I'm learning about multithreading in Python, and am experimenting to see how GIL(Global Interpretor Lock) works. Given the following python code: def cpu_bound(): i = 0 for _ in range(10000000): i += 1 With the above code, my computer runs in 0.45 seconds. In joblib.Parallel(n_jobs=8, prefer="threads"), I create 12 threads to execute this code. As far as I know, in python there is GIL so actually only 1 python thread is running at a time. If so, to run all 12 threads will take at least 0.45 * 12 = 5.4 seconds (not including the time for context-switching, creating threadpool), but in reality it only takes 3.25 seconds to run 12 threads. Please help me explain the reason. Continue reading...
