Responsible: Jenni Deadline: late June, but only after T1.2 is completed.

• Investigate parallelization strategies using OpenMDAO and MPI. [WE]
  • Two levels of parallelization is identified, namely;
    1. Multistart / random initial parameters
    2. Parallelization of the wake model, e.g. 1 processor per wind direction
  • Implement multistart
  • Fuga is so fast that MPI is found not to be feasible.
  • There are still possibilities for speed-up of the code which is more beneficial than implementing MPI. (Milestone 1.1)
• Consider if the α version can be parallelized in a simple manner using Python’s build-in multiprocessing module. [GW]
  • N/A