| 2 |
|
|
| 3 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
| 4 |
============================== |
============================== |
| 5 |
|
o fix bug in dxV,dyU definition of cubed-sphere grid (ini_curvilinear_grid) |
| 6 |
|
|
| 7 |
|
checkpoint51c_post |
| 8 |
|
o introducing integer flag debugLevel |
| 9 |
|
- debLevZero: suppress all output |
| 10 |
|
- debLevA: standard |
| 11 |
|
- debLevB: equivalent to current debugMode=.TRUE. |
| 12 |
|
o introducing pathname variable mdsioLocalDir for mdsio |
| 13 |
|
that can be specified at runtime |
| 14 |
|
(relevant for cluster I/O to local disk) |
| 15 |
|
` |
| 16 |
|
>>>checkpoint51b_post |
| 17 |
|
adjoint of ptracers |
| 18 |
|
o disentangled ALLOW_PTRACERS using new ALLOW_GCHEM |
| 19 |
|
o extended state in checkpoint_lev?, etc. |
| 20 |
|
o alternative ctrl/cost with tr1 or ptracer |
| 21 |
|
o new comlev1_bibj_ptracers |
| 22 |
|
o new ptracers.flow |
| 23 |
|
o made ptracers_forcing same as external_forcing_s |
| 24 |
|
o added verification for adjoint with ptracers |
| 25 |
|
- adm. grdchk fail so far |
| 26 |
|
- tlm. grdchk OK |
| 27 |
|
- checked tapelev/comlev recomputations: OK |
| 28 |
|
<<<checkpoint51b_pre |
| 29 |
|
|
| 30 |
|
checkpoint51a_post |
| 31 |
|
o added code for biogeochemistry package |
| 32 |
|
- new pkgs: dic, gchem |
| 33 |
|
- new subroutine in ptracers |
| 34 |
|
o modify code for biogeo packages |
| 35 |
|
- in model/src and pkg/ptracers |
| 36 |
|
|
| 37 |
|
checkpoint51 |
| 38 |
|
o merged improved packages from ecco-branch (ecco_c51_e34) to MAIN |
| 39 |
|
- autodiff, ctrl, exf, grdchk, seaice |
| 40 |
|
kept separate packages |
| 41 |
|
- cost, ecco, obcs |
| 42 |
|
o updating verification/internal_wave, verification/lab_sea |
| 43 |
|
(all verif's pass on faulks.lcs.mit.edu RH7.3, g77) |
| 44 |
|
|
| 45 |
|
checkpoint50i_post |
| 46 |
|
o Preparing next differentiable checkpoint and sync |
| 47 |
|
of MAIN vs. ecco-branch |
| 48 |
|
- 1st step: this one is differentiable |
| 49 |
|
- 2nd step: next(! not this) checkpoint will have updated packages |
| 50 |
|
from ecco-branch) |
| 51 |
|
|
| 52 |
|
checkpoint50h_post |
| 53 |
|
o new package "land": provide surface BC for AIM. |
| 54 |
|
o pkg/ptracers: apply exchange to all tracers after reading a pickup. |
| 55 |
|
o AIM: use true log-P extrapolation for near surface temp (= as in |
| 56 |
|
F.Molteni paper) ; update aim.5l_cs output. |
| 57 |
|
o New variable in PARM03 for more flexible pickups |
| 58 |
|
o somewhat cleaned package initialisation sequence for |
| 59 |
|
ctrl/ cost/ ecco/ in packages_... routines |
| 60 |
|
|
| 61 |
|
checkpoint50g_post |
| 62 |
|
o merged with release1_p17 (pkg/seaice and verification/lab_sea) |
| 63 |
|
- added SEAICE_MULTILEVEL for 8-category sea-ice thermodynamics |
| 64 |
|
- LSR sea-ice dynamic solver moved to SouthWest B-grid location and |
| 65 |
|
made the default because of faster convergence than ADI |
| 66 |
|
|
| 67 |
|
checkpoint50f_post |
| 68 |
|
o exf totally restructured |
| 69 |
|
