--- MITgcm/doc/tag-index 2007/06/01 12:57:48 1.826 +++ MITgcm/doc/tag-index 2008/08/02 23:12:23 1.1012 @@ -1,9 +1,592 @@ -$Header: /home/ubuntu/mnt/e9_copy/MITgcm/doc/tag-index,v 1.826 2007/06/01 12:57:48 mlosch Exp $ +$Header: /home/ubuntu/mnt/e9_copy/MITgcm/doc/tag-index,v 1.1012 2008/08/02 23:12:23 jmc Exp $ $Name: $ Notes on tags used in MITgcmUV ============================== +o exch2: + change index bounds in rx2_cube exchanges (new S/R: EXCH2_GET_UV_BOUNDS) + - no longer depend on the order sequence (N,S,E,W). + - 3rd exchange no longer needed (tested with 24 tiles). + - same modif to hand-written adjoint S/R (global_ocean.cs32x15: zero diff) + - exch_UV_A-grid readily available (but not yet tested). + comment out 3rd exch call and remove copy of V to U at SE & NE face-corners + (both are no longer needed) in exch_uv: no effect in forward run; but + get truncation error differences with adjoint exch. +o exch2: + - change index-bounds storage (moved from target to local tile, + more intuitive this way) ; rename/remove some variables. + - change utils/exch2/matlab-topology-generator & code_mods accordingly. + +checkpoint61a +o exch2/matlab-topology-generator: + fix bug in index bounds for N-S internal exchange (hard to tell if this + had any effect on the solution). +o pkg/mnc: small fix in mnc_handle_err: make sure that length of msgbuf + is exceeded +o testreport: modify syntax of awk, so that it works also with + awk, and not only with nawk and gawk + +checkpoint61 +o autodiff + Change default for TAF usage by removing argument '-nonew_arg' + This changes TAF subroutine argument list (ordering of active var.) + Omission of this flag leads to incompatibility w.r.t. TAMC + To restore TAMC-compatibility, need following modifs: + * use adoptfile tools/adjoint_options/adjoint_tamc_compatibility + * use CPP option #define AUTODIFF_TAMC_COMPATIBILITY + Tested TAF version is 1.9.22 + N.B.: exch2 hand-written adjoint code currently not TAMC compatible + +checkpoint60 +--->>> This is the last checkpoint for which generating the adjoint +--->>> is done via the TAMC-compatibility TAF-flag "-nonewarg" +--->>> by default. +--->>> Future checkpoints will ommit this flag by default. +o pkg/generic_advdiff, os7mp: simplify and avoid division by zero. +o pkg/grdchk: Add gradient check to etan0 perturbations. +o pkg/exf: read ice area from file (#ifdef ALLOW_ICE_AREAMASK). +o pkg/autodiff: more flexibility in zeroadj. + model/src: modifying do_oceanic_phys.F accordingly. +o pkg/ctrl: Bounds for (kapgm,kapredi,eddypsi) parameter estimation. +o model/src and pkg/gmredi: GMREDI_WITH_STABLE_ADJOINT CPP option + special setup of gmredi for which the adjoint is stable + enough for parameter optimization. +o bridging the gap between eddy stress and GM. + -> eddyTau is replaced with eddyPsi (eddyTau = f x rho0 x eddyPsi) + along with a change in CPP option (now ALLOW_EDDYPSI). + -> when using GM w/ GM_AdvForm: + The total eddy streamfunction (Psi = eddyPsi + K x Slope) + is applied either in the tracer Eq. or in momentum Eq. + depending on data.gmredi (intro. GM_InMomAsStress). + -> ALLOW_EDDYPSI_CONTROL for estimation purpose. + The key modifications are in model/src/taueddy_external_forcing.F + pkg/gmredi/gmredi_calc_*F pkg/gmredi/gmredi_*transport.F + +checkpoint59r +o new S/R (pkg/mdsio/mdsio_facef_read.F) reading of sNx+1 x sNy+1 array + from per-face file (take S/R READSYMTILE_RS out of ini_curvilinear_grid.F); + fix option to read coriolis(@ cell-corner) from files with CS-grid. +o pkg/mnc: add some attributes to standard 1D-coordinate variables +o pkg/diagnostics: enforce missing_value attribute for scalar variables at + mass points, when using mnc for output + - add runtime parameter (missing_value, missing_value_int) to override + default (which is the value of getcon('UNDEF')) +o pkg/mnc: prepare mnc for the actual use of missing values with the + diagnostics package: small fix in mnc_cw_readwrite.template and more + initialisation of never used fields + - replace mnc_common.h and mnc_id_header.h with corresponding upper case + versions +o pkg/thsice/thsice_ini_vars.F added default initial values so that + thermodynamics sea ice solver does not blow up when some initial fields + are missing. +o move from external_forcing_surf.F to ini_forcing.F + the p-coord. hack