| 3 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
| 4 |
============================== |
============================== |
| 5 |
|
|
| 6 |
|
o clean up build options files for SunOS, in particular |
| 7 |
|
- remove -traditional from xmakedepend |
| 8 |
|
- change -xarch=v9 to -xarch=native for f77 |
| 9 |
|
- restore use of make (not gmake) |
| 10 |
|
- add CC=gcc if using g77 |
| 11 |
|
o changes per discussion of AM, JMC, & EH3 today: |
| 12 |
|
- move redundant "grid" variables out of "state" -- in preparation of |
| 13 |
|
creating proper CF-style coordinate variables and in recognition |
| 14 |
|
that any/all grid variables can be easily copied (eg. w/NCO) from |
| 15 |
|
the grid files into any of the other NetCDF files |
| 16 |
|
- *always* output grid variables with MDSIO (this will eventually be |
| 17 |
|
changed) |
| 18 |
|
- output MNC grid file if useMNC is true |
| 19 |
|
o every instance of _END_MASTER() has been found and replaced with |
| 20 |
|
_END_MASTER( myThid ) in order to satisfy certain picky Sun |
| 21 |
|
preprocessors |
| 22 |
|
o re-arrange diagnostics pkg initialisation: |
| 23 |
|
- allow each package to extend the available diagnostics list |
| 24 |
|
- add some checking and fix small problems (multi-threaded, ...) |
| 25 |
|
- new S/R diagnostics_fill (replace fill_diagnostics): |
| 26 |
|
look through the short list of active diag. (instead of the long list) |
| 27 |
|
create function DIAGNOSTICS_IS_ON to tell if a diagnostics is active |
| 28 |
|
|
| 29 |
|
checkpoint57a_post |
| 30 |
|
o this set of changes restores TAMC compatibility |
| 31 |
|
tested for global_ocean.90x40x15 adjoint on trough |
| 32 |
|
(what a pain) |
| 33 |
|
|
| 34 |
|
checkpoint57a_pre |
| 35 |
|
o some modifs in tools/build_options for sunos |
| 36 |
|
o call ctrl_pack for ctrl at end for optimcycle=0 |
| 37 |
|
|
| 38 |
|
checkpoint57 |
| 39 |
|
o add horizontal bi-harmonic viscosity for vertical velocity |
| 40 |
|
- new parameter viscA4W (defaults to viscA4) |
| 41 |
|
- affects only non-hydrostatic code |
| 42 |
|
- to be done: no-slip lateral boundary conditions |
| 43 |
|
o switch back to "sflux = sflux - runoff" in exf_getforcing.F |
| 44 |
|
o implement Implicit Vertical advection for pTracers |
| 45 |
|
o depth convergence accelerator: replace deltaTtracer by dTtracerLev(k) |
| 46 |
|
(implemented also for implicit vertical diffusion & advection) |
| 47 |
|
o OBCS: extend application of tracer open boundary conditions |
| 48 |
|
to a strip beyond the boundary of width Olx/Oly |
| 49 |
|
o add call monitor to TAF set of required calls (flow directives) |
| 50 |
|
and include in natl_box_adjoint |
| 51 |
|
o untangle a few ifdefs related to cost, ctrl, autodiff |
| 52 |
|
untangel a few ALLOW_ECCO_OPTIMIZATION |
| 53 |
|
o Apply Cooper & Haines representative estimates also at sub-surface (k=Nr) |
| 54 |
|
(theta2, salt2 in ecco_cost_weights) |
| 55 |
|
|
| 56 |
|
checkpoint56c_post |
| 57 |
|
o OBCS as control variables |
| 58 |
|
- update ctrl_ad_diff.list, obcs_ad_diff.list |
| 59 |
|
- remove balance of obcs controls from default |
| 60 |
|
- fix index bug nobcs in ctrl_init |
| 61 |
|
- fix dummy fields filen in ctrl_pack |
| 62 |
|
