wiki:IPSLCM5_v5

Version 22 (modified by jgipsl, 12 years ago) (diff)

--

New set up of the coupled model of IPSL. This configuration contains possibility to run IPSLCM5A, IPSLCM5B or IPSLCM5_C and all smaller configurations such as LMDZOR.

Working branch of modipsl

A new branch of modipsl is created to test new configurations. Extract as follow :

svn co http://forge.ipsl.jussieu.fr/igcmg/svn/modipsl/branches/newconfig modipsl

New fonctionalities in this branch :

  • mod.def : contains the configurations IPSLCM5_v5, LMDZOR_v5, LMDZREPR_v5, LMDZORINCA_v5 in v5 family.
  • model : possibility to change a model component version. See revision 1733.

Extract IPSLCM5_v5 configuration

The new configuration name IPSLCM5_v5 is not yet known in the script fait_config. Use following workaround before extracting the configuration :

  • Modify mod.def at ligne 71, from IPSLCM5_v5 to IPSLCM5A, to have following line :
    #-C- IPSLCM5_v5  CONFIG/UNIFORM/v5/IPSLCM5_v5  HEAD         8  IPSLCM5A       config
    
  • Modify model ligne 419 from IPSLCM5A to IPSLCM5_v2, to have following :
      ( IPSLCM5_v5 )
      ../modeles/UTIL/fait_config IPSLCM5A ;;
    

LMDZ5 can also be used, in that case change the line for extraction of LMDZ into :

#-C- IPSLCM5_v5  LMDZ5/trunk        1575        11  LMDZ       modeles

Contents of the configuration directory IPSLCM5_v4

> ls
AA_make
AA_make.ldef
scripts
GENERAL
USER_SPEC

The AA_make* files and the directory scripts are used for the compiling. These are the same as in IPSLCM5A. There are no differences in compiling compared to IPSLCM5A.

For the running part, there are 2 directories USER_SPEC and GENERAL. The directory GENERAL contains the sub-directories POST, PARAM and DRIVERS. The files in GENERAL are used for all types of experiments using this configuration. USER_SPEC contains different sub-directories with all cards needed for 5 proposed experiments with IPSLCM5.

> ls USER_SPEC
LMDZ
LMDZOR
IPSLCM5

> ls USER_SPEC/IPSLCM5
Chistorical
CpiControl
EXP00
historical
piControl

For one experiment, a config.card and a directory COMP exist. The COMP directory contains all components card for the experiment. For example the experiment historical :

> ls USER_SPEC/IPSLCM5/historical
config.card
COMP

> ls USER_SPEC/IPSLCM5/historical/COMP
lim2.card
lmdz.card
oasis.card
opa9.card
orchidee.card
pisces.card
stomate.card


How to create a job and run directory

Combining the GENERAL directory with one of the sub-directories in USER_SPEC makes a complete run directory. The ins_job contains the creation of such a run directory.

Choose in USER_SPEC the experiment to be done and copy the corresponding config.card to modipsl/config/IPSLCM5_v4/. Modify the config.card as usual, for at least the JobName. Apply ins_job and the creation of a new run directory will be done :

  • the run directory will be created with the same name as JobName in config.card. If a directory with the same name already exists, nothing will be done.
  • the directory GENERAL will be copied into the new run directory
  • one of the COMP directories in USER_SPEC will be copied into the new run directory, according to the variable UserSpec in config.card
  • the config.card will be copied into the run directory
  • a job will be created in the run directory

Once the run directory has been created with ins_job, this directory is used as a normal run directory. For example it is possible to remove the job, change the config.card inside the directory and re-create the job with ins_job. To create a second run directory, modify the upper config.card, in IPSLCM5_v4 directory, with at least the JobName, and re-run ins_job. If a directory with the JobName already exist, nothing will be done using this config.card.

DRIVER directory

Another difference with origin configuration structure (such as IPSLCM5A) is the separation between cards and drivers. All cards (expect config.card) are kept in COMP directory and all drivers are in a new DRIVER directory. To use this structure the tag libIGCM_v1_12 or later is needed.

Remaining questions

How to make clear that a variable in a parameter file might be edited by the drivers ?

  • If the variable always is edited, it is then possible to put in parameter file var=_AUTO_
  • But if a variable only is some cases is modified, then a default value is needed. A possibility might be to decide a syntax for example var=_AUTO_:X where X is the default value. Then before starting the execution all variables that still keeps _AUTO_, the X value must be adapted...

New functionnality

Ensemble functionnality : ensemble section into config.card and ensemble.card and ensemble driver. More information : wiki:Ensembles

To do list

Avant fin mai

  • Améliorer ins_job (Sébastien D.)
    • Ajouter dans ins_job option pour MPI/OpenMP (Arnaud)
  • MPI/OpenMP : enlever les parametres obsolètes de config.card liés à l'execution. (Arnaud)
  • Ajouter dans mod.def les nouveaux configurations
  • Changer pour ORCHIDEE sous SVN (Josefine)
  • Changer pour OASIS sous SVN (Arnaud)
  • Moyenne résolution :
    • Mettre les fichiers poids sur les comptes commune. Vérifier qu'on pointe sur les bons fichiers pour les différents résolutions. Documenter. (Olivier ?)
    • Vérifier que les paramètres sont bons pour la moyenne résolution. pmagic aussi!
    • A faire à la main ou automatiser :
      • Il faut aussi changer TagName : IPSLCM5A-MR
      • Il faut réduire les tailles des Chunck dans les POST
  • Revoir les variables "auto" dans les namelist NEMO. Dans le driver, éventuellement utiliser le même type de _sed que dans orchidee.driver et lmdz.driver.
  • Validation entre IPSLCM5_u1 et IPSLCM5A
  • Documentation dans modispslforBeginners1
  • Adapter libIGCM pour Curie. Fait dans tag libIGCM_v2_beta1
  • Mettre le fichier Bands en restart.
  • Avoir un fichier de sortie texte par processeur par défaut
  • Séparer les environnements par défaut : calcul et post-traitement
  • passage à netcdf4?

Plus tard

  • Changer environnement de compilation pour NEMO : utiliser NEMO v3_4 avec fcm et allocation dynamique pour le nombre de processus
  • Amélioration dans la gestion des diagnostiques: choix des fichiers, fréquences, time-series,..., simplification et documentation
  • Déterminer si souhaitable d'ajouter NEMO et PISCES offline (Chrisitan)
  • Ajoute IPSLCM5INCA_u1 (Anne)

Attachments (1)

Download all attachments as: .zip