consisting in using pLoad to set phi0surf ; + update output of tutorial_global_oce_in_p (affected @ truncation level). +o Freshen up a bit hs94.1x64x5, notably: + remove rotationPeriod and tRef=5*0. (adjoint only), and add + exactConserv=.TRUE. +o pkg/generic_advdiff: + change option GAD_ALLOW_SOM_ADVECT to GAD_ALLOW_TS_SOM_ADV which only + applies to files where Temperature & Salinity 2nd Order moments are used. +o pkg/ptracers: add second-order moment advection schemes (80 and 81); + this uses a dynamically allocated internal state data structure + (#define PTRACERS_ALLOW_DYN_STATE in PTRACERS_OPTIONS.h) + and requires a fortran 90 compiler + +checkpoint59q +o pkg/mom_vecinv: Add energy & enstrophy conserving scheme (selectVortScheme=3) + (from Sadourny, described by Burridge & Haseler, ECMWF Rep.4, 1977) +o Refresh tutorial_global_oce_biogeo, notably: + - change bathymetry (put back seamounts in ACC, use hFacs) + - change EOS (POLY3 --> JMD95Z) + - use Bryan-Lewis vertical diffusivity + - use multidimensional advection + - remove old virtual virtual in dic pkg, now done through ptracer pkg + - change a couple of parameters in dic pkg +o pkg/ctrl: change formula in adctrl_bound.F and add a parameter to ctrl_map_ini_gen.F +o pkg/obcs: change default to OBCSfixTopo = .true., + - set OBCSfixTopo=.false. in exp4/input/data.obcs to reproduce old result + - replace "debugMode" with "debugLevel .GE. debLevB" +o implementation of TAF directives for gchem,cfc,dic +o remove "exp5" (replaced by tutorial_deep_convection). +o pkg/obcs: add code to fix topography gradients normal to open boundaries; + this requires computing the index fields OB_Jn/Js/Ie/Iw in obcs_readparms; + the code can be turned on by setting OBCSfixTopo = .true. in data.obcs + --> changes results in exp4, seaice_obcs +o tutorial_deep_convection : + update the set-up and increase resolution and domain size ; start + from pseudo initial conditions = output at t=120mn in order to see + something happening when running testreport. +o pkg/seaice: add os7mp as an advection scheme + add check for advection schemes + add a flag for restoring T and S underneath sea ice +o pkg/my82: add netcdf-support and diagnostics +o thsice pkg: + - Now hsMax limits snow height by turning it into ice following + flooding scheme (and now conserving energy) + - Slight rewriting of the formula in thsice_calc_thick.F + - Parameter rhowi eliminated to be replaced by floodFac = (rhosw-rhoi)/rhos + --> changes results of global_ocean.cs32x15.icedyn and global_ocean.cs32x15.thsice + +checkpoint59p +o gchem / dic / cfc : + - add useDIC, useCFC , read from "data.gchem" (& hold in GCHEM.h); + - add new parameter files: "data.dic" & "data.cfc" ; + - various re-arrangement in initialisation & header file ; + - fix tutorial_global_oce_biogeo for multi-threaded test. +o pkg/dic: test for "pickup_dic" file when restarting ; + if no pickup-file, can restart with a re-initialised pH (same as it did + before) providing pickupStrictlyMatch is set to False. +o change argument list of S/R PTRACERS_WRITE_PICKUP & MYPACKAGE_WRITE_PICKUP + + fix call to S/R GGL90_WRITE_PICKUP & DIC_WRITE_PICKUP +o add a "warning" (to Error msg file) when an Active Tracer is not using AB-2 + and staggerTimeStep is off (needed for stability of Internal-wave dynamics). +o move out of PARAMS.h & data the last pieces of old-ptracer code: tauTr1ClimRelax, + & lambdaTr1ClimRelax ; put those 2 params in PTRACERS_PARAMS.h & data.ptracers + (note: not clear to me what lambdaTr1ClimRelax is still used for ...) +o pkg/mom_vecinv: - new parameter "selectVortScheme" to select which + scheme to use for vorticty-term ; + - compute upwindVorticity=T case within mom_vi_u/v_coriolis_c4.F + (with default selectVortScheme, return to original discretisation) + - remove weird combination (e.g.: upwindVorticity & highOrderVorticity) +o pkg/seaice: add a line to the summary to indicate grid type (B/C) +o in mom_calc_visc, fix average value of viscAh_W & viscA4_W next to the + bottom ; only used with variable horizontal viscosity and non-hydrostatic. + also fix diagnostics VISCAHW & VISCA4W. + +checkpoint59o +o dome/code/obcs_calc.F: avoid division by zero (detected using gfortran). +o change the loop ordering in most gad_*_adv_x.F to improve vectorization: + move a few statemens out the