- replaced all exf_set_ routines by one generic exf_set_gen.F |
| 70 |
|
and made sure it's still differentiable |
| 71 |
|
- split exf_getffields.F into new exf_bulkformulae.F |
| 72 |
|
- exf_inscal_/exf_outscal_ suffixes consistent with field names |
| 73 |
|
- updated adjoint/makefile and relevant verif. accordingly |
| 74 |
|
|
| 75 |
|
checkpoint50f_pre |
| 76 |
|
o AIM: near surface temp evaluated from Pot.Temp: fixed for Part Cell ; |
| 77 |
|
+ change Evap over land. (update aim.5l_cs output) |
| 78 |
|
o added missing .cvsignores and made uniform across expts |
| 79 |
|
o changed configuration of tidal_basin_2d expt |
| 80 |
|
o added pkg/debug functions for tracing code in fatal crashes |
| 81 |
|
o tidied pkg/monitor to use MONITOR_OPTIONS.h |
| 82 |
|
o new fn in pkg/monitor to stop model if solution grows too big |
| 83 |
|
o removed lots of single quotes from comments |
| 84 |
|
|
| 85 |
|
checkpoint50e_post |
| 86 |
|
o make KPP work with PTRACERS |
| 87 |
|
- fix gad_calc_rhs to call new routine kpp_transport_ptr, which is |
| 88 |
|
nearly a copy of kpp_transport_s |
| 89 |
|
- there is no analogue to SurfaceTendencyS, so I have to use |
| 90 |
|
gPtr(of the surface layer) instead |
| 91 |
|
o add a new platform SunFire+mpi (SunFire 15000) to genmake |
| 92 |
|
checkpoint50e_pre |
| 93 |
|
|
| 94 |
|
checkpoint50d_post |
| 95 |
|
o change kpp output from multiple-record state files to single-record state |
| 96 |
|
files analogous to write_state.F |
| 97 |
|
o reduce the output frequency of cg3d-related stuff to the monitor frequency, |
| 98 |
|
analogous to the cg2d-related output. |
| 99 |
|
o fix small problem with in ptracers_write_checkpoint.F: len(suff)=512, |
| 100 |
|
so that writing to internal file fn (with length 512) fails. |
| 101 |
|
checkpoint50d_pre |
| 102 |
|
|
| 103 |
|
checkpoint50c_post |
| 104 |
|
o fix problem in ini_procs.F when using MPI (was there since checkpoint48f) |
| 105 |
|
o add monitor diagnostic of vorticity (mon_vort3.F) and potential energy. |
| 106 |
|
o add time-average diagnostic of U*V and vertical diffusive flux for theta. |
| 107 |
|
o bug fixed in convective adjustment diagnostic (multi-tile set-up). |
| 108 |
|
o 4th Order Adv scheme: return to previous (c40pre7) masking near boundaries. |
| 109 |
|
|
| 110 |
|
checkpoint50c_pre |
| 111 |
|
Merging from release1_p13: |
| 112 |
|
o bug fix for pkg/seaice dynamic solver |
| 113 |
|
o Added SEAICE_initialHEFF to pkg/seaice |
| 114 |
|
|
| 115 |
|
checkpoint50b_post |
| 116 |
|
o store u*,v* in gU,V instead of in guNm1,gvNm1 |
| 117 |
|
o in order to put the momForcing out of the Adams-Bashforth: |
| 118 |
|
move forcing & CD-scheme calls from mom_fluxform & mom_vecinv |
| 119 |
|
to timestep.F |
| 120 |
|
o new flag "useCDscheme" (default=F); replace guCD,gvCD by local arrays |
| 121 |
|
o re-generate output for highly sensitive test-exp: |
| 122 |
|
global_ocean_pressure, ideal_2D_oce, lab_sea, natl_box |
| 123 |
|
|
| 124 |
|
checkpoint50b_pre |
| 125 |
|
