🏭 Python Multiprocessing Mastery

What a process actually is, why it's the only way past the GIL for real CPU parallelism, and how isolated memory changes everything — read this first.

Create, start, and join a Process — and the __main__ guard you can't skip.

The modern way. submit, map, Futures — and the one-line swap from threads.

No shared memory means you pass messages. Queue and Pipe for IPC.

When copying is too slow: Value, Array, shared_memory, and Manager.

The classic worker pool: map, imap, starmap, and why chunksize matters.

Cross-process Lock, Event, Semaphore, and Barrier — and why they're heavier than threads'.

fork vs spawn vs forkserver — the source of half of all multiprocessing bugs.

Everything crossing a process boundary is pickled. What can't cross, and how to make it cheap.

How a child crash surfaces, terminate vs graceful stop, and signals across processes.

The reason you're here: real speedups on CPU work, chunking, and Amdahl's hard limit.

Why one GIL per process is the whole point, and what free-threaded Python 3.13+ changes.

The decision guide from the process side: when the serialization tax is worth paying.

multiprocessing vs concurrent.futures vs joblib vs subprocess vs Dask/Ray — what each is for.