process origin nodes in parallel?

[knaaptime]

thanks for this! I finally got around to playing with an example. I noticed that running preprocess doesnt actually destroy the CPU like i thought it might. I wonder if it would make sense to split the origin nodes by n_cpus, then run the dijkstra function on each chunk in parallel using joblib? It might be a little taxing on memory since you'd still need to send all edge weights and destination nodes to each core, but still seems like it could be worth a shot. Curious what you think

[fscottfoti]

Yeah seems like that would work. I think I was considering doing the parallelization inside the for loop which I think numba doesn't support, but it does seem like it would work in a multiprocessing situation.