Question

所以我正在进行迭代深化的多进程国际象棋搜索算法，它目前工作。不幸的是，有时程序会在except块中的某处停顿。当它停止时会有7个僵尸孩子（我在8核机器上运行，所以我创建了7个进程）。这种情况相当罕见，但在我的程序实际播放时，每千次移动会发生几次。

似乎只发生在linux / mac机器上。我希望我的搜索在10秒后停止并返回前一个movelist的值。

有关如何调试或解决此问题的任何建议？

def search(board, maxDepth, maxTime):
    start = time.time()

    threadData = [[i, None, 0, board] for i in board.legal_moves]
    threads = min(len(threadData), cpu_count()-1)
    pool = ThreadPool(processes=threads)
    moveList = []
    gotDepth = -1
    for depth in range(0, maxDepth):
        # set the current depth to search
        for i in range(len(threadData)):
            threadData[i][2] = depth

        result = pool.map_async(moveThreading, threadData)

        try:
            threadData = result.get(maxTime - (time.time() - start))
            gotDepth = depth
            threadData = sorted(threadData, key=itemgetter(1), reverse=True)
            moveList = [[item[0], item[1]] for item in threadData]
        except TimeoutError:
            #Diagnostic printing for analysis after game is played
            if DIAGNOSTIC:
                with open("diagnostic.txt", "a") as f:
                    f.write("@"+str(depth-1)+'==')
                    for i in moveList:
                        f.write("['"+str(i[0])+"'_"+str(i[1])+']')
                    f.write('\n')
            pool.terminate()
            pool.join()
            pool = None
            break

    return gotDepth, moveList

Python多处理池停滞/僵尸儿童

0 个答案: