L6·Expert

🧵Threads, Processes & the GIL

Concurrency vs parallelism in CPython.

threadingmultiprocessingGIL

§ 1The GIL

CPython's Global Interpreter Lock allows only one thread to execute Python bytecode at a time. Threads are still great for I/O.

§ 2Threads

threading.Thread(target=fn).start() — cheap, share memory, fine for I/O-bound.

§ 3Processes

multiprocessing bypasses the GIL by spawning separate interpreters. Best for CPU-bound.

Example

1from concurrent.futures import ThreadPoolExecutor
2import time
3
4def slow(i):
5    time.sleep(0.2)
6    return i * i
7
8t = time.perf_counter()
9with ThreadPoolExecutor(max_workers=5) as pool:
10    print(list(pool.map(slow, range(5))))
11print(f"{time.perf_counter()-t:.2f}s")  # ~0.2s not 1.0s

Try it live · Python runs in your browser

Loading playground…

Checkpoint quiz

Question 1 / 2

For CPU-bound Python work, prefer:

async / await & Event Loop

Metaclasses