Python多个进程不会循环(Python multiple processes do not loop on for)

 我正在研究几种算法的实现，以计算图上的最短路径。 
 
 我已经设法顺序实现了Dijkstra的算法，现在我正在尝试通过Python的多处理模块来优化我的算法。 
 
 整个代码都有效。 我在这里要做的是： 
 
 
 
 首先用nb_cpu = mp.cpu_count()来检查我可以处理多少cpus 
 
 
 然后相应地划分我在图表中的所有节点 
 
 
 最后调用方法subprocess_dijkstra ，它应该为每个节点计算dijkstra算法，它作为一个参数给出（这个想法是每个进程只需要为图的一小部分计算算法）。 
 
 
 当我运行我的脚本（从main.py文件中调用我只是根据我的需要格式化数据）时，我有4个进程正在启动。 
 
 但是，它们似乎不在subprocess_dijkstra for node in nodes定义的for node in nodes循环中执行for node in nodes 。 
 
 每个进程只计算一次代码，然后无限期地暂停... 
 
 这是我在Python下进行多处理的第一次尝试，所以我可能错过了一个细节。 有人有想法吗？ 
 
 当我中断脚本时，python告诉我中断发生在p.join()行上。 
 
 感谢有人帮助我:) 
 
 这是我的代码： 
 
import multiprocessing as mp

def subprocess_dijkstra(do_print, nodes, tab_contenu, tab_distances):
    tab_dist_initial = dict(tab_distances)
    tab_dist = dict()
    for node in nodes:
        visited_nodes = list()
        tab_dist = dict(tab_dist_initial)
        dmin = -1
        resultat = ""
        filename = "dijkstra"+str(node)+".txt"

        if do_print:
            dt = open(filename, 'w')
            tab_dist[node] = 0

            """Ligne de résultat initiale"""
            for valeur in tab_dist.values():
                resultat += str(valeur)
                resultat += " "
            resultat += "\n"

            dt.write(resultat)

        while len(visited_nodes) != len(tab_contenu):
            """ On se place sur le noeud non visité qui a la distance minimale de notre départ """
            for cle, valeur in tab_dist.items():
                if cle not in visited_nodes:
                    if dmin ==-1 or valeur<dmin:
                        dmin = valeur
                        node = cle


            """ On vérifie que le noeud n'a pas déjà été visité """
            if (node not in visited_nodes):
                """ On regarde les fils de ce noeud et la longueur des arcs"""
                for cle,valeur in tab_contenu[node].items():
                    tab_dist[cle] = min(tab_dist[cle], tab_dist[node]+valeur)

                visited_nodes.append(node)

                if do_print:
                    resultat = ""
                    """ Ligne de résultat """
                    for valeur in tab_dist.values():
                        resultat += str(valeur)
                        resultat += " "
                    resultat += "\n"

                    dt.write(resultat)

        if do_print:
            dt.close()


def main(do_print,donnees):

    tab_contenu = donnees[1]
    nb_nodes = int(donnees[0])
    tab_distances = {x: float('inf') for x in range(nb_nodes)}
    args=[(do_print, x, tab_contenu, tab_distances) for x in range(nb_nodes)]
    nb_cpu = mp.cpu_count()


    pool = mp.Pool(processes = nb_cpu)
    pool.starmap(subprocess_dijkstra, args)
    pool.close()
    pool.join()

I'm working on the implementation of several algorithms to compute shortest paths on graphs.
 
I have managed to implement Dijkstra's algorithm sequentially and I'm now trying to optimize my algorithm through the multiprocessing module of Python.
 
As a whole the code works. What I am trying to do here is :
 
 
 
First to check how many cpus I can work on with nb_cpu = mp.cpu_count()
 
 
Then dividing all the nodes I have in my graph accordingly
 
 
Finally calling the method subprocess_dijkstra that should compute the dijkstra algorithm for each of the nodes it is given as an argument (the idea being that each process only has to compute the algorithm for a smaller part of the graph).
 
 
When I run my script (called from a main.py file where I just format the data to suit my needs), I have 4 processes launched as I should.
 
However, they do not seem to execute the for node in nodes loop defined in subprocess_dijkstra.
 
Each process only computes the code once and then they go on hold indefinitely...
 
It is my first attempt at multiprocessing under Python so I may have missed a detail. Does anybody have an idea ?
 
When I interrupt the script, python tells me that the interruption takes place on the p.join() line.
 
Thanks to anyone helping me :)
 
Here is my code : 
 
import multiprocessing as mp

def subprocess_dijkstra(do_print, nodes, tab_contenu, tab_distances):
    tab_dist_initial = dict(tab_distances)
    tab_dist = dict()
    for node in nodes:
        visited_nodes = list()
        tab_dist = dict(tab_dist_initial)
        dmin = -1
        resultat = ""
        filename = "dijkstra"+str(node)+".txt"

        if do_print:
            dt = open(filename, 'w')
            tab_dist[node] = 0

            """Ligne de résultat initiale"""
            for valeur in tab_dist.values():
                resultat += str(valeur)
                resultat += " "
            resultat += "\n"

            dt.write(resultat)

        while len(visited_nodes) != len(tab_contenu):
            """ On se place sur le noeud non visité qui a la distance minimale de notre départ """
            for cle, valeur in tab_dist.items():
                if cle not in visited_nodes:
                    if dmin ==-1 or valeur<dmin:
                        dmin = valeur
                        node = cle


            """ On vérifie que le noeud n'a pas déjà été visité """
            if (node not in visited_nodes):
                """ On regarde les fils de ce noeud et la longueur des arcs"""
                for cle,valeur in tab_contenu[node].items():
                    tab_dist[cle] = min(tab_dist[cle], tab_dist[node]+valeur)

                visited_nodes.append(node)

                if do_print:
                    resultat = ""
                    """ Ligne de résultat """
                    for valeur in tab_dist.values():
                        resultat += str(valeur)
                        resultat += " "
                    resultat += "\n"

                    dt.write(resultat)

        if do_print:
            dt.close()


def main(do_print,donnees):

    tab_contenu = donnees[1]
    nb_nodes = int(donnees[0])
    tab_distances = {x: float('inf') for x in range(nb_nodes)}
    args=[(do_print, x, tab_contenu, tab_distances) for x in range(nb_nodes)]
    nb_cpu = mp.cpu_count()


    pool = mp.Pool(processes = nb_cpu)
    pool.starmap(subprocess_dijkstra, args)
    pool.close()
    pool.join()

原文：https://stackoverflow.com/questions/29447447