- add dummy weights for obcs |
| 63 |
|
o lsopt: comment out fort.94 output |
| 64 |
|
o optim: |
| 65 |
|
- remove unnecessary header files |
| 66 |
|
- adjusted namelists |
| 67 |
|
- add xerbla.F to Makefile |
| 68 |
|
o exf: |
| 69 |
|
- multiple modifications of exf_check_range |
| 70 |
|
- unit change in exf_init_runoff from [m/year] to [m/s] |
| 71 |
|
o mdsio: |
| 72 |
|
- i/o-vector change debugLevel from debLevA to debLevB |
| 73 |
|
|
| 74 |
|
checkpoint56b_post |
| 75 |
|
o collect all ptracers output in an new routine (PTRACERS_OUTPUT) that |
| 76 |
|
is called from DO_THE_MODEL_IO |
| 77 |
|
- rename PTRACERS_STATV_WRITE to PTRACERS_WRITE_TIMEAVE |
| 78 |
|
- introduce new parameter PTRACERS_dumpFreq (defaults to dumpFreq) |
| 79 |
|
- clean up PTRACERS_WRITE_TIMEAVE a little |
| 80 |
|
o add another verification experiment that tests both the cfc-pkg and |
| 81 |
|
the gchem-pkg with GCHEM_SEPARATE_FORCING undefined, that is with |
| 82 |
|
reactive tendencies treated inline with regular timestepping |
| 83 |
|
o testreport previously did a lousy job of comparing small numbers, so |
| 84 |
|
we (Ed and Martin) changed the way the cut-off of 1.e-12 is |
| 85 |
|
implemented in the little c-progam tmp_cmpnum.c which is embedded in |
| 86 |
|
testreport. |
| 87 |
|
o ptracers_monitor: bug fix for MPI |
| 88 |
|
|
| 89 |
|
checkpoint56a_post |
| 90 |
|
o GCHEM: finish reorganizating the package |
| 91 |
|
- forward_step calls GCHEM_CALC_TENDENDY, which computes gchemTendency |
| 92 |
|
(introduces another 3D-array for each passive tracer, but only if |
| 93 |
|
GCHEM_SEPARATE_FORCING is undefined. For GCHEM_SEPARATE_FORCING |
| 94 |
|
gchemTendency is not needed because the timestep is done separately) |
| 95 |
|
- gchemTendency is added to gPtr in GCHEM_ADD_TENDENCY |
| 96 |
|
- GCHEM is now---more or less cleanly---separate from PTRACERS |
| 97 |
|
- to bo done: handling of surfaceForcingPtr that is needed for KPP not |
| 98 |
|
so clear to me how that should be done in a general way. |
| 99 |
|
o CFC: improve formatting of output in CFC_ATMOS |
| 100 |
|
o PTRACERS: |
| 101 |
|
- rename GCHEM_MONITOR to PTRACERS_MONITOR and call it from MONITOR, so |
| 102 |
|
that ALL experiments with ptracers enable can be checked. This makes |
| 103 |
|
GCHEM_MONITOR obsolete. |
| 104 |
|
- include a runtime parameter PTRACERS_monitorFreq that defaults to |
| 105 |
|
monitorFreq |
| 106 |
|
- set default PTRACERS_write_mdsio to false if PTRACERS_write_mnc is true |
| 107 |
|
o add PTRACERS (PTR0[1-5]) to testreport output |
| 108 |
|
o exf_check_range: |
| 109 |
|
* default changed to .TRUE. |
| 110 |
|
* will check values for niter0 if debLevA, every timestep if debLevB |
| 111 |
|
* STOP in exf_check_range if 'out of range' |
| 112 |
|
* range chosen generous to allow for potential extema |
| 113 |
|
TBD: a version with stricter range |
| 114 |
|
permitted range would be runtime parameters |
| 115 |
|
o GMredi: |
| 116 |
|
- fix ldd97 slope limit ; extend valid domain of Psi-Bolus ; |
| 117 |
|
- change S/R GRAD_SIGMA to be valid in the overlap on CS-grid. |
| 118 |
|
- change the default GMREDI_OPTIONS.h: compile everything except Visbeck |
| 119 |
|
o GCHEM: |
| 120 |
|
