/[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 1.1259 by jmc, Thu Apr 29 13:56:40 2010 UTC revision 1.1270 by jmc, Mon May 17 02:32:49 2010 UTC
# Line 4  $Name$ Line 4  $Name$
4      Notes on tags used in MITgcmUV      Notes on tags used in MITgcmUV
5      ==============================      ==============================
6    
7    o eesupp/src (exch-1):
8      - Fix AD EXCH code for 2-D domain special cases (Nx=1 or Ny=1).
9      - Separate buffer filling and MPI sending: allow EXCH-1 to work for local
10        array (non-shared) when using MPI+MTH. Also reduces number of BARRIER
11        (even without using MPI).
12      - Message mode: move RecvAck setting (indicator of buffer being ready)
13        from send_put to recv_get S/R (was useless before, but not sure if
14        it's much more useful now);
15      - switch the order of sync: MPI-proc 1rst and then threads;
16      - take out spin-waiting code (#undef EXCH_USE_SPINNING), use BARRIER instead.
17    o model/src:
18      initialise dPhiHydX,dPhiHydY inside calc_grad_phi_hyd.F (so that TAF does
19      not drop it in S/R mddynamics, producing NaN in hs94.1x64x5 AD-test exp).
20    o utils/matlab/rdmnc.m: adapt to be able to use the generic matlab netcdf API
21    o pkg/ctrl: require a specific set of control variables and exf-flags for
22      adjoint of seaice (stop in ctrl_check.F if conditions are not satisfied)
23    o model/src:
24     a) 1rst version of a new driver S/R to turn-off output freq and flags after the
25       1rst forward sweep, to avoid over-writing output files during recomputations
26       recomputations (& grdchk); called from cost_final.F (& ctrl_init_ctrlvar.F ?)
27     b) 2nd version:
28      - add specific pkg S/R for seaice & ptracers (called from S/R TURNOFF_MODEL_IO);
29      - switch-off more output flags (exf_monFreq, chkPtFreq, dumpInitAndLast,
30         writePickupAtEnd) + seaice & ptracers output;
31      - do not switch off MNC flags (otherwise, won't get admonitor netcdf file)
32    o model/src:
33      - new S/R rotate_uv2en_rl (..._rs) that rotates vector fields.
34    o pkg/exch2:
35      - remove "simulationMode" from arg. list of S/R EXCH2_RX1,2_CUBE ;
36      - add argument "signOption" to EXCH2_RX1_CUBE (will be needed for SM exch).
37    o pkg/autodiff & exch2:
38      process exch_uv_agrid_3d_rx.f & exch2_uv_agrid_3d_rx.f through TAF
39      (for now, only if ALLOW_EXCH2 is defined).
40    o pkg/autodiff:
41      - include PACKAGES_CONFIG.h in exch*ad.flow & exch*ad.F
42        (not supposed to be in CPP_OPTIONS.h) to get ALLOW_EXCH2 defined correctly.
43       + clean-up comments (_RS is not always real*4)
44      - fix bug in ADEXCH_UV_3D_RX (without EXCH2, was calling the forward exch S/R)
45      - add a stop in ADEXCH_UV_AGRID_3D_RX if trying to use it with EXCH2:
46        was calling a forward exch2, which is wrong ; used in SEAICE & EXF pkgs.
47    o testreport (AD-tests):
48      no longer link files from dir "../input" (1rst link files from
49      "../input_ad.[xx]" if secondary test "xx" and then from "../input_ad")
50    
51  checkpoint62f (2010/04/29)  checkpoint62f (2010/04/29)
52  o model/src/the_main_loop.F:  o model/src/the_main_loop.F:
53    adjust store directives for seaice evp and multicategory according to    adjust store directives for seaice evp and multicategory according to
54    pkg/ecco/the_main_loop.F    pkg/ecco/the_main_loop.F
55  o adjoint:  o adjoint:
56    global_sum_ad was incomplete (noted many times by many people ;)    global_sum_ad was incomplete (noted many times by many people ;)
57    nice description by Jean U. as follows:    nice description by Jean U. as follows:
58    * allreduce(var) in global_sum.F can be interpreted as    * allreduce(var) in global_sum.F can be interpreted as
59      reduce(sum) followed by broadcast of sum.        reduce(sum) followed by broadcast of sum.
60    * The adjoint of that is    * The adjoint of that is
61      sum(adjoint) followed by broadcast(summed adjoint value)      sum(adjoint) followed by broadcast(summed adjoint value)
62      or in other words, allreduce(sum, adjoint vars)      or in other words, allreduce(sum, adjoint vars)
63    We were missing the sum(adjoint)!    We were missing the sum(adjoint)!
64    Benign, except(!) for "global" scalars appearing in tile loops (bi,bj)    Benign, except(!) for "global" scalars appearing in tile loops (bi,bj)
Legend:
Removed from v.1.1259  
changed lines
  Added in v.1.1270
  ViewVC Help
Powered by ViewVC 1.1.22