StartNew(Action<object> action,object state,CancellationToken cancellationToken,TaskCreationOptions creationOptions,TaskScheduler scheduler);

当工作代码实际开始执行时.

我创建了一个测试来显示：

public static class TestTaskFactory
{
    private class TaskInfo
    {
        public int Number;
    }

    private static int NUM_TASKS = 5;
    private static int NumFinished = 0;

    public static void Run()
    {
        for (int n = 1; n <= NUM_TASKS; n++)
        {
            Log("Starting task #" + n + " ...");
            var task_info = new TaskInfo { Number = n };
            var task = Task.Factory.StartNew(Worker,task_info,CancellationToken.None,TaskCreationOptions.None,TaskScheduler.Default);
            Thread.Sleep(0);
        }

        Log("Waiting for tasks to finish ...");
        while (NumFinished < NUM_TASKS)
        {
            Thread.Sleep(1);
        }

        Log("All done");
    }

    private static void Worker(object state)
    {
        var task_info = (TaskInfo)state;
        Log("Task #" + task_info.Number + " running");

        // Do something
        Thread.Sleep(2000);

        // Done
        ++NumFinished;
    }

    private static void Log(string msg)
    {
        Console.WriteLine(DateTime.Now.ToString("HH.mm.ss.fff") + ": Thread " + Thread.CurrentThread.ManagedThreadId + ": " + msg);
    }
}

Mac下Mono下的输出：

16.57.31.420: Thread 1: Starting task #1 ...
16.57.31.508: Thread 1: Starting task #2 ...
16.57.31.508: Thread 1: Starting task #3 ...
16.57.31.508: Thread 1: Starting task #4 ...
16.57.31.508: Thread 1: Starting task #5 ...
16.57.31.508: Thread 1: Waiting for tasks to finish ...
16.57.31.510: Thread 5: Task #1 running
16.57.32.009: Thread 6: Task #2 running <-- Approx 500 msec later
16.57.32.511: Thread 7: Task #3 running <-- Approx 500 msec later
16.57.33.012: Thread 8: Task #4 running <-- Approx 500 msec later
16.57.33.513: Thread 9: Task #5 running <-- Approx 500 msec later
16.57.35.515: Thread 1: All done

就好像Mono想要在产生新线程之前等待500毫秒来重用现有线程.如果我将工作时间减少到500毫秒以下,则延迟会减少.例如,将worker Thread.Sleep(2000)更改为Thread.Sleep(50)：

...
17.13.20.262: Thread 5: Task #1 running
17.13.20.314: Thread 5: Task #2 running <-- approx 50 msec later
17.13.20.365: Thread 5: Task #3 running <-- approx 50 msec later
17.13.20.416: Thread 5: Task #4 running <-- approx 50 msec later
17.13.20.466: Thread 5: Task #5 running <-- approx 50 msec later

但是在MS Framework 4.0下,工作代码开始之前没有大的延迟：

...
17.05.42.238: Thread 9: Waiting for tasks to finish ...
17.05.42.256: Thread 11: Task #1 running
17.05.42.256: Thread 12: Task #3 running <-- little delay
17.05.42.256: Thread 13: Task #4 running <-- little delay
17.05.42.257: Thread 10: Task #2 running <-- little delay
17.05.43.264: Thread 14: Task #5 running <-- little delay

在我提交关于Mono的错误报告之前,我想要理智地检查我是否错过了我需要在Mono上进行的一些调整或使用Task.Factory错误.我实际上在我的真实应用程序中使用了最大并发调度程序.

所以我的问题是：这是Mono / MonoTouch中的一个错误吗？

更新：我已经从使用Mono *下的ThreadPool切换到Ami Bar的智能线程池(github; Code Project article). GSerjo的Extended Thread Pool看起来也不错,但是我在移动设备上试图避免使用很多依赖项.我在Xamarim thread编写了一些简单的测试.我可能错过了其他100个线程池实现,但到目前为止我对SmartThreadPool感到满意.使用WINDOWS_PHONE模式在MonoTouch下编译.