- move call to GCHEM_FORCING_SEP, so that it is now called before |
| 121 |
|
the blocking exchanges to avoid computing halos in GCHEM_FORCING_SEP |
| 122 |
|
- introduced a useGCHEM runtime flag and applied it everywhere. |
| 123 |
|
- renamed PTRACERS_SEPARATE_FORCING to GCHEM_SEPARATE_FORCING; |
| 124 |
|
this flag is now only visible within GCHEM-pkg routines. |
| 125 |
|
- moved gchem_forcing_int temporarily into ptracers_forcing, this will |
| 126 |
|
be replaced by a 3D array that is computed before thermodynamics; for |
| 127 |
|
this, the tracer ID iTracer had to be included into the parameter list |
| 128 |
|
of ptracers_forcing |
| 129 |
|
- set up forward_step to call gchem_forcing, that will replace |
| 130 |
|
gchem_forcing_int in ptracers_forcing |
| 131 |
|
- cleaning up |
| 132 |
|
o small cleaning of exch2: remove files that are also in eesupp/src |
| 133 |
|
|
| 134 |
|
checkpoint56 |
| 135 |
|
o AIM: read land-Fraction earlier (from ini-fixed) and monitor land at nIter0 |
| 136 |
|
o mnc: cleanup "full" files--fixes out-of-memory problem |
| 137 |
|
o ECCO-GODAE |
| 138 |
|
* model/, pkg/, code/, input/ now complete w.r.t. ECCO-GODAE production setup |
| 139 |
|
* fully merged and updated from SIO code |
| 140 |
|
windstress sign change affects |
| 141 |
|
- exf_inscal_[u/v]stress |
| 142 |
|
- sign of xx_tau[u/v] adjustments |
| 143 |
|
- signs in cost_scat |
| 144 |
|
* ctrl_get_gen: avoid update of tau[u/v] for first few days |
| 145 |
|
* natl_box_adjoint updated and compressed |
| 146 |
|
* 2-level checkpoint switch added |
| 147 |
|
(tested in EXP. 1 of natl_box_adjoint) |
| 148 |
|
* new switch doInitXX to avoid init. of xx_... in optimcycle=0 |
| 149 |
|
o SVD |
| 150 |
|
* first complete, non-hacked version for SVD calculations |
| 151 |
|
* all interfaces clean w.r.t to handling of xx_..., adxx_..., g_xx_... |
| 152 |
|
performed via direct access files admtlm_vector.it???? |
| 153 |
|
(via ctrl_[un]pack) |
| 154 |
|
* optimcycle is used for ARPACK iteration counter |
| 155 |
|
(note: adxx_... is iterated, rather than xx_...; latter is constant!) |
| 156 |
|
* EV problem ok w.r.t. L2-norm in tangent space; need to check other norms |
| 157 |
|
and add linear transforms where necessary |
| 158 |
|
* controls are initial/final SST, SSS (can be extended; need to check |
| 159 |
|
ctrl_set_[un]pack_xyz.F according to ctrl_set_[un]pack_xy.F |
| 160 |
|
o mdsio: filename info to both stderr, stdout |
| 161 |
|
o exf: exf_inscal_runoff: separated default conversion [m/year] -> [m/sec] |
| 162 |
|
from sign change |
| 163 |
|
|
| 164 |
|
checkpoint55j_post |
| 165 |
|
o isolate dissipation tendency (allow to keep it out off AB) in V.I form. |
| 166 |
|
affects (truncation level) output.txt of global_ocean.cs32x15(&.thsice) |
| 167 |
|
o add consistent build directories and .cvsignore files to verification |
| 168 |
|
o added viscA4GridMin and viscA4GridMax for more control on biharmonic viscosity |
| 169 |
|
o make all subroutines 30 or less characters in length for certain older |
| 170 |
|
compilers to work (a thorough search only turns up two routines in |
| 171 |
|
pkg/ptracers) |
| 172 |
|
o allow (vecinv only) a different horiz. viscosity for Divergence and Vorticity; |
| 173 |
|
affects (truncation level) output.txt of global_ocean.cs32x15(&.thsice) |
| 174 |
|
|
| 175 |
|
checkpoint55i_post |
| 176 |
|
o change arrays for total vertical diffusivity (2.D if explicit, |
| 177 |
|
3.D if implicit) ; each ptracer uses its own background value. |
| 178 |
|
o added mnc/NetCDF output capability to ocn_compon_interf |
| 179 |
|
- added bits to verification/cpl_aim+ocn so that it can run on |
| 180 |
|
ITRDA and other clusters without the mpirun "-wd" option |
| 181 |
|
o fixed wayward endif in mom_fluxform.F that messed up cyclindrical |
| 182 |
|
configuration |
| 183 |
|
o give MNC the ability to create additional files rather than |
| 184 |
|
exceeding a specified file size limit "MNC_MAX_FILESIZE" |
| 185 |
|
- the default is just less than 2GB which is a practical limit |
| 186 |
|
on many systems |
| 187 |
|
o Compute the non-local term using either KPP vertical diffusivity (=default) |
| 188 |
|
or the total vertical diffusivity (when KPP_ghatUseTotalDiffus=T) |
| 189 |
|
|
| 190 |
|
checkpoint55h_post |
| 191 |
|
o add verification test "vermix" |
| 192 |
|
o remove all the *_ioinc flags and replace with the single global |
| 193 |
|
outputTypesInclusive flag |
| 194 |
|
o allow to set a vertical profile of vertical diffusivity for T & S |
| 195 |
|
o call PACKAGE_BOOT & PACKAGES_READPARMS just after INI_PARMS |
| 196 |
|
o useMNC moved from data.mnc to data.pkg |
| 197 |
|
|
| 198 |
|
checkpoint55g_post |
| 199 |
|
o Renaming carbon AD experiments to comply with testreport conventions |
| 200 |
|
o Updating adjoint of Held-Suarez hs94.1x64x5 |
| 201 |
|
o Fixing directives for non-ECCO forward_step, after ECCO directives |
| 202 |
|
have been fixed to avoid recomp. of do_oceanic_phys |
| 203 |
|
Works well for global_ocean.90x40x15, |
| 204 |
|
but recomp. issue remains for carbon |
| 205 |
|
o Updating adjoint output for natl_box_adjoint after |
| 206 |
|
pkg/ecco has been modified (see taga c55d_post) |
| 207 |
|
|
| 208 |
|
checkpoint55f_post |
| 209 |
|
o rename to timeave_init_fixed |
| 210 |
|
|
| 211 |
|
checkpoint55e_post |
| 212 |
|
o some delicate re-shuffle of store directives to avoid one |
| 213 |
|
extra call of do_oceanic_physics |
| 214 |
|
affects ecco_the_main_loop.F, do_oceanic_physics.F |
| 215 |
|
checkpoint_lev1_directives.h |
| 216 |
|
o NB: this may break global_ocean adjoint temporarily, |
| 217 |
|
but it is clear how to fix it. Will do later, need this now. |
| 218 |
|
|
| 219 |
|
checkpoint55d_post |
| 220 |
|
o ECCO specific cost function terms (up-to-date with 1x1 runs) |
| 221 |
|
o ecco_cost_weights is modified to 1x1 runs |
| 222 |
|
o modifs to allow observations to be read in as |
| 223 |
|
single file or yearly files |
| 224 |
|
o enable to read exf forcing fields as either |
| 225 |
|
single file or yearly files (flag useExfYearlyFields) |
| 226 |
|
checkpoint55d_pre |
| 227 |
|
|
| 228 |
|
o switch between 2-level and 3-level checkpointing |
| 229 |
|
o add ability of MNC to write local and "partial" (eg. 2D slices where |
| 230 |
|
the full 3D field is never actually stored) arrays to NetCDF files |
| 231 |
|
with the correct (that is, the complete multi-dimensional) set of |
