| 4 |
Notes on tags used in MITgcmUV |
Notes on tags used in MITgcmUV |
| 5 |
============================== |
============================== |
| 6 |
|
|
| 7 |
|
o multi-threading and IO: |
| 8 |
|
- replace some call to MDSIO S/R by the corresponding READ/WRITE_FLD S/R. |
| 9 |
|
- remove most of BEGIN_MASTER/ END_MASTER around READ/WRITE_FLD S/R calls |
| 10 |
|
(needed to work with SingleCpuIO): tested in all verif. exp. |
| 11 |
|
o GAD: use simpler DST3 formulation as default (old version left within #ifdef) |
| 12 |
|
affects the results (but none of the verification test uses this scheme). |
| 13 |
|
o Adding some pkg/diagnostics capability to pkg/seaice |
| 14 |
|
(will certainly be subject to changes) |
| 15 |
|
|
| 16 |
|
checkpoint58q_post |
| 17 |
|
o added CPP_EEOPTIONS.h flag DISABLE_MPI_READY_TO_RECEIVE. Speeds up |
| 18 |
|
1/6-deg Southern Ocean configuration by 8% on IBM Datastar and |
| 19 |
|
high_res_cube configuration by 4% on Columbia. |
| 20 |
|
o clean-up multi-threaded problems (reported by debugger tcheck on ACES); |
| 21 |
|
add (in global_with_exf/code) a version of EXF_INTERP_READ without |
| 22 |
|
dynamical allocation (seems to work in multi-threaded). |
| 23 |
|
o new function "MASTER_CPU_IO" (=T if master thread & master CPU, if using |
| 24 |
|
single CPU io) so that TAF does not have to see EESUPPORT.h |
| 25 |
|
o pkg/seaice: implement sea-ice loading (when using RealFreshWaterFlux) |
| 26 |
|
|
| 27 |
|
checkpoint58p_post |
| 28 |
|
o obcs: stop if obcs is used with useCDscheme = .true. |
| 29 |
|
o shelfice: |
| 30 |
|
- fix a sign error so that no slip at shelfice interface works |
| 31 |
|
- remove stop for combination non-hydrostatic code and shelfice. That does |
| 32 |
|
not mean, that the results are correct, but at least they are numerically |
| 33 |
|
stable |
| 34 |
|
o diagnoctics: modify UVELPHI and VVELPHI to include hfac (i.e. mass-weighted) |
| 35 |
|
o added adjoint exch routined for all exch*3d* routines |
| 36 |
|
and updated flow directives |
| 37 |
|
|
| 38 |
|
checkpoint58o_post |
| 39 |
|
o replace calls to EXCH_Z & EXCH_UV_AGRID: use the new _3D_ version ; |
| 40 |
|
remove old subroutines (_XY_ & _XYZ_); update the adjoint version. |
| 41 |
|
o change controling S/R for writing pickup file (new S/R do_write_pickup) |
| 42 |
|
and clean-up reading/writing of pickup files. |
| 43 |
|
o fix local version of read/write pickup (NH part) in rotating_tank/code |
| 44 |
|
o simplify EXCH interface: |
| 45 |
|
- change _XYZ_ interface to _3D_ subroutine (with 3rd dim in argument list) |
| 46 |
|
- not often used EXCH S/R (exch_z, exch_uv_agrid): keep only the _3D_ version |
| 47 |
|
o added the regrid package |
| 48 |
|
o shelfice: |
| 49 |
|
- add a simple boundary layer scheme to reduce noise |
| 50 |
|
- fix the indexing in shelfice_u/v_drag so that there will be actually drag |
| 51 |
|
|
| 52 |
|
checkpoint58n_post |
| 53 |
|
o fix globalArea multi-threaded calculation: need to apply GLOBAL_SUM to |
| 54 |
|
local variable (not in common block). |
| 55 |
|
o lot of small changes for multi-threaded. Most tests (using OMP & ifort) pass. |
| 56 |
|
- still problems with pkgs: mnc, exp2, dic, cal, exf (and probably more); |
| 57 |
|
- still BEGIN_MASTER/END_MASTER to remove around READ/WRITE_FLD S/R calls |
| 58 |
|
to work with SingleCpuIO. |
| 59 |
|
o add "_d 0" to real constant of pkg/cfc subroutines ; |
| 60 |
|
update exp. cfc_exemple & tutorial_cfc_offline output files. |
| 61 |
|
o multi-threading and IO: |
| 62 |
|
- all threads call READ_FLD S/R (needed if using SingleCpuIO): |
| 63 |
|
(no BEGIN_MASTER/ END_MASTER around the call, but inside MDSIO S/R. |
| 64 |
|
- barrier before and after reading. |
| 65 |
|
- stop if other than master is calling MDS_WRITELOCAL (if #undef IO_THREAD_SAFE) |
| 66 |
|
- avoid using MDS_WRITELOCAL for LAND pkg output. |
| 67 |
|
- new S/R AIM_WRITE_PHYS (replaces AIM_WRITE_LOCAL) to write AIM physics |
| 68 |
|
common-block variables ; Allows multi-threading with master-thread IO. |
| 69 |
|
o change default MAX_NO_THREADS from 32 to 4 ; |
| 70 |
|
and stop cleanly when MAX_NO_THREADS is too small. |
| 71 |
|
o add S/R BAR_CHECK to check multi-threaded barrier synchronization |
| 72 |
|
|
| 73 |
|
checkpoint58m_post |
| 74 |
|
o add a Non-Hydroatatic test (with 3-D Coriolis, NHMTerms, biharmonic Visc. |
| 75 |
|
and side-drag) on CS-grid : global_ocean.cs32x15.viscA4 |
| 76 |
|
o make "tutorial_global_oce_latlon" from standard "global_ocean.90x40x15" exp. |
| 77 |
|
keep global_ocean.90x40x15 (for AD testing) with former secondary test |
| 78 |
|
(".qshyd") as the main test (mv output.qshyd.txt to output.txt) |
| 79 |
|
o CVS modules: update MITgcm_verif_basic; remove MITgcm_verif_ocean & _verif_atmos; |
| 80 |
|
add MITgcm_tutorials (= code + all the tutorials) |
| 81 |
|
o ecco, option to compute cost for total ssh (mean+anom) |
| 82 |
|
o consistent with addition of ptracer and ssh in pkg/profiles |
| 83 |
|
o profiles: adding ptracer and ssh |
| 84 |
|
o testreport: change multi-test output-file name |
| 85 |
|
(from output.txt.{name} to output.{name}.txt) |
| 86 |
|
o profiles: now ok for mutli-tiles case |
| 87 |
o seaice: add 4 parameters for different advection schemes for area, |
o seaice: add 4 parameters for different advection schemes for area, |
| 88 |
(effective) seaice thickness, enthalphy, and snow height. Their default |
(effective) seaice thickness, enthalphy, and snow height. Their default |
| 89 |
is the standard parameter SEAICEadvScheme |
is the standard parameter SEAICEadvScheme |
Parent Directory
| 
Revision Log
| 
Revision Graph
| 
Patch