main loop into separate loops +o pkg/shelfice: + - sort out freshwater flux; convert freshwater fluxes to mass fluxes + (kg/m^2/s), does not break tests, because in verification/isomip, + rhoConstFresh=rhoConst (by accident) + - rearrange computation of tLoc, sLoc, pLoc to make code more efficient + - move S/R find_rho_scalar, pkg/ecco/sw_ptmp.F, pkg/ecco/sw_adtg.F into + new file model/src/seawater.F, so that they are available for all pkgs. + - convert potential temperature into in-situ temperature. This changes the + isomip tests => update tests +o add a flag writePickupAtEnd (default=true) to be able to suppress + writing a pickup at the last timestep +o tools/mpack-1.6 : + - disable automake (not working) and autoconf (not needed). + - update to version 1.6.4 which fix MD5 code on some 64 bits platforms + also cleaner with much less compilation warnings. +o pkg/profiles: + rename profiles_interp_mean_genericgrid as profiles_interp_mean_gg + (for "less than 32 characters name" compiler requirement) +o pkg/obcs and pkg/exf: added capability to specify seaice open boundary + conditions that have a different sampling frequency than S/T/U/V obcs +o pkg/diagnostics (stats-diags): + - fix global Min/Max in case 1 processor get only empty tiles +o genmake2: replace (4 times) + make genmake_tc_1.o >> genmake_warnings 2>&1 + with + COMM="$CC $CFLAGS -c genmake_tc_1.c" + echo $COMM >> genmake_warnings + $COMM >> genmake_warnings 2>&1 + to make genmake2 always see the value of CC as specified in the optfile. +o generic_advdiff: - prather advection scheme (SOM) coded for CS-grid. + - add diagnostics for 1srt & 2nd Order moments and sub-grid variance. + - modify verification advect_cs test to test SOM-advection on CS-grid. +o pkg/ctrl: + - Rewrite of ctrl_map_ini_ecco.F, + introducing a generic routine (ctrl_map_ini_gen.F). + - Modification of ctrl_bound/adctrl_bound handling control vector bounds. + +checkpoint59n +o pkg/diagnostics: for large number of diagnostics, store diagnostic mate + number in dedicate array "hdiag": + - new version of S/R DIAGNOSTICS_ADD2LIST : DIAGNOSTICS_ADDTOLIST + with 1 more argument (mate number); modify old version DIAGNOSTICS_ADD2LIST. + - many small changes (integer format) for large number of diagnostics + - start to use DIAGNOSTICS_ADDTOLIST (necessary for diagnostics with + a counter mate) in pkg diagnostics, ptracers, thsice & aim_v23. +o add parameter rotateGrid and Euler angles EulerPhi/Theta/Psi. For + usingSphericalGrid, this allows to define the rotated grid coordinates + via phiMin/thetaMin/dxSpacing/dySpacing, etc., but then re-compute the + geographical coordinates according to the inverse of the rotation defined + by the Euler angles, so that Coriolis parameter and online-interpolation by + exf works with the geographical coordinates on XC/YC/XG/YG. + Naturally, this feature does not work with all packages, so some + combinations are prohibited in config_summary (flt, flt_zonal, ecco, + profiles), because there the coordinates are assumed to be regular + spherical grid coordinates. +o pkg/gmredi,ctrl,ecco,autodiff,grdchk and model/src,inc: + - introduce isopycnal diffusion coefficient control. +o tools/adjoint_options: + - additional control vector components. +o pkg/mnc: have mnc create marginally meaning full coordinate variables + (that is the grid indices) if usingCurvilinearGrid and NOT EXCH2 +o pkg/mypackage: + - add example of read/write pickup for state-var ; + - add entry points to the main code for state-vars diags & write-pickup. +o pkg/exf: rewrite of exf_getffieldrec.F to properly deal with year + transitions for useExfYearlyFields + - add new subroutine that determines the file to read from for + use*YearlyFields = .TRUE. and .FALSE. + - add a verification experiment to global_with_exf that tests + useExfYearlyFields with USE_EXF_INTERPOLATION defined +o pkg/obcs/exf: add useOBCSYearlyFields (=.FALSE.) to obcs with exf in + analogy to useExfYearlyFields; rearrange obcs_prescribe_read: two new + subroutines (which could be generated from a template, because they are + almost identical), for now they are included in obcs_prescribe_read.F for + simplicity +o pkg/exf: + - add a little more diagnostic to exf_interp_read (in analogy to + mds_read_field, can be turned off with debugLevel < debugLevA=1) + - improve