Hi all,
I'm running the docking_protocol application by mpirun, and I want to save silent results to individual directories determined by process ID.
I found the option "-out:path:mpi_rank_dir" may be helpful but it didn't work. How do I use this option properly?
Category:
Post Situation:
Most Rosetta protocols (including docking_protocol) are trivially parallelizable. That is, there's no real difference between running a single (MPI) run of 10,000 structures versus 10 runs of 1000 structures each. Each output structure is independent of any others. In some respects, the thing MPI gives you is the coordination of output such that you can put things into a single file.
The -mpi_rank_dir option comes from an older job distribution system, one which is not used by docking_protocol. Given the limited usefulness, it really hasn't been ported to the newer running systems.
What's the reason you're looking for MPI and separate output files? Is there a problem with letting MPI do the single silent output? If you need the multiple outputs, is there a reason why a non-MPI run doesn't work for you? I don't know your particular use case, so I don't know if there might be a workaround to accomplish what you want.
I want to do this just because I'm afraid the output file will be too big to read (about 3G for 20k structures), so I'm going to run 10 separate runs or more. Thanks a lot.
If filesize is an issue, note that you can get Rosetta to automatically read/write gzipped silent files by using a `.gz` extension on the filename