CPython concurrency in CPU-intensive applications

Python is an interpreted language, depending on the underlying protocol there are many versions, the most common is the C-based CPython, by default we call Python is CPython.

Python's Gil (Global Interpreter Lock):

Used to resolve data integrity and state synchronization between multithreading, so that no matter how many CPUs the thread is distributed on, the interpreter only allows 1 threads to run at the same time.

So Python is thread_safe. In fact, Python is almost only single-threaded programming.

CPU Intensive Applications: (frequent calculation)

Starting from 1 accumulate, to 100 million

1 def foo (): 2     num = 03for in      Range (100000001):4         num + = i

Now do a task, complete three functions foo () run, calculate the total time, give four kinds of scenarios

1, single thread, 1,2,3,4 serial

2, multi-threading, 1,2,3,4 parallel

3, single process, 1,2,3,4 serial

4, multi-process, 1,2,3,4 parallel

Experiment to verify the best way to implement Python concurrency under CPU cheats

To test your computer's CPU:

intel_i7-4700 (4-Core 8 threads)

Programme 1;

1 if __name__=="__main__":2T_list = []3Start_time =time.time ()4      forIinchRange (5):5i = Threading. Thread (target=foo)6 t_list.append (i)7 I.start ()8 I.join ()9End_time =time.time ()Ten     Print("totally time:{}". Format (End_time-start_time))

Programme 2;

1 if __name__=="__main__":2T_list = []3Start_time =time.time ()4      forIinchRange (5):5i = Threading. Thread (target=foo)6 t_list.append (i)7 I.start ()8      forIincht_list:9 I.join ()TenEnd_time =time.time () One     Print("totally time:{}". Format (End_time-start_time))

Programme 3;

1 if __name__=="__main__":2P_list = []3Start_time =time.time ()4      forIinchRange (5):5i = multiprocessing. Process (target=foo)6 p_list.append (i)7 I.start ()8 I.join ()9End_time =time.time ()Ten     Print("totally time:{}". Format (End_time-start_time))

Programme 4;

1 if __name__=="__main__":2P_list = []3Start_time =time.time ()4      forIinchRange (5):5i = multiprocessing. Process (target=foo)6 p_list.append (i)7 I.start ()8      forIinchp_list:9 I.join ()TenEnd_time =time.time () One     Print("totally time:{}". Format (End_time-start_time))

The correspondence between the scheme and practice:

Solution 1:--42.97582244873047

Solution 2:--44.63868308067322

Solution 3:--45.94909477233887

Solution 4:--15.7623131275177

Summarize:

The efficiency of multi_thread here is lower than single_thread!. Therefore, compute-intensive applications, do not use multi-threading, but increase the burden! (Gil-induced).

The most effective performance is the multi-process!

CPython concurrency in CPU-intensive applications