| 232 |
|
array indicies |
| 233 |
|
- used in mom_vecinv() to write the diagFreq output |
| 234 |
|
- tested (demonstrated) in verification/aim.5l_cs |
| 235 |
|
o further mnc-ification |
| 236 |
|
- add verification/lab_sea to the list with mnc setups |
| 237 |
|
- fix undefined mnc v-names in cd_code |
| 238 |
|
- add kpp to mnc output (both snapshot and timeave) |
| 239 |
|
o non-hydrostatic code |
| 240 |
|
- added horizontal viscosity parameter viscAhW for w, that replaces viscAh |
| 241 |
|
in calc_gw.F |
| 242 |
|
- defaults to viscAh |
| 243 |
|
- this is useful, when I want to use viscAh=0 (e.g., only biharmonic |
| 244 |
|
friction or Leith) in non-hydrostatic mode, as long as there is |
| 245 |
|
no biharmonic friction and/or Leigth or equivalent for vertical velocity. |
| 246 |
|
o pkg/cost |
| 247 |
|
extend simple cost functions to treat time mean quantities |
| 248 |
|
(T,U,V) via cost_accumulate_state from cost_tile |
| 249 |
|
o multidimAdvection on CS-grid: implement the full 3 pass ; overlap needs to |
| 250 |
|
be 2 times larger than the minimum size the advection scheme requires. |
| 251 |
|
now get a perfect conservation of salt in advect_cs test-exp (updated) |
| 252 |
|
o GGL90 (Gaspar's TKE mixed layer) |
| 253 |
|
- add horizontal diffusion of TKE, requires exchanges in |
| 254 |
|
do_fields_blocking_exchanges, horizontal diffusivity is zero by |
| 255 |
|
default. In OPA there is no horizontal diffusion of TKE but the |
| 256 |
|
mixing coefficients are computed from a horizontal average of TKE of |
| 257 |
|
6 points or so. I think that diffusion has a little more physical |
| 258 |
|
justification. |
| 259 |
|
- clean up ggl90_calc in the hope of reducing memory usage (this hope |
| 260 |
|
was in vain) |
| 261 |
|
- mask tke-variable in ggl90_init |
| 262 |
|
o fix mask Pb in multidimAdvection on CS-grid (still a small Pb with ipass) |
| 263 |
|
o fix biharmonic diffusion on CS-grid. |
| 264 |
|
o more obcs stuff |
| 265 |
|
- added obcs_external_fields_load as an alternative to the EXF-pkg option |
| 266 |
|
for reading boundary values. This routine is called from |
| 267 |
|
obcs_prescribe_read |
| 268 |
|
- updated obcs_check |
| 269 |
|
- changed sponge layer relaxation parameter units for days to seconds |
| 270 |
|
- turned off sponge layer by default |
| 271 |
|
- changed argument list of OBCS_CALC to myTime+deltaTclock |
| 272 |
|
- fixed about 35 bugs in obcs_external_fields_load |
| 273 |
|
|
| 274 |
|
checkpoint55c_post |
| 275 |
|
o obcs stuff |
| 276 |
|
- new runtime flags useOBCSprescribe, useOBCSbalance |
| 277 |
|
- re-arranged obcs_calc |
| 278 |
|
- moved exf_getobcs to obcs_prescibe_read |
| 279 |
|
obcs_prescibe_read is called after ORLANSKI stuff |
| 280 |
|
and independent of it (executable can contain all). |
| 281 |
|
- updated obcs_check |
| 282 |
|
- added exchanges for U,V,gU,gV for useOBCSprescribe=.TRUE. |
| 283 |
|
These may be necessary only if tangential velocities are |
| 284 |
|
prescribed at OB, not sure, needs testing. |
| 285 |
|
o make the MNC_CW_ADD_VATTR_* subroutines all take scalar arguments |
| 286 |
|
which simplifies the calls and passes the ifc check-bounds test |
| 287 |
|
|
| 288 |
|
checkpoint55b_post |
