Hello,
I have been running the ProteinInterfaceMS mover with rosettascripts. It runs successfully, though one single design seems to take about 5 days to complete. This is on one cpu. I was wondering if ProteinInterfaceMS mover is capable of making use of MPI? If so how does it split the calculations? And how would you submit the command?
Thanks
Category:
Post Situation:
Unfortunately, the ProteinInterfaceMS mover is intrinsically single threaded, and you can't spread the genetic algorithm across multiple processors with MPI or threading. (At least for a single run through. If you're asking about having multiple -nstruct or multiple input structures, those can be split trivially or with MPI. But for a single run-through, it's strictly single threaded.)
If you are looking for an MPI-capable multistate design protocol, you might want to look at the mpi_msd application: https://www.rosettacommons.org/docs/latest/application_documentation/design/mpi-msd and https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0020937 -- this approach does distribute the genetic algorithm out across multiple processors.