/[MITgcm]/MITgcm/doc/tag-index
ViewVC logotype

Diff of /MITgcm/doc/tag-index

Parent Directory Parent Directory | Revision Log Revision Log | View Revision Graph Revision Graph | View Patch Patch

revision 1.2161 by mlosch, Mon Jun 15 15:48:41 2015 UTC revision 1.2192 by jmc, Fri Oct 23 12:31:57 2015 UTC
# Line 4  $Name$ Line 4  $Name$
4      Notes on tags used in MITgcmUV      Notes on tags used in MITgcmUV
5      ==============================      ==============================
6    
7    checkpoint65p (2015/10/23)
8    o pkg/seaice/seaice_evp.F: fix store directives and key computations, since
9        there is no adjoint test for EVP, we don't know what it does to the adjoint
10        solution, but at least there are no more recomputation warnings left
11    o pkg/seaice/seaice_evp.F: slight modification of averaging so that EVP stands
12        a chance of giving the same results as a fully converged VP (JFNK) solver
13    o tools/adjoint_options/adjoint_diva:
14      - fix bug (typo) in Tang-Linear FLAGS setting for MPI + DIVA
15    o pkg/streamice:
16      - spotted bug in streamice_driving_stress using dyG where should have been dxG
17    o pkg/exf:
18      - remove STOP statement that prevented to use wind or (wind-)stress
19        interpolation for constant-in-time forcing fields (wind/stress period = 0).
20    o pkg/seaice:
21      - replace SEAICE_EVP_USE_ADAPTATION by run time parameters, requires
22        5 additionals 2D fields and 6 new store directives
23      - change logic in seaice_readparms.F: EVP code can now be turned on in
24        various ways, setting SEAICE_deltaTev is no longer required,
25        alternatively set SEAICE_evpAlpha/Beta, SEAICEuseEVPstar, SEAICEuseEVPrev,
26        or SEAICEaEVPcoeff
27    o pkg/shelfice/shelfice_thermodynamics.F
28      - redefining uLoc/vLoc to u/v, instead of 2u/v
29    o pkg/seaice/seaice_evp.F: introduce "adaptive EVP" by Madlen Kimmritz
30      - this is an intermediate state of a truely converging EVP that has
31        the potential of outperforming implicit VP methods, I do not recommend
32        using it yet.
33      - for now, "adaptive EVP" is turned on by setting SEAICE_EVP_USE_ADAPTATION
34        and all parameters are hardwired
35      - for convergence it still requires very smooth regularisations, that means,
36        if SEAICE_DELTA_SMOOTHREG is defined, deltaCreg=deltaC+deltaMin, which is
37        different from the VP case with deltaCreg=sqrt(deltaC**2+deltaMin**2); also
38        specific averaging appears to be required, still subject to tuning
39      - still needs cleaning and I need to turn parameter into runtime parameters
40      - add code to compute and print residuals of iteration, compile with
41        defining ALLOW_SEAICE_EVP_RESIDUAL to enable it
42    o pkg/seaice/seaice_evp.F: fix a bug (factor of four in front of e12Csq)
43      that entered with r1.54 on Sep04, 2015, -> update results
44    
45    checkpoint65o (2015/09/14)
46    o pkg/mom_vecinv & mom_common:
47      - add deep-atmosphere geometry factor in mom_calc_visc.F + in routines
48        called from there.
49      - add anelastic and deep-atmosphere geometry factor in pkg/mom_vecinv ; this
50        allows to use Vector-Invariant form in deep atmos and anelastic formulation
51    o pkg/seaice: introduce fast ice parameterisation following Itkin et al. (2015)
52      and Koenig-Beatty+Holland (2012)
53      - two new parameters SEAICE_tensilFac and SEAICE_tensilDepth
54      - global 2D field tensileStrength, computed in seaice_calc_ice_strength
55      - requires an extra input argument for S/R seaice_calc_viscosities
56      - tensileStrength's depth dependence is different from Itkin et al. (2015)
57        (to be changed and played with, for now exponential)
58    o pkg/seaice: introduce new parameter SEAICEscaleSurfStress (default = .FALSE.)
59      - if TRUE scale ice-ocean and ice-atmosphere stress acting on sea ice
60        by AREA according to Connelly et al. (2004), JPO.
