Python GIL impact

benchmark.md

GIL https://wiki.python.org/moin/GlobalInterpreterLock

Benchmark

5 threads:

python3 ./multithreading.py                               
Execution time with GIL:  20.465982913970947
Execution time without GIL:  4.411532878875732

10 threads:

python3 ./multithreading.py                               
Execution time with GIL:  43.809110164642334
Execution time without GIL:  6.819617748260498

multithreading.py

import time
import threading

NUM_PROCESSES = 10
NUM_THREADS = 10

def count_down(n):
    while n > 0:
        n -= 1


def run_test():
    threads = []

    for i in range(NUM_THREADS):
        # Create two threads to execute the count_down function
        threads.append(threading.Thread(target=count_down, args=(100000000,)))
        # Start the threads
        threads[i].start()

    for i in range(NUM_THREADS):
        # Wait for both threads to finish
        threads[i].join()       

if __name__ == '__main__':
    # Measure the execution time with the GIL
    start_time = time.time()
    run_test()
    end_time = time.time()
    execution_time_with_gil = end_time - start_time
    print("Execution time with GIL: ", execution_time_with_gil)

    # Measure the execution time without the GIL using multiprocessing
    import multiprocessing

    multiprocessing.freeze_support()
    start_time = time.time()

    processes = []

    for i in range(NUM_PROCESSES):
            # Create two threads to execute the count_down function
            processes.append(multiprocessing.Process(target=count_down, args=(100000000,)))
            # Start the threads
            processes[i].start()

    for i in range(NUM_PROCESSES):
        # Wait for both threads to finish
        processes[i].join()

    end_time = time.time()
    execution_time_without_gil = end_time - start_time
    print("Execution time without GIL: ", execution_time_without_gil)