source: LMDZ4/branches/LMDZ4-dev-20091210/libf/dyn3dpar/abort_gcm.F @ 3241

Last change on this file since 3241 was 1222, checked in by Ehouarn Millour, 15 years ago

Changes and cleanups to enable compiling without physics
and without ioipsl.

IOIPSL related cleanups:

  • bibio/writehist.F encapsulate the routine (which needs IOIPSL to function)

with #ifdef IOIPSL flag.

  • dyn3d/abort_gcm.F, dyn3dpar/abort_gcm.F and dyn3dpar/getparam.F90: use ioipsl_getincom module when not compiling with IOIPSL library, in order to always be able to use getin() routine.
  • removed unused "use IOIPSL" in dyn3dpar/guide_p_mod.F90
  • calendar related issue: Initialize day_ref and annee_ref in iniacademic.F (i.e. when they are not read from start.nc file)

Earth-specific programs/routines/modules:
create_etat0.F, fluxstokenc.F, limit_netcdf.F, startvar.F
(versions in dyn3d and dyn3dpar)
These routines and modules, which by design and porpose are made to function with
Earth physics are encapsulated with #CPP_EARTH cpp flag.

Earth-specific instructions:

  • calls to qminimum (specific treatment of first 2 tracers, i.e. water) in dyn3d/caladvtrac.F, dyn3d/integrd.F, dyn3dpar/caladvtrac_p.F, dyn3dpar/integrd_p.F only if (planet_type == 'earth')

Interaction with parallel physics:

  • routine dyn3dpar/parallel.F90 uses "surface_data" module (which is in the physics ...) to know value of "type_ocean" . Encapsulated that with #ifdef CPP_EARTH and set to a default type_ocean="dummy" otherwise.
  • So far, only Earth physics are parallelized, so all the interaction between parallel dynamics and parallel physics are encapsulated with #ifdef CCP_EARTH (this way we can run parallel without any physics). The (dyn3dpar) routines which contains such interaction are: bands.F90, gr_dyn_fi_p.F, gr_fi_dyn_p.F, mod_interface_dyn_phys.F90 This should later (when improving dyn/phys interface) be encapsulated with a more general and appropriate #ifdef CPP_PHYS cpp flag.

I checked that these changes do not alter results (on the simple
32x24x11 bench) on Ciclad (seq & mpi), Brodie (seq, mpi & omp) and
Vargas (seq, mpi & omp).

EM

  • Property svn:eol-style set to native
  • Property svn:keywords set to Author Date Id Revision
File size: 1.1 KB
Line 
1!
2! $Id: abort_gcm.F 1222 2009-08-07 11:48:33Z idelkadi $
3!
4c
5c
6      SUBROUTINE abort_gcm(modname, message, ierr)
7     
8#ifdef CPP_IOIPSL
9      USE IOIPSL
10#else
11! if not using IOIPSL, we still need to use (a local version of) getin_dump
12      USE ioipsl_getincom
13#endif
14      USE parallel
15#include "iniprint.h"
16 
17C
18C Stops the simulation cleanly, closing files and printing various
19C comments
20C
21C  Input: modname = name of calling program
22C         message = stuff to print
23C         ierr    = severity of situation ( = 0 normal )
24
25      character (len=*) :: modname
26      integer ierr
27      character (len=*) :: message
28
29      write(lunout,*) 'in abort_gcm'
30#ifdef CPP_IOIPSL
31c$OMP MASTER
32      call histclo
33      call restclo
34      if (MPI_rank .eq. 0) then
35         call getin_dump
36      endif
37c$OMP END MASTER
38#endif
39c     call histclo(2)
40c     call histclo(3)
41c     call histclo(4)
42c     call histclo(5)
43      write(lunout,*) 'Stopping in ', modname
44      write(lunout,*) 'Reason = ',message
45      if (ierr .eq. 0) then
46        write(lunout,*) 'Everything is cool'
47      else
48        write(lunout,*) 'Houston, we have a problem ', ierr
49      STOP
50      endif
51      END
Note: See TracBrowser for help on using the repository browser.