The multiprocessing scheduler I wrote wasn't at the time written to support passing data in memory.
I think this scheduler offers considerable benefit as a totally lightweight and transparent scheduler (without all the heuristics and complexity of dask and ray). This scheduler might continue to be of use for testing/debugging/small projects.
To that end, supporting the passing of data in memory by some means would be beneficial followed by testing of the implementation.
The multiprocessing scheduler I wrote wasn't at the time written to support passing data in memory. I think this scheduler offers considerable benefit as a totally lightweight and transparent scheduler (without all the heuristics and complexity of dask and ray). This scheduler might continue to be of use for testing/debugging/small projects. To that end, supporting the passing of data in memory by some means would be beneficial followed by testing of the implementation.
Issues