| 289 |
|
o add a 2D scalar, Z location, exchange_2 routine. |
| 290 |
|
o EXCH_UV for CS-grid: add one u,v in the corner-halo region, allows |
| 291 |
|
to compute vorticity on a wider stencil [e.g., vort3(0,1)&(1,0)] |
| 292 |
|
affects advect_cs results (for the wrong reason): update output.txt |
| 293 |
|
o use a more standard discretization for biharmonic viscosity ; |
| 294 |
|
(original version still available with #define MOM_VI_ORIGINAL_VISCA4 ) |
| 295 |
|
affects results of dome set-up: update output.txt |
| 296 |
|
o fix gad_advection.F (CubedSphere part), back to version 1.23 |
| 297 |
|
o fix bug in mom_vi_del2uv.F (CubedSphere part). |
| 298 |
|
|
| 299 |
|
checkpoint55a_post |
| 300 |
|
o merged code to |
| 301 |
|
* prescribe/read time-dependent open boundaries |
| 302 |
|
(works in conjunction with exf, cal) |
| 303 |
|
* sponge layer code for open boundaries |
| 304 |
|
* each boundary N/S/E/W now has its own CPP option |
| 305 |
|
(healthy for the adjoint) |
| 306 |
|
|
| 307 |
|
checkpoint55 |
| 308 |
|
o up-to-date adjoint of previous breaks |
| 309 |
|
- natl_box_adjoint: problem for exp 12, 13: cost = 0. needs fix |
| 310 |
|
- global_ocean_ebm: adjoint still needs fix (TLM ok) |
| 311 |
|
- global_ocean.90x40x15: extended GMREDI_OPTIONS.h |
| 312 |
|
- carbon: * remove tr1-related adjoint verification |
| 313 |
|
* add ptracer as control variable |
| 314 |
|
o remove all tr1-related code (ALLOW_PASSIVE_TRACER) |
| 315 |
|
(adjoint stuff still has some tr1 'names', but all use ptracer arrays) |
| 316 |
|
|
| 317 |
|
checkpoint54f_post |
| 318 |
|
o in verification/carbin/ remove experiment related to |
| 319 |
|
ALLOW_PASSIVE_TRACER (array tr1 etc) |
| 320 |
|
code_ad/ input_ad/ results_ad |
| 321 |
|
|
| 322 |
|
o add two new packages |
| 323 |
|
- ggl90, TKE-model of Gaspar et al. (1990), Blanke+Delecluse (1993) |
| 324 |
|
+ subject to optimization (memory) |
| 325 |
|
- opps, OPPS (ocean penetrative plume scheme) of Paluszkiewicz+Romea (1997) |
| 326 |
|
+ subject to reordering (many GOTOs in the original code) |
| 327 |
|
+ so far opps is called for every (i,j) point, and then operates in |
| 328 |
|
the vertical; one may want to reorder the scheme in order operate in |
| 329 |
|
horizontal slabs at the cost of (a lot of) memory, at least ten new 3D |
| 330 |
|
fields) |
| 331 |
|
o overhaul of IO so that we now have flags for MDSIO and/or MNC |
| 332 |
|
- all verification tests compile and run with linux_ia32_g77 |
| 333 |
|
- defaults are compatible with current input files--nothing |
| 334 |
|
should change if you were not previously using MNC |
| 335 |
|
- MNC output has been added in numerous places (eg. timeave) |
| 336 |
|
but there are still a few writes not yet do-able with MNC |
| 337 |
|
(this is in progress) |
| 338 |
|
- flags now allow for either/or/both use of MDSIO and MNC and |
| 339 |
|
documentation will soon follow |
| 340 |
|
- numerous small formatting cleanups for ProTeX |
| 341 |
o add version info (via BUILD_INFO.h) to MNC and stdout |
o add version info (via BUILD_INFO.h) to MNC and stdout |
| 342 |
|
|
| 343 |
checkpoint54e_post |
checkpoint54e_post |
Parent Directory
| 
Revision Log
| 
Revision Graph
| 
Patch