61      - For EVP and the LSR solver, the implicit part of the drag term is
62        the only non-zero term in the denominator when there is no ice (which
63        prevented running the model with zero ice-ocean drag). If
64        SEAICEscaleSurfStress = .TRUE., this results in a division by zero
65        (or zero main diagonals BU/BV) which need to be caught. A practical
66        consequence is that for open water, the momentum equation reduce
67        to trivially 0 = 0 + 0 (for EVP). For LSR, BU/BV are reset to 1, if
68        they happen to be zero, often leading to a non-zero solution over
69        open water.
70    o pkg/seaice: revise EVP code (seaice_evp.F) in preparation for a more
71        efficient EVP method.
72      - introduce area-weighted averages for e12 (analogously to
73        seaice_calc_viscosities.F); this totally changes lab_sea.hb87 (2 digits
74        remain, but also changing the order of terms in the previous average
75        changed the results as much) -> update experiment
76      - add SEAICE_DELTA_SMOOTHREG code
77      - disentangle computation of zetaC/deltaC and zetaZ/deltaZ, also for
78        TEM-code
79      - adjust loop bounds so that only required fields are computed
80      - add new averaging code for zetaZ/deltaZ (again similar to
81        seaice_calc_viscosities.F), inactive by default for now (will change
82        results again), but will replace old code eventually
83    o eesupp/src:
84      - add an other alternative for multi-proc GLOBAL_SUM (in global_sum_tile.F)
85        to always cumulate tile local-sum in the same order
86        (option #define GLOBAL_SUM_ORDER_TILES in CPP_EEOPTIONS.h).
87        Still uses MPI_Allreduce (like standard code) but on array of tiles
88        and should be faster than GLOBAL_SUM_SEND_RECV code.
89      - switch almost all verification experiments that use GLOBAL_SUM_SEND_RECV
90        (defined in CPP_EEOPTIONS.h_mpi) to GLOBAL_SUM_ORDER_TILES instead.
91      - add CPP_EEOPTIONS.h_mpi with GLOBAL_SUM_ORDER_TILES defined in 14 more exp.:
92         aim.5l_cs, cheapAML_box, deep_anelastic, dome, fizhi-cs-32x32x40,
93         fizhi-cs-aqualev20, fizhi-gridalt-hs, global_ocean_ebm, matrix_example,
94         natl_box, solid-body.cs-32x32x1, tutorial_advection_in_gyre,
95         tutorial_baroclinic_gyre and tutorial_held_suarez_cs.
96    
97    checkpoint65n (2015/07/29)
98    o pkg/thsice:
99      - add a warning when used with pkg/aim but without updating SST
100        (stepFwd_oceMxL=F & useCoupler=F & tauRelax_MxL<>-1 ).
101    o pkg/offline:
102      - change where Offline-Fields are loaded, from load_fields_driver.F to
103        directly within FORWARD_STEP, just after updating iter number and time;
104      - move setting of GM and KPP diffusivity out of OFFLINE_FIELDS_LOAD into new
105        S/R OFFLINE_GET_DIFFUS that is called towards the end of DO_OCEANIC_PHYS;
106       These changes allow:
107        a) to recover exactly on-line results using deltaToffline=deltaTClock,
108        tested (using offlineIter0= nIter0 -1 and offlineTimeOffset= -deltaT/2)
109        both with Lin & Non-Lin Free-Surf, both with staggerTimeStep & synchronous
110        time-step, both with GM & KPP files and with recomputing these fields;
111        b) no needs to switch on staggerTimeStep anymore;
112        c) simplify rescaling of offline horiz. velocity with Non-Lin Free-Surf.
113      - update output of experiment tutorial_dic_adjoffline (only 3 digits left).
114    o pkg/offline:
115      - change (rarely used) offset-iteration parameter to an offset time (in s):
116        can easily recover previous offset-iter effect by setting offset-time
117        (offlineTimeOffset) to offset-iter x deltaToffline;
118      - do not turn-off Temp & Salinity surface relaxation when using KPP (used
119        in KPP as part of surface buoyancy forcing) in offline run.
120      - fix ptracers KPP non-local term in the case where KPP is re-computed.
121    o OpenAD
122      - Adding preliminary ggl90 test to OpenAD verification.
123        Config. with ggl90 currently diverges after 64 time steps
124        so will need more work to become more stable
125      - Adding preliminary kpp test to OpenAD verification.
