4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
5 |
============================== |
============================== |
6 |
|
|
7 |
o pkg/ecco : |
o verification, AD & TLM output: |
8 |
|
- new TAF version 2.3.10 (previously 2.3.8) add more parenthesis in multiple |
9 |
|
additions and multiplications which affect results at machine truncation |
10 |
|
level. Update few AD output (1D_ocean_ice_column, bottom_ctrl_5x5, |
11 |
|
lab_sea.noseaicedyn & offline_exf_seaice) and TLM output |
12 |
|
(1D_ocean_ice_column & tutorial_tracer_adjsens.som81) which had only 12 |
13 |
|
digits of matching. |
14 |
|
o pkg/seaice: |
15 |
|
modify lsr-algorithm so that it vectorizes AND converges by using a |
16 |
|
"zebra" method: split vector loop into two with step size 2; this |
17 |
|
solves a general convergence problem for the JFNK solver on a vector |
18 |
|
computer. It has the potential to even improve convergence for |
19 |
|
non-vector machines, but it does change the results so it is turned |
20 |
|
off by default. To turn on define CPP-flags SEAICE_VECTORIZE_LSR and |
21 |
|
SEAICE_VECTORIZE_LSR_ZEBRA in SEAICE_OPTIONS.h |
22 |
|
|
23 |
|
checkpoint64e (2013/03/05) |
24 |
|
o pkg/ecco and several verification experiments: |
25 |
|
Remove special versions of the_model_main.F, the_main_loop.F, forward_step.F |
26 |
|
from ecco package, and revert to unified routine in model/src/ |
27 |
|
Adjust verification experiments accordingly, especially gloal_oce_llc90 |
28 |
|
Should conclude long-standing work by Gael to reconsile the_main_loop with |
29 |
|
ECCO and will enable OpenAD's revolve loop in ECCO setup. |
30 |
|
o pkg/generic_advdiff: |
31 |
|
- implement compressible flow method for multi-dim advection (similar to |
32 |
|
gad_som_advect.F); add new option "GAD_MULTIDIM_COMPRESSIBLE" (since |
33 |
|
TAF generates many recomputations) to use it; no yet coded with implicit |
34 |
|
vertical advection. |
35 |
|
- change experiment advect_cs (new output) and advect_xy (same output) |
36 |
|
to test compressible flow multi-dim advection. |
37 |
|
o pkg/generic_advdiff: |
38 |
|
- fix SOM advection for unusual cases (linear Free-Surf in r*, other |
39 |
|
cases where linear Free-Surf is not @ k=1, e.g., useShelfIce) |
40 |
|
o pkg/seaice: |
41 |
|
- fix bug in the no-slip implementation in seaice_calc_strainrates.F, this |
42 |
|
bug affects two verification experiments (lab_sea and lab_sea.hb87), |
43 |
|
recover old (wrong) results by setting noSlipFac = 2. in |
44 |
|
seaice_calc_strainrates.F; |
45 |
|
- add another averaging method for etaZ (etaZmethod=3), a mix of 2 and 0 |
46 |
|
- add runtime parameter SEAICE_JFNKepsilon for that finite difference |
47 |
|
gradient in the Jacobian times vector operation (s/r seaice_jacvec) |
48 |
|
- fix line search in seaice_jfnk.F (s/r seaice_jfnk_update) |
49 |
|
o pkg/seaice: reorganize seaice_lsr.F to be able to reuse more code: |
50 |
|
- new subroutines seaice_lsr_tridiagu/v |
51 |
|
- introduce loop bounds i/jmin/max to be able to use a restricted |
52 |
|
additive Schwarz method, but not used for now. |
53 |
|
- rename SEAICE_CALC_LSR_COEFFS into SEAICE_LSR_CALC_COEFFS |
54 |
|
- use new subroutines seaice_lsr_tridiagu/v in s/r seaice_preconditioner |
55 |
|
o OpenAD support |
56 |
|
Reorganize S/R call for overall time-stepping loop to facilitate |
57 |
|
OpenAD revolve loop replacement |
58 |
|
(new S/R main_do_loop.F) |
59 |
|
o model/src: |
60 |
|
- add 3-D array 3rd dimension to argument list of S/R ADAMS_BASHFORTH2/3; |
61 |
|
- pkg/cheapaml: remove local S/R ADAMS2d and use instead standard S/R |
62 |
|
ADAMS_BASHFORTH2. |
63 |
|
o pkg/cheapaml: |
64 |
|
- add new parameter (default=F, same as before) to disable full domain |
65 |
|
(including land) periodicity in X & Y dir. |
66 |
|
- new S/R CHEAPAML_COPY_EDGES to copy relaxation field into edges of |
67 |
|
prognostics field ; merge the 2 S/R cycle_2d_tracer and timestep_2d_tracer |
68 |
|
into single S/R CHEAPAML_TIMESTEP. |
69 |
|
|
70 |
|
checkpoint64d (2013/02/17) |
71 |
|
o pkg/monitor (angular momentum): |
72 |
|
