You can have (non transparent, but supporting the buffer interface, enough for eg numpy arrays) shared memory with the current multiprocessing stuff: https://do

You can have (non transparent, but supporting the buffer interface, enough for eg numpy arrays) shared memory with the current multiprocessing stuff: https://docs.python.org/3/library/multiprocessing.shared_memory.html https://docs.python.org/3/library/multiprocessing.shared_mem... A tangent but I find it amusing to contrast the perpetual Python GIL debate with all the new computation platforms that claim to be focused on scalability. Those are mostly single threaded or max out at a few virtual CPUs (eg "serverless" platforms) and there people applaud it. There people view the isolation as supporting scalability.