vectorizability of exf_interp for TARGET_NEC_SX, makes code + ugly but fast by unrolling short inner loops, inserting a unroll=8 + directive in routine lagran, and splitting a loop and making do while the + outermost loop (the latter requires the intoduction of index arrays) +o pkg/obcs: replace obcs_apply_area/heff/hsalt/hsnow with obcs_apply_seaice +o pkg/seaice: moved DIFFERENT_MULTIPLE from seaice_model to seaice_dynsolver + http://forge.csail.mit.edu/pipermail/mitgcm-devel/2008-January/003173.html +o verification/seaice_obcs: updated for restart test and latest options +o pkg/seaice: add a new flag SEAICEuseTEM to use a slightly different + ice rheology (so-called truncated ellipse, that gets rid of tensile stress); + for now this is only available for the C-grid code and also within + #ifdef SEAICE_ALLOW_TEM, because this option may slow down the EVP code + +checkpoint59m +o pkg/seaice: use different time constant for melting (SEAICE_gamma_t) + and freezing (SEAICE_gamma_t_frz) ice due to mixed layer temperature +o gad_som : fix flux output (& diagnostics) to have same units as other + advection scheme fluxes (was previously multiplied by deltaT). +o added verification/seaice_obcs both to test seaice open boundaries + and to provide an example matlab scrpt for generating them +o added verification/lab_sea/input.salt_plume for testing pkg/salt_plume +o pkg/salt_plume: clean up salt_plume_frac.F to + remove redundancy in PlumeMethod 1 and 4 +o pkg/kpp: added saltplume diagnostics +o pkg/obcs and pkg/seaice: + - do not set uice and vice to zero when OB*icefile are unspecified + - OBCS_SEAICE_AVOID_CONVERGENCE is undefined by default + - moved seaice tracer obcs after call to seaice_growth +o pkg/salt_plume: added various options for salt plume depth + criterion and for vertical distribution of the salt + +checkpoint59l +o pkg/ptracers: new version of ptracers pickup: + read meta file and write only fields which are needed to restart. + ( same logic as main pickup file ; also using pickupStrictlyMatch ) +o pkg/seaice: + - add timescale for melting due to warm mixed layer as a means of + setting SEAICE_availHeatFrac more conveniently + - add a CPP-flag/runtime parameter check for flooding algorithm + - reformulated availHeat in seaice_growth.F to be independent of AREA +o change units attribute of EmPmR netcdf variables to kg/m^2/s to reflect recent + unit changes +o pkg/obcs: added OBCS_SEAICE_AVOID_CONVERGENCE constraint +o cg2d.F: add 3 compiler directives that speed up this routine by 30% on a NEC + SX vector computer. +o pkg/seaice: + - removed SEAICE_FFIELDS.h and seaice_get_forcing.F + seaice forcing fields can now be read only through pkg/exf + - removed some unused halos +o no longer write pickup from THE_MODEL_MAIN but always from FORWARD_STEP +o pkg/exf: added useRelativeWind EXF_PARAM.h option for subtracting U/VVEL + or U/VICE from U/VWIND before computing U/VSTRESS + +checkpoint59k +o pkg/seaice: ICE2WATR=SEAICE_rhoIce/rhoConst for consistency between + dynamics and thermodynamics. Set ICE2WATR=0.92 to reproduce older results. +o fix un-initialised variable problem (in mom_vecinv & shap_filt): + all experiments pass when compile with un-initialised var. checking option. +o added pkg/salt_plume +o script to test restart (for now, in MITgcm_contrib/jmc_script) +o pkg/seaice: new read/write pickup version. + read meta file and write only fields which are needed to restart. + ( same logic as main pickup file ; also using pickupStrictlyMatch ) +o pkg/seaice: fix bug in seaice_ocean_stress (call of seaice_calc_strainrates) +o lab_sea.hb87: modify the domain & update output.hb87.txt : + lat-long grid & no-slip BC in seaice-dynamics produce different sigma12 + at j=1 & j=Ny+1 : this case is not treated correctly in the restart + process. Fixed here by closing the domain at the Northern boundary. +o pkg/rw: cleaning-up: call directly new MDSIO S/R MDS_READ/WRITE_FIELD. +o pkg/mdsio: add new arguments to MDS_READ/WRITE_FIELD for reading/writing + a set of consecutive levels in/from a 3-D array. +o pkg/ptracers: + use a 2 characters string ioLabel to identify tracer (if more than 99 tracers). + to-do: fix pkg/diagnostics when large number of diagnostics are defined. + +checkpoint59j +o in pkgs: dic, obcs, offline, rbcs: + standard way to compute