How to avoid repeated data movement between processes?

[Michael_Eastwood]

@Chong_Wang

The problem is that each time a task starts, a large object (nm) is moving from local process to remote process, which costs some unnecessary computation time. For the simple example above, a SharedArray can be used. However, in my program, a quite complex type is assigned to nm. In this case, how to avoid this overhead?

Try using a CachingPool: https://docs.julialang.org/en/stable/stdlib/parallel/#Base.Distributed.CachingPool