Changes between Version 2 and Version 3 of IncaCiclad

Timestamp:

01/22/20 16:39:52 (4 years ago)

Author:

xuezhou

Comment:

--

IncaCiclad

@@ -1 @@
-= Lancer une simulation LMDZORINCA sur CICLAD =
-Les tests lmdzorinca sur Ciclad sont faits avec l'aide des informations sur le site :  [https://forge.ipsl.jussieu.fr/igcmg_doc/wiki/Doc/ComputingCenters/ESPRImesocenter]
+LMDEZORINCA tests at Ciclad are done with the help of information on the site :  [https://forge.ipsl.jussieu.fr/igcmg_doc/wiki/Doc/ComputingCenters/ESPRImesocenter]
 
-Comme sur Irène, le répertoire de soumission est généré par le script ins_job. Un seul MPI est utilisé, par conséquence il faut demander plus de mémoire. 
-Par exemple, pour un test AER (96*95*19) avec 48MPI, 1OMP, nous demandons :
-#PBS -l mem=60gb
-#PBS -l vmem=300gb 
+First of all, the environment should be defined by adding 
+{{{
+. /home/igcmg/.atlas_env_ciclad_ksh
+}}} in {{{${HOME}/.bashrc}}}\\
+
+= Compilation at Ciclad =
+The LMDEZORINCA can be extracted as usual. At Ciclad, ifort is the default compiler. The models LMDZOR_v6 and ORCHIDEE have been tested successfully. Therefore only in INCA model arch-ifort_CICLAD.fcm and  arch-ifort_CICLAD.path should be added.
+Here is a example: 
+arch-ifort_CICLAD.fcm:
+{{{
+%COMPILER            /usr/lib64/openmpi/1.4.5-ifort/bin/mpif90 
+%LINK                /usr/lib64/openmpi/1.4.5-ifort/bin/mpif90
+%AR                  ar
+%MAKE                gmake
+%FPP_FLAGS           -P -traditional
+%FPP_DEF             CPP_FLUSH
+%BASE_FFLAGS         -i4 -r8 -fp-model precise  
+%PROD_FFLAGS         -O3
+%DEV_FFLAGS          -g -O2 -traceback -fp-stack-check -ftrapuv -check bounds
+%DEBUG_FFLAGS        -fpe0 -O0 -g -traceback -fp-stack-check -ftrapuv -check bounds -check all
+%MPI_FFLAGS
+%OMP_FFLAGS          -openmp -openmp-threadprivate compat
+%BASE_LD             -i4 -r8
+%MPI_LD
+%OMP_LD              -openmp
+}}}
+
+arch-ifort_CICLAD.path:
+
+{{{
+NETCDF_INCDIR="-I/opt/netcdf42/ifort/include"
+NETCDF_LIBDIR="-L/opt/netcdf42/ifort/lib"
+NETCDF_LIB="-lnetcdf -lnetcdff -lstdc++"
+
+IOIPSL_INCDIR="-I${INCADIR}/../../lib -I${INCADIR}/../IOIPSL/inc"
+IOIPSL_LIBDIR="-L${INCADIR}/../../lib -L${INCADIR}/../IOIPSL/lib"
+IOIPSL_LIB="-lioipsl"
+
+XIOS_INCDIR="-I${INCADIR}/../XIOS/inc"
+XIOS_LIBDIR="-L${INCADIR}/../XIOS/lib"
+XIOS_LIB="-lxios"
+
+ORCH_INCDIR="-I${INCADIR}/../../lib"
+ORCH_LIBDIR="-L${INCADIR}/../../lib"
+}}}
+
+A pre_proc_ifort_CICLAD.x file should also be added, which can copy pre_proc_X64_IRENE.x and just modifier inca_pp.x64 by inca_pp.ciclad.
+ 
+As at other servers, the script ins_make is used to generate makefiles. 
+In Makefile, for xios : --netcdf_lib netcdf4_seq is added. 
+{{{
+xios :
+        (cd  ../../modeles/XIOS ; ./make_xios  \
+--netcdf_lib netcdf4_seq \
+--prod --arch ${FCM_ARCH} --job 8 ; cp bin/xios_server.exe ../../bin/. ; )
+}}}
+
+P.S. Without this line, it seems to work too. 
+
+= Run a simulation LMDZORINCA at CICLAD =
+As at Irène, a submit directory is created by launching ins_job. At Ciclad only one OMP and one MPI for xios, therefore more memory is required. 
+For instance, for a test AER (96*95*19) with 48MPI and 1OMP #PBS -l mem=60gb #PBS -l vmem=300gb can be set.