CODEWITHSUNDEEP
pythonpython-multiprocessing
Alecs Gavyn
Alecs Gavyn

Reputation: 23

When implementing Python's multiprocessing, function gives different return values?

A function is giving different results when I run it from one process from when I'm calling it with multiprocessing. I'm not sure why. I'm creating a list from the manager, and calling the same target function with different parameters for each process. The target function calls a function that is imported from a different module. It is the imported function that is giving me different results based on when I'm calling it from multiple processes or one process.

For example:

from foo import foo_function
from multiprocessing import Process, Manager


def another_function(indices, a_list, return_list):
    for i in indices:
        for j in a_list:
            return_list.append(foo_function(i, j))


if __name__ == '__main__':
    jobs = []
    manager = Manager()
    return_list = manager.list()
    all_indices = [[0, 1, 2], [3, 4, 5], [6, 7, 8]]
    all_lists = [['a', 'b', 'c'], ['d', 'e', 'f'], ['g', 'h', 'i']]
    for i in range(3):
        jobs.append(Process(target=another_function, args=(all_indices[i], all_lists[i])))
        jobs[i].start()
    for i in range(3):
        jobs[i].join()

And foo_function() is giving me different results when I call it from multiple Processes vs when I call it from one.

EDIT: Here is what the actual foo_function is:

def battle(ally: CustomClass, enemy: CustomClass, shields: int):
    ally.starting_shields = shields
    enemy.starting_shields = shields

    ally.reset()
    enemy.reset()

    turns = 0

    # Main Battle Loop
    while ally.is_alive() and enemy.is_alive():
        ally.reduce_cooldown()
        enemy.reduce_cooldown()

        turns += 1
        if ally.can_act():
            if not ally.use_charge_move(enemy):
                ally.use_move(ally.fast_move, enemy)

        if enemy.can_act():
            if not enemy.use_charge_move(ally):
                enemy.use_move(enemy.fast_move, ally)

    # There are 2 points for using enemy shields and 3 for using enemy health.
    ally_rating = enemy.starting_shields - enemy.get_shields()
    enemy_rating = ally.starting_shields - ally.get_shields()
    ally_rating += 5 * (enemy.starting_health - enemy.get_health()) / enemy.starting_health
    enemy_rating += 5 * (ally.starting_health - ally.get_health()) / ally.starting_health
    if ally.get_health() > 0:
        ally_rating += 3 * ally.energy / 100
    if enemy.get_health() > 0:
        enemy_rating += 3 * enemy.energy / 100

    total_rating = ally_rating + enemy_rating
    return int(round(1000 * ally_rating / total_rating, 0)), int(round(1000 * enemy_rating / total_rating, 0))

As you can see, it's only calling the CustomClasses methods and only uses local variables.

Upvotes: 1

Views: 118

Answers (2)

Alecs Gavyn
Alecs Gavyn

Reputation: 23

The issue ended up being one not related to multiprocessing. Sorry.

Upvotes: 1

Luca Bezerra
Luca Bezerra

Reputation: 1188

Pretty hard to say without knowing foo_function, but it's probably due to different processes accessing a reference to the same list and finding it to have different values for each, since there doesn't seem to be any concurrence handling there.

Upvotes: 0

Related Questions