynamics.F |
188 |
- fix exponents for velocity dependent exchange coefficients: change |
- fix exponents for velocity dependent exchange coefficients: change |
189 |
from 2/3(=0) to 0.66666666666666667 |
from 2/3(=0) to 0.66666666666666667 |
190 |
- move computation of velocity dependent exchange coefficients into |
- move computation of velocity dependent exchange coefficients into |
191 |
separate loop, so that it applies to both thermodynamics |
separate loop, so that it applies to both thermodynamics |
192 |
- move some divisions by constants out of loops |
- move some divisions by constants out of loops |
193 |
|
|