Parallel execution

If a parallel version of ADF has been installed you should be aware of a few special aspects of running ADF in parallel. Partially this depends on the platform and on the installation settings.

First of all, you may specify (by command-line options in the run-script and/or by defining suitable environment variables) explicitly how many parallel processes are to be used. Secondly, you should realize that most of the files that you would have in a single-node run are in a parallel run distributed over the parallel processes. Some parts of the file may be identical across the processes while other parts are not and would only after a recombination yield the data of the corresponding single-node file. The normal result files, (standard output, the logfile and the binary result file TAPE21) are complete at the master process.