Libraries like numpy, scipy, and polars utilize tools like BLAS or rayon which are often implicitly multithreaded for things like matrix decomposition. However this poses an issue when used in the context of a multiprocessing environment like we use, where each Spyral process might now try to spawn its own independent BLAS threads and you'll quickly run out of resources. We can use tools like environment variables or threadpoolctl to disable these features.
Libraries like numpy, scipy, and polars utilize tools like BLAS or rayon which are often implicitly multithreaded for things like matrix decomposition. However this poses an issue when used in the context of a multiprocessing environment like we use, where each Spyral process might now try to spawn its own independent BLAS threads and you'll quickly run out of resources. We can use tools like environment variables or threadpoolctl to disable these features.