o restore the use of momAdvection & useCoriolis flags (mom_vecinv) |
| 126 |
|
o corrected CPP option mismatch in exf_Bulk... functions for case |
| 127 |
|
#define ALLOW_ATM_WIND && #undef ALLOW_ATM_TEMP |
| 128 |
|
|
| 129 |
|
checkpoint50a_post |
| 130 |
|
o Bug fix for merging between c50 and KPP. |
| 131 |
|
ikey was passed from thermodynamics to kpp_calc via |
| 132 |
|
common block rather than being recomputed in kpp_calc, |
| 133 |
|
in contradiction with new key itdkey. |
| 134 |
|
New key ikppkey created, and tamc.h headers updated. |
| 135 |
|
|
| 136 |
|
checkpoint50 |
| 137 |
|
- 2nd step of merge with ecco-branch: |
| 138 |
|
merge changes in ecco-branch back into main trunk |
| 139 |
|
- bug fix in find_alpha |
| 140 |
|
- more storing in gmredi_calc_tensor.F, kpp_routines.F |
| 141 |
|
- avoid iterative loop in ini_pressure for ALLOW_AUTODIFF_TAMC |
| 142 |
|
- merged some sliced AD support routines |
| 143 |
|
|
| 144 |
|
checkpoint49 |
| 145 |
|
merged from ecco-branch: |
| 146 |
|
o exf: |
| 147 |
|
- Enable initialisation of forcing fields to constant |
| 148 |
|
(runtime) values. |
| 149 |
|
- in exf_getffields.F |
| 150 |
|
Reduce i-/j-loop to interior domain, discarding overlaps. |
| 151 |
|
That also fixes wrong TAF-key computations for key_1, key_2 |
| 152 |
|
with bulf formulae. |
| 153 |
|
- exf_init.F modify #ifdef for exf_init_evap |
| 154 |
|
- exf_getffieldrec.F, ctrl_getrec.F |
| 155 |
|
The following INT-usages are not safe: |
| 156 |
|
fldsecs = int(fldsecs/fldperiod)*fldperiod |
| 157 |
|
fldcount = int(fldsecs/fldperiod) + 1 |
| 158 |
|
and were modified. |
| 159 |
|
o autodiff: checkpoint_lev... |
| 160 |
|
- updated storing for bulk formulae (some arrays were missing) |
| 161 |
|
- added state for seaice package |
| 162 |
|
- synchronized with main branch |
| 163 |
|
o DIVA: |
| 164 |
|
Modif.s in |
| 165 |
|
- the_main_loop.F |
| 166 |
|
- the_model_main.F |
| 167 |
|
and added cost_final_restore.F cost_final_store.F |
| 168 |
|
o With these changes, it is possible to fully separate |
| 169 |
|
the first forward run to compute the cost function (mdthe_main_loop) |
| 170 |
|
and adjoint runs over individual DIVA intervals (adthe_main_loop) |
| 171 |
|
into separate model executions. |
| 172 |
|
This enables better tuning of DIVA to given queue sizes. |
| 173 |
|
o Loss of cost function value in consecutive DIVA runs is |
| 174 |
|
fixed by storing to 'costfinal' file. |
| 175 |
|
o Appropriate call of ctrl_unpack/_pack,grdchk_main at |
| 176 |
|
beginning/end of adjoint integration is accounted for. |
| 177 |
|
The current changes apply to TAF-1.4.26. |
| 178 |
|
TAF has one remaining bug, requiring one manual change |
| 179 |
|
(see MITgcm manual). |
| 180 |
|
o cal: synchronized between MAIN and ecco-branch |
| 181 |
|
|
| 182 |
checkpoint48i_post |
checkpoint48i_post |
| 183 |
o Changes to restore differentiability of code w.r.t. previous tag |
o Changes to restore differentiability of code w.r.t. previous tag |
Parent Directory
| 
Revision Log
| 
Revision Graph
| 
Patch