WaitHandle.WaitAll 64 个句柄限制的解决方法？

Question

我的应用程序通过 ThreadPool.QueueUserWorkItem 生成大量不同的小型工作线程，我通过多个ManualResetEvent 实例对其进行跟踪。我使用 WaitHandle.WaitAll 方法阻止我的应用程序关闭，直到这些线程完成为止。

然而，我以前从未遇到过任何问题，因为我的应用程序承受更多负载，即创建更多线程，我现在开始遇到此异常：

WaitHandles 必须小于或等于 64 - 缺少文档

最好的替代解决方案是什么？

代码片段

List<AutoResetEvent> events = new List<AutoResetEvent>();

// multiple instances of...
var evt = new AutoResetEvent(false);
events.Add(evt);
ThreadPool.QueueUserWorkItem(delegate
{
    // do work
    evt.Set();
});

...
WaitHandle.WaitAll(events.ToArray());

解决方法

int threadCount = 0;
ManualResetEvent finished = new ManualResetEvent(false);

...
Interlocked.Increment(ref threadCount);
ThreadPool.QueueUserWorkItem(delegate
{
    try
    {
         // do work
    }
    finally
    {
        if (Interlocked.Decrement(ref threadCount) == 0)
        {
             finished.Set();
        }
    }
});

...
finished.WaitOne();

Answer 1

创建一个跟踪正在运行的任务数量的变量：

int numberOfTasks = 100;

创建一个信号：

ManualResetEvent signal = new ManualResetEvent(false);

每当任务完成时就减少任务数量：

if (Interlocked.Decrement(ref numberOftasks) == 0)
{

如果没有剩余任务，则设置信号：

    signal.Set();
}

同时，在其他地方，等待信号发出放：

signal.WaitOne();

Answer 2

从 .NET 4.0 开始，您还有另外两个（在我看来，更干净的）选项可供您使用。

第一种是使用 CountdownEvent类。它不需要您自己处理递增和递减：

int tasks = ...; /* however many tasks you're performing */

// Dispose when done.
using (var e = new CountdownEvent(tasks))
{
    // Queue work.
    ThreadPool.QueueUserWorkItem(() => {
        // Do work
        ...

        // Signal when done.
        e.Signal();
    });

    // Wait till the countdown reaches zero.
    e.Wait();
}

但是，有一个更强大的解决方案，那就是使用 Task 类，如下所示：

// The source of your work items, create a sequence of Task instances.
Task[] tasks = Enumerable.Range(0, 100).Select(i =>
    // Create task here.
    Task.Factory.StartNew(() => {
        // Do work.
    }

    // No signalling, no anything.
).ToArray();

// Wait on all the tasks.
Task.WaitAll(tasks);

使用 Task< /code> 类以及对 WaitAll 更干净，IMO，因为您在整个代码中编织了更少的线程原语（注意，没有等待句柄）；您不必设置计数器，处理递增/递减，您只需设置任务然后等待它们。这使得代码更能表达您想要做什么，而不是如何做（至少在管理并行化方面）。

.NET 4.5 提供了更多选项，您可以通过调用 Task 类上的静态 Run 方法：

// The source of your work items, create a sequence of Task instances.
Task[] tasks = Enumerable.Range(0, 100).Select(i =>
    // Create task here.
    Task.Run(() => {
        // Do work.
    })

    // No signalling, no anything.
).ToArray();

// Wait on all the tasks.
Tasks.WaitAll(tasks);

或者，您可以利用 TPL DataFlow 库（它位于 System 命名空间中，因此它是官方的，即使它是从 NuGet 下载的，如实体框架）并使用 < code>ActionBlock，如下所示：

// Create the action block.  Since there's not a non-generic
// version, make it object, and pass null to signal, or
// make T the type that takes the input to the action
// and pass that.
var actionBlock = new ActionBlock<object>(o => {
    // Do work.
});

// Post 100 times.
foreach (int i in Enumerable.Range(0, 100)) actionBlock.Post(null);

// Signal complete, this doesn't actually stop
// the block, but says that everything is done when the currently
// posted items are completed.
actionBlock.Complete();

// Wait for everything to complete, the Completion property
// exposes a Task which can be waited on.
actionBlock.Completion.Wait();

请注意，ActionBlock 默认情况下一次处理一项，因此如果您想让它处理多项一次操作，您必须通过传递 ExecutionDataflowBlockOptions 实例并设置 MaxDegreeOfParallelism 属性：

var actionBlock = new ActionBlock<object>(o => {
    // Do work.
}, new ExecutionDataflowBlockOptions { MaxDegreeOfParallelism = 4 });

如果您的操作确实是线程安全的，那么您可以设置 < code>MaxDegreeOfParallelsim 属性为 DataFlowBlockOptions.Unbounded：

var actionBlock = new ActionBlock<object>(o => {
    // Do work.
}, new ExecutionDataflowBlockOptions { 
    MaxDegreeOfParallelism = DataFlowBlockOptions.Unbounded
});

重点是，您可以对选项的并行程度进行细粒度控制。

当然，如果您想要将一系列项目传递到 ActionBlock 实例中，那么您可以链接 ISourceBlock实现来提供 ActionBlock，如下所示：

// The buffer block