What does the underlying implementation of worker in SCOOP? Multi-threading or multiprocessing?

Question

I am trying to use SCOOP on a single machine to accelerate genetic algorithm, as shown in DEAP documentation page. But I found out that the computational efficiency is not t