Running TELEMAC on multiple platforms links from Python for TELEMAC


Tools to run TELEMAC on multiple platforms : runcode.py

runcode.py


Arguments available under runcode.py are listed with runcode.py -help.
They are :

  • -c CONFIGNAME (alternatively configname=CONFIGNAME)
  • -f CONFIGFILE (alternatively configfile=CONFIGFILE)
  • -r ROOTDIR (alternatively rootdir=ROOTDIR)
  • -s (alternatively sortiefile)
  • -t (alternatively tmpdirectory)
  • -x (alternatively compileonly)
  • -w WDIR (alternatively workdirectory=WDIR)
  • jobname=JOBNAME
  • queue=HPC_QUEUE
  • walltime=WALLTIME
  • email=EMAIL
  • hosts=HOSTS
  • ncsize=NCSIZE
  • nctile=NCTILE
  • ncnode=NCNODE
  • sequential
  • mpi
  • split
  • merge
  • run
  • use-link


-c allows the user to specify the configuration to be used from the configurations given in the configuration file. Failing that, if the configuration file does not refer to a specific configuration, one will be selected arbitrarily.


-f allows the user to specify a different configuration file (./configs/systel.cfg or the file given by the environment variable SYSTELCFG are considered by default). The full name of the file (i.e. including path) is then required.


-r allows the user to specify a different root directory (generally corresponding to a different version of the TELEMAC-MASCARET system).


-s prints the run progress and useful information to a text file.


-t preserves the temporary directory.


-x if specified,the source code will be compiled and the executable created but the model will not be run.


-w allows the user to specify the temporary directory to run into. This saves splitting the input files again when the same number of processors is used in a parallel run.


jobname assigns a jobname to help track the run on HPC.


queue specifies the HPC queue where the model is to be run.


walltime specifies a walltime (real time, not CPU time) for the job corresponding to the maximum time it should take. If this limit is exceeded, the HPC queue manager will stop the job.


email defines the list of users (email addresses) to which the HPC queue manager has to send mail, if any, about the job.


hosts if specified, the job will only run on oneor several of the hosts. ';' delimited


ncsize gives the number of processors to be used for the job.


nctile gives the number of nodes to be used for the job.


ncnode specifies the number of cores to be used on each node (cannot be greater than the number of cores in a node). This can be useful if memory requirements for the job are high.


sequential if specified, submit the jobs in arguments (list of steering files) sequentially. This can be useful if memory requirements for a job are high and no more than one job is allowed on a node at any one time.


mpi (internal command).


split if specified, will only prepare and split the input files but will not run the job.


merge if specified, will merge the different contributions to the job output into one result file. merge usually follows a run command.


run if specified, will only run the multi-processor job but will not recombine the output into one result file at the end of the job. run usually follows a split command.


use-link allows the user to setup symbolic links to the input files rather than making copies of the files in the temporary working folder. Available on linux platforms only.


It is possible to use combinations of the above. For example:
runcode.py telemac2d -s t2d_test.cas ncsize 36 nctile 3 -c hydry queue highp jobname test email=Joe.Bloggs@hrwallingford.com

python/runcode_in_parallel.txt · Last modified: 2015/07/28 16:19 by n.durand