coeff. for time interpolation between 2 time reccords + which retains full precision (& make results less dependent on compiler). + update output of exp: tutorial_cfc_offline & tutorial_global_oce_biogeo. +o pkg/cfc: if compiled with exf and seaice, cfc now can use fields from those + packages instead of reading ice cover and wind speed separately +o split PTRACERS.h in 2 header files: PTRACERS_FIELDS.h & PTRACERS_PARAMS.h + (keep a temporary version of PTRACERS.h in verification/OpenAD/code_ad_openad + to allow to compile customized code before the update) +o cleaning-up: + - fix CPP instruction syntax in pkg/obcs + - replace (some) call to MDSIO S/R with call to RW interface S/R ; and + add #ifdef ALLOW_MDSIO elsewhere (allows to compile without pkg/mdsio) + - fix some threads issue in IO parts. +o pkg/dic: + - add tons of "_d 0" (--> much better testreports !) + and then update outputs of tutorial_global_oce_biogeo +o pkg/mom_fluxform: + add a special version of momentum advection intended to conserve momentum + next to a bathymetry step or a coastline edge. (default = off = #undef) +o new read/write pickup version: + - for now, only implemented for main (mdsio) pickup file. + - write out only what is needed to restart; merge pickup_nh , pickup_ph + and main pickup together. + - read meta file; read only what is needed to restart; stop if a field + is missing (if pickupStrictlyMatch=T, = the default) or try to restart + if it make sense (even if the restart is not perfect). + - if no meta file, assume that pickup to read corresponds exactly to + a pickup that would currently be written. + - read pickup written in the old format (until checkpoint59i) if it + recognizes an old meta file (no field-list). + - add a meta file in exp. which starts from a pickup: 2 cases: + a) minimal meta file (with just file precision) + pickupStrictlyMatch=F + e.g.: global_ocean.90x40x15/input/pickup.0000036000.meta + b) a meta file describing the content of the associated pickup file + e.g.: tutorial_global_oce_biogeo/input/pickup.0004248000.meta +o pkg/mdsio & pkg/rw: + - new routine to read a meta file ; new set of routine to read + Multiple-fields file (with meta file). +o added open boundary conditions capability for seaice + HSNOW, HSALT, UICE, and VICE + +checkpoint59i +o adjoint: + * fix multiDimAdvection adjoint + (partial recomputation problem by TAF for fVerT(:,:,kDown) + and comlev1 size definition (maxpass vs. max. number of tracers) + * verification/global_ocean.90x45x15 adjoint now tests + NLFS and multiDimAdvection using DST3 (=30) + * verification/lab_sea now tests multiDimAdvection using DST3 (=30) +o pkg/kpp: add "_d 0" to real constant (if missing); + update output of exp: natl_box (10.digits); vermix (std, 7.digits) + and lab_sea, 3.FW (only 6,4,6 digits match) & AD (8.digits). +o prepare for "clever pickup" implementation: add new header file "RESTART.h" + for internal parameters related to restart process (formerly from PARAMS.h) + and add startAB parameter to argument list of S/R ADAMS_BASHFORTH2. +o more standard interface to {PKG}_WRITE_PICKUP with pickup suffix passed + as argument. (only atm2d & obcs write_pickup have not yet been updated) +o pkg/monitor: + - use GLOBAL_SUM_TILE instead of _GLOBAL_SUM + - fix mon_stats_rs (statistics over the full domain, even where field is zero) + - account for deep Atmosphere factor and density factor (anelastic) +o added open boundary conditions capability for seaice HEFF and AREA +o pkg/shelfice: add option for convservative form of Hellmer&Olbers (1989) + thermodynamics + - requires a little reorganization, which affects the testreport results + of isomip.htd (only 7 digits of cg2d agree) => update/modify experiment, + so that now the conservative form is tested + - add a few comments +o add missing cvs Header: or Name: in eesupp/*, pkg/* & verification/code* +o add a little bit of code to gad_os7mp_adv_?.F to help vectorization +o add new diagnostics to seaice: advective, diffusive fluxes (only for + multidim-advection, for scheme 2 they are not filled) and transports, + thermodynamic growth rates +o change units of EmPmR (inside the code, + within coupling interface) + from m/s to kg/m2/s (mass flux) but leave input file unchanged. + Affects results @ truncation level (cg2d match) and update output of + experiments: lab_sea(6), global_ocean.90x40x15(10), + global_ocean.cs32x15(11x3,10), natl_box(11), tutorial_global_oce_latlon(11) + + adjoint results: global_ocean_ebm(11) and lab_sea (12). + Coupled set-up: allow to restart from previous pickups (cpl_oldPickup=T). + Next: will retire convertEmP2rUnit (simply using mass2rUnit instead). + +checkpoint59h +o pkg/exf: + - improve vectorization by moving if-statement out of loops in + exf_mapfields, exf_radiation + - exf_bulkformulae.F: introduce 2D fields and vectorize code +o pkg/kpp: added the salt plume scheme (ALLOW_SALT_PLUME) to KPP package +o testreport: + - allow to test for Qnet & SeaIce monitor output; clean-up ALLOW_BULK_OFFLINE + monitor & update offline_exf_seaice exp. output. + - AD testing: pass/FAIL according to Grad. output accuracy (instead of Cost) +o pkg/seaice: + - bug fix in the evp solver: boundary conditions were affected be + masking of zMax/zMin (changes results) + - minor bug fix in boundary conditions in seaice_calc_strainrates + (changes results for no_slip) + - seaice_lsr.F: make interpolation of eta dependent on boundaries + - replace hardcoded lower limit for zeta (zMin) by runtime parameter + - add a test to lab_sea, that tests the Hibler and Bryan (1987) ocean + stress formulation and no-slip boundary conditions +o pkg/ptracers: Temporary fix for PTRACERS_useGMRedi & PTRACERS_useKPP + in gad_calc_rhs.F ; add few safety stops in initialisation phase. +o pkg/seaice: a first HSALT time stepping equation has been checked in +o pkg/seaice: replaced maskW with seaiceMaskU and maskS with seaiceMaskV + in seaice_lsr.F; this solves the pkg/seaice-pkg/obcs problem + without requiring DO-IF statements +o pkg/thsice: + use continuous power-law function for partition of energy between lateral + melting/freezing and thinning/thickening ; old code using step function + still available with #undef THSICE_FRACEN_POWERLAW + update output of exp: aim.5l_cs.thSI, global_ocean.cs32x15.icedyn & .thsice +o testreport: change AD testing : check against results/output_adm.txt + using same scripts as for forward test. + +checkpoint59g +o adjoint testing: in preparation for changing testreport: + - output format: cost and grad written on separated lines + - ad output results in directory "results" (instead of "results_ad") +o pkg/seaice: adding ice salinity HSALT as a prognostic variable +o pkg/seaice: changes for using pkg/seaice with pkg/obcs +o pkg/ecco: allow user defined cost function contributions +o testreport: + Allow to choose (per experiment) the list of tested output variables + and the main variable (used to issue a pass/FAIL ) ; + - for now, done by adding the specific list in file: [exp]/input/tr_checklist + - default is unchanged: >cg2d< , (T,S,U,V,5 tracers)_[min,max,mean,StD] ; + Also modify printing of summary: + a) ">" NN "<" indicates the "main" variable (used for pass/FAIL) + b) echo the list of tested variables in the individual summary.txt of + each experiment. +o new S/R GLOBAL_SUM_TILE to compute global sum independently of number of + threads and number of processes (with #define GLOBAL_SUM_SEND_RECV) + (but still depend on the tile decompostion of the domain). +o pkg/seaice: change a discretization detail to make the B-grid code + "more symmetric". This change is likely to remove all "ridges/leads" that + where quasi-aligned with the grid in B-grid solutions. It may also be + the solution to the stress coupling issues with the B-grid. +o each tested exp: add a "run" directory with a standard ".cvsignore" file ; + remove .cvsignore from input dir. (since testreport no longer run here). +o pkg/thsice: + - cleaning: move common block out of "THSICE_2DYN.h" (removed) to "THSICE_VARS.h" + - fix reshaping of sea-ice just after advection: + melt all the ice only if ice-volume is too small. + - re-visit lateral/thinning repartition of melting (modifs from May 04, 2007): + prevent lateral melting if sea-ice fraction is too close to iceMaskMin ; + and update experiment global_ocean.cs32x15.icedyn (+start from pickup files). +o external_forcing_surf : extend the use of salt_EvPrRn & temp_EvPrRn (fresh-water + tracer content) to linFS or no RealFreshWaterFlux ; default remains unchanged. +o define mass2rUnit & rUnit2mass as units conversion factor, used for surface + forcing. mass2rUnit: from