В каких случаях Python не удерживает глобальный интерпретаторный замок (GIL) при выполнении операций?

The Global Interpreter Lock (GIL) in Python prevents multiple threads from executing bytecode simultaneously within a single process, which limits multithreading in CPU-bound tasks. However, the GIL is released during operations that free it, typically input/output operations or computations implemented at the C level, which can run in parallel.

Examples of such cases:

• Input/output operations (file reading/writing, network requests)
• Calls to functions from C extensions that explicitly release the GIL (e.g., in NumPy, SciPy libraries)
• Long computations implemented outside the interpreter (e.g., via Cython with GIL release)

Thus, during blocking or long-running operations that do not require the Python interpreter, the GIL is released, allowing other threads to work in parallel.