[Python] It seems that global variables cannot be referenced in Multiprocessing

I didn't have a report on this, so I'll drop a note.

In Python, the standard module Multiprocessing allows parallel computing. Hardware parallel computing like GPU is not supported, it is just multi-process processing on CPU.

The elements of the pooled argument list are passed to the arguments of the function you want to process in parallel.

If you write a script that refers to a global variable in a function that is processed in parallel, it will be as follows.

`example.py`


import multiprocessing as mp

a = 0

def init_global():
    global a
    a = 100

def func(proc):
    global a
    print("Inside MultiProcess: ", proc, a)
    return a

if __name__ == "__main__":

    print("Before Change", a)

    init_global()

    print("After Change", a)

    pool = mp.Pool(4)
    callback = pool.map(func, range(4))
    pool.terminate()
    pool.join()

    print(sum(callback))

result As you can see, you can refer to and assign in a normal function (init_global), but you cannot refer to a global variable in a function that is processed in parallel.

If you really want to refer to the value of a global variable in a parallel processing function, you have to specify the argument from the caller. (If you want to pass multiple arguments, you need to generate a tuple)

`example.py`


import multiprocessing as mp

a = 0
b = 1111

def init_global():
    global a
    a = 100

def func(proc):
    print("Inside MultiProcess: ", proc[0], proc[1])
    return proc[0]

if __name__ == "__main__":

    print("Before Change", a)

    init_global()

    print("After Change", a)

    pool = mp.Pool(4)
    args = [(a,b) for i in range(4)]
    callback = pool.map(func, args)
    pool.terminate()
    pool.join()

    print(sum(callback))

result I was able to hand it over.