mass per unit area [kg/m2] to r-coordinate unit + (z-coord: =1/rhoConst ; p-coord: =gravity) ; rUnit2mass is just the reverse; + and remove horiVertRatio (& recip_horiVertRatio). Affects truncation error + of exp. global_oce_in_p: 10 16 16 14 16 13 16 16 16 13 13 13 16 13 13 12 16 + -> update output.txt +o pkg/cd_code: add brackets in multiple terms sum (S/R CD_CODE_SCHEME): + this affects truncation error: -> update output.txt of exp: + global_ocean.90x40x15 (fw&adm), lab_sea (fw&adm) & lab_sea.lsr, + ideal_2D_oce, isomip.htd, natl_box, tutorial_global_oce_latlon. + +checkpoint59f +o pkg/autodiff: zero_adj resets an ajoint variable to zero (in adjoint mode) +o pkg/ctrl: ctrl_bound forces parameter corrections to stay in + a bounded range (in forward mode) +o pkg/ecco, ctrl, autodiff and mom_common: + internal parameter controls, completements of implementations +o add few EXCH S/R (UV_BGRID, UV_DGRID, SM) for B & D-Grid velocity & 2nd Moment Sxy. +o use UV_BGRID exch for dxV,dyU: for now, only with pkg/exch2 & forward ; + (not seem any difference in testing results). +o pkg/cd_code: use the right EXCH calls (for CS-grid); but problem remains + at face corners. +o pkg/generic_advdiff: + - in GAD.h, define as parameters (iMinAdvR,iMaxAdvR,jMinAdvR,jMaxAdvR) + loop range indices for computing vertical advection tendency + and do the calculation in the interior only (no overlap) ; + for now, only implemented for SOM advection (to save CPU time). + - add argument "withSigns" to S/R FILL_CS_CORNER_TR_RL (needed for SOM_xy moment) +o exch2 adjoint + First set of changes: + - add exch2 hand-written adjoint templates + - exch2 hand-written no longer use TAF-flag 'nownew_arg' + This will require cleanup in pkg/autodiff/*.F + (changed routines currently kept separate in verif. code/) + - Comment all relevant #ifndef ALLOW_AUTODIFF_TAMC + that used to hide exch2 or cubed-sphere specific code + (commented via 'cph-exch2') +o pkg/seaice + - Close southern and western boundaries for UICE/VICE when useOBCS=.TRUE. +o pkg/bulk_force: fix loading into Qsw of Short-Wave flux; and update + output of global_ocean.cs32x15.thsice . + +checkpoint59e +o exch2: new S/R that only use the cube-tracer (center-position) exchange: + for B-grid vector (not yet tested) and C-grid vector (<- only used + when option W2_USE_R1_ONLY in W2_OPTIONS.h is defined). +o starting implementation of Duffy et al. (GRL 1999) salt plume scheme + (#ifdef ALLOW_SALT_PLUME) +o gather_2d/scatter_2d: removed halo regions +o gmredi: add option (GM_taper_scheme='fm07') for Ferrari & McWilliams 2007 + scheme and the corresponding test case (under front_relax dir). For now, + only available with Skew-flux form and not yet tested in realistic set-up. + adjoint: store directives updated, but gradient checks deteriorated, + so maybe GM/Redi adjoint is now broken. +o seaice adjoint + * add AREA, HEFF, HSNOW as control variable + (should not change lab_sea results) + * CTRL_OPTIONS.h needs new flag: CTRL_SET_OLD_MAXCVARS_40 +o pkg/seaice: + * added computation of saltFlux in seaice_growth; + For time being, sea ice salinity is assumed constant + contribution of snow flooding to freshening of sea ice is neglected. + SEAICE_salinity is set in SEAICE_PARM01 of data.seaice and it + defaults to 0.0 in order to maintain backward compatibility. + Typical values for SEAICE_salinity are 4 to 10 g/kg. + * replaced dRf(1) with dRf(kSurface)*hFacC(kSurface) in seaice_growth.F + * added SEAICE_availHeatFrac, which specifies fraction of surface level + heat content used to melt or grow ice + * commented out cap of HICE: it helps keep ice thinner + * reintroduced cap of HEFF via MAX_HEFF and SEAICE_CAP_HEFF option + * applied OBCS T/S mask to HEFFM sea ice mask + +checkpoint59d +o standard name: {PKG}_DIAGNOSTICS_STATE to fill diagnostics array + for {PKG} state variables, and called from DO_STATEVARS_DIAGS. + (keep {PKG}_DIAGNOSTICS_FILL for non-state variables diagnostics) +o pkg/gmredi (gmredi_readparms.F): add some missing "_d 0" + and change default value of parameter "GM_Small_Number"; + update output of exp: lab_sea (x2), cfc_example, global_ocean.90x40x15, + global_ocean.cs32x15 (x4), global_ocean_ebm, global_with_exf, + tutorial_global_oce_biogeo, tutorial_global_oce_latlon +o pkg/ecco: cost function when applying pkg/smooth