126        required small tweakes in kpp_routines.F
127      - Merge and update from Krishna Narayanan's contrib area:
128        * Add one variable for preliminary OpenAD DIVA
129        * Add replacement time-stepping for OpenAD DIVA
130        * genmake2 flag -diva (but only for OpenAD; preliminary)
131        * required modifs for OAD_support
132        successful preliminary test for tutorial_global_oce_biogeo/
133    o pkg/seaice:
134      - initialise deltaC in seaice_lsr, cosmetic changes in seaice_lsr
135      - add cpp-flag SEAICE_DELTA_SMOOTHREG for option of regularising
136        delta with a smooth function in s/r seaice_calc_viscosity
137        (no effect on EVP)
138      - rename local suffix "sqr" into "sq" for more consistent variable names
139    o pkg/seaice:
140        add new parameter SEAICE_deltaMin that is used ***only***
141        for regularizing Delta (and nothing else, like the parameter
142        which is also used for all sorts of things SEAICE_EPS).
143        Defaults to SEAICE_EPS for backward compatibility only.
144    o pkg/seaice: replace erroneously storing e12Csqr on the wrong tape with
145        the wrong key by inialising it before use -> fixes a recomputation
146        problem but does not fix the gradient
147    o pkg/layers:
148      - move resetting of flux arrays to outside the iLa loop
149        (this caused layers_diapycnal to not work with more than one layer coord.)
150      - fix bug in calculation of layers_recip_delta in layers_init_fixed.F
151    o pkg/layers:
152      - remove from common block 3D layers fields (used to diagnose transport
153        in tracer coords) and declare them as local arrays in layers_calc.F
154  o pkg/seaice:  o pkg/seaice:
155    - Rename SEAICE_VECTORIZE_LSR_ZEBRA to SEAICE_LSR_ZEBRA and define it    - Rename SEAICE_VECTORIZE_LSR_ZEBRA to SEAICE_LSR_ZEBRA and define it
156      in lab_sea forward experiment (changes results -> updated)      in lab_sea forward experiment (changes results -> updated)
157  o pkg/seaice: modify seaice_lsr.F in order to improve the gradient    - update adjoint experiment lab_sea after cleaning up seaice_lsr_tridiagu/v
158    computations; for SEAICE_VECTORIZE_LSR  o pkg/seaice: modify seaice_lsr.F in order to improve the gradient
159       computations; for SEAICE_VECTORIZE_LSR
160    - move the loops over which the tridiagonal solvers (seaice_lsr_tridiagu/v)    - move the loops over which the tridiagonal solvers (seaice_lsr_tridiagu/v)
161      sweep to the outside,      sweep to the outside,
162    - remove store directives and add "CADJ loop sequential" directives    - remove store directives and add "CADJ loop sequential" directives
163      in analogy to model/src/solve_tridiagonal.F      in analogy to model/src/solve_tridiagonal.F
164    - replace many "#ifdef SEAICE_VECTORIZE_LSR_ZEBRA" by variable loop    - replace many "#ifdef SEAICE_VECTORIZE_LSR_ZEBRA" by variable loop
165      boundaries to yield more readable code. This has also the charming side      boundaries to yield more readable code. This has also the charming side
166      effect that your can use SEAICE_VECTORIZE_LSR_ZEBRA without      effect that your can use SEAICE_VECTORIZE_LSR_ZEBRA without
167      SEAICE_VECTORIZE_LSR (i.e. adjoint without recomputations in      SEAICE_VECTORIZE_LSR (i.e. adjoint without recomputations in
168      seaice_lsr_tridiagu/v now requires either of these flags, vectorization      seaice_lsr_tridiagu/v now requires either of these flags, vectorization
169      still requires SEAICE_VECTORIZE_LSR with SEAICE_VECTORIZE_LSR_ZEBRA as      still requires SEAICE_VECTORIZE_LSR with SEAICE_VECTORIZE_LSR_ZEBRA as
170      an additional option)      an additional option)
171    The tridiagonal solvers are now completely analoguous to solve_tridiagonal.F     The tridiagonal solvers are now completely analoguous to solve_tridiagonal.F
172    but the adjoint gradients (which are affected by this change) still explode.     but the adjoint gradients (which are affected by this change) still explode.
173  o coupler (atm & ocn _compon_interf):  o coupler (atm & ocn _compon_interf):
174    - avoid using Exch2 IO mapping if not suitable.    - avoid using Exch2 IO mapping if not suitable.
175    
 >>>>>>> 1.2160  
176  checkpoint65m (2015/06/15)  checkpoint65m (2015/06/15)
177  o pkg/layers:  o pkg/layers:
178    - clean-up old parameters setting (from Sep 2012):    - clean-up old parameters setting (from Sep 2012):
Legend:
Removed from v.1.2161  
changed lines
  Added in v.1.2192
  ViewVC Help
Powered by ViewVC 1.1.22