- do not add Eta contribution to the total AM if using rigid-lid; |
73 |
|
- correct for Adams-Bashforth in zonal-wind part (instead of in AM_eta). |
74 |
|
- horizontal discretisation: compute Zonal wind contribution separately for |
75 |
|
each component (instead of from cell centered averaged) and use the 4 |
76 |
|
grid-cell corner averaged value of Omega*(R*cos_Phi)^2 in AM_eta part. |
77 |
|
- Check that AM is conserved with linear dynamics (momAdvection=F and linear |
78 |
|
FreeSurf) if using vectorInvariant momentum. |
79 |
|
o model/src (horizontal grid): |
80 |
|
- move initialisation of all horizontal grid arrays to null value |
81 |
|
from specific horizontal grid S/R to main S/R INI_GRID; |
82 |
|
- add 2 new arrays to project model C-grid velocity on to zonal direction |
83 |
|
(similar to AngleCosC & -AngleSinC but at U and V point location) and |
84 |
|
compute them in S/R CALC_GRID_ANGLES. |
85 |
|
o pkg/seaice: replace a hard coded parameter by a runtime parameter |
86 |
|
SEAICE_JFNK_tolIter |
87 |
|
o pkg/seaice: introduce a restricted additive Schwarz (RAS) method in |
88 |
|
preconditioner for JFNK (run LSR into the overlaps as far as possible, |
89 |
|
i.e OLx/y-2, but overwrite the solution in the overlaps) |
90 |
|
- extend loop range in seaice_calc_viscosity |
91 |
|
- add new parameters SEAICE_OLx/y (default = OLx/y-2) |
92 |
|
- add flexible loop ranges to seaice_preconditioner, seaice_calc_lsr_coeffs |
93 |
|
- slightly modify loops in tridiagonal solver lsr-preconditioner |
94 |
|
- by default remove the extra exchange in seaice_preconditioner with a |
95 |
|
replacement of an undocumented CPP flag by SEAICE_PRECOND_EXTRA_EXCHANGE |
96 |
|
- modify offline_exf_seaice so that old results without RAS are reproduced |
97 |
|
for output.dyn_jfnk.txt |
98 |
|
in seaice_fgmres.F (unrelated to RAS) |
99 |
|
- add a cpp-flag option for a re-ordered mapping (potentially more |
100 |
|
efficient), but because this changes the results, turn it off by |
101 |
|
default |
102 |
|
- replace the save statement by local common blocks |
103 |
|
o utils/matlab: |
104 |
|
- fix tile-offset in rdmnc.m when output was produced using EXCH2. |
105 |
|
- update cs_grid/bk_line scripts to use load_grid.m (in utils/matlab) |
106 |
|
instead of local script "load_cs.m". |
107 |
|
o pkg/seaice: |
108 |
|
- fix a serious bug in seaice_fgmres.F, s/r seaice_scalprod, connected |
109 |
|
to the scalarProductMetric and the unrolling. By chance this does not |
110 |
|
change the verification experiment, because of the particular domain |
111 |
|
decomposition |
112 |
|
o pkg/diagnostics: |
113 |
|
- change default missing_value from UNSET_RL to -999. |
114 |
|
o pkg/layers: improve vectorization of layers_fluxcalc.F |
115 |
|
- step 1: split loops, make TatU/V 2D arrays. This alone reduces the cpu |
116 |
|
time spent in this routine by more than a factor of 3 on an NEC-SX8 |
117 |
|
- step 2: replace search algorithm by a bisection algorithm found in |
118 |
|
Press et al. Numerical Recipes in Fortran. This algorithm can be |
119 |
|
vectorized and finds the correct bin in approx. log2(Nlayers) steps. |
120 |
|
For now, use this algorithm only for TARGET_NEC_SX defined. |
121 |
|
- moved the search algorithm into a separate subroutine, that is kept |
122 |
|
within the same file for automatic inlining. |
123 |
|
- modified catching potential errors |
124 |
|
o pkg/ecco: |
125 |
- add compile switches ALLOW_SHALLOW_ALTIMETRY and ALLOW_HIGHLAT_ALTIMETRY. |
- add compile switches ALLOW_SHALLOW_ALTIMETRY and ALLOW_HIGHLAT_ALTIMETRY. |
126 |
o model/src, pkg/autodiff, pkg/ctrl, pkg/grdchk : |
o model/src, pkg/autodiff, pkg/ctrl, pkg/grdchk : |
127 |
- modifications for ALLOW_GENTIM2D_CONTROL |
- modifications for ALLOW_GENTIM2D_CONTROL |
128 |
-- forward_step.F, the_main_loop.F : pass options and arrays for adjoint tapes. |
-- forward_step.F, the_main_loop.F : pass options and arrays for adjoint tapes. |
129 |
-- load_fields_driver.F : add call to CTRL_MAP_GENTIM2D( myTime, myIter, myThid ) |
-- load_fields_driver.F : add call to CTRL_MAP_GENTIM2D( myTime, myIter, myThid ) |