to controls +o pkg/ctrl: add forcingPrecond run time parameter +o pkg/seaice: + - replace ZMIN=4.e8 with ZMIN=0. in dynsolver.F as in C-grid code + - fix bug in metric terms of u-equation in LSR solver (implicit part), + changes verification/lab_sea + - rearrange implicit part of u-equation in LSR solver to make it easier to + read, understand, and debug. This causes changes at the truncation level. + - fix (serious) bug in the boundary conditions (sign error in the implicit + part) of the LSR solver. Changes all LSR results. + - fixed problem with HSNOW initialization to avoid thick snow when AREA + is initilized with small values - does not change lab_sea results. + - added HsnowFile initialization capability +o pkg/smooth: addition to model/src and pkg/ctrl so allow + initialization of pkg/smooth and application to control vector +o pkg/seaice: fix another bug in the LSR no-slip boundary conditons +o dependencies: pkg/profiles requires pkg/cal +o pkg/profiles: + - added option ALLOW_PROFILES_GENERICGRID + -> pkg/profiles can thus be used with any grid (e.g. cs32x15) assuming + that interpolation information (grid points and coefficients) is provided + within the netcdf input files. + -> sample matlab scripts can be found in MITgcm_contrib/gael + that allow you to prepare netcdf input files. + - also modified: cleaner include statements, cleaner error print statements, + cleaner stops when error, and more tests of the netcdf input files consistency. +o additional test (in global_ocean.cs32x15) using seaice+thsice pkgs. +o pkg/seaice: + - move seaice-diagnostics from seaice_do_diags to do_statevars_diags; this + requires a new routine: seaice_diagnostics_fill.F + - make sure that seaice does not move if not forced (pressure replacement + method, Hibler and Ip, 1995), changes results + - and new damping criterion for EVP solver according to Hunke, JCP, 2001 + - SEAICE_evpDampC = > 0. turns it on (615. is the recommended value). + It is turned off by default (=-1.), because it changes the results. + - make it possible to read AREA from a file, too. + +checkpoint59c +o add S/R and diagnostics for oceanic mixed-layer depth +o move GMRedi call after all vertical mixing schemes (including KPP) +o pkg/thsice: + - ensure iceMask > iceMaskMin : this problem appeared after May-04 modifs + (melt only laterally if thin ice): update output of global_ocean.cs32x15.thsice o pkg/obcs: - bug fix for balance code o pkg/seaice: @@ -18,12 +601,18 @@ - no longer depend on tapering scheme; instead, compute slope (limited by GM_maxSlope). this is equivalent to what was comuted before with clipping or linear tappering. - - fix vertical averaging (wrong by 1 level) + - fix vertical averaging (wrong by 1/2 level) - leave the old version (but will not be maintained) within #ifdef OLD_VISBECK_CALC / #endif + - update output of exp. ideal_2D_oce o change error catch in calc_r_star to improve vectorization o exf runoff and seaice: remove ALLOW_SEAICE from exf pkg files and add #define ALLOW_RUNOFF in SEAICE_OPTIONS.h (when not using exf). +o adjoint: + Re-organized adjoint checkpointing according to Matt Mazloff + (but with modifs/completions/cleanups) + Storing is now bundled into large arrays to optimized I/O + (and second step in cleanup of the_main_loop) checkpoint59b o coupler (pkg/atm_ocn_coupler, pkg/ocn_compon_interf, pkg/atm2d): @@ -77,7 +666,7 @@ o pkg/thsice: fix few little problems, in thsice_calc_thickn.F: - growth vertically (and not laterally) if iceFrac == iceMaskMax - - melt laterally only if hIce < hThinIce (as the comments say) + - melt only laterally if hIce < hThinIce (as the comments say) and in thsice_extend.F: - allow to form ice even when iceFrac == iceMaskMax (by increasing thickness) - start to form ice as soon as the minimum ice-volume is reachable. @@ -114,7 +703,9 @@ ATMOSPHERIC_LOADING is #define). o pkg/kpp: replaced _KPP_RL with _RL o bug fixes for vertical diffusivity computations when both KPP and - 3D diffusivity arrays are used. + 3D diffusivity arrays are used; pkg/kpp/kpp_calc_diff_s & _t.F + are no longer updating the vertical diffusivity array + (passed as argument) but are (re-) setting those arrays. o pkg/seaice: fix bug in evp solver, while doing that - change (improve?) discretization of stress computations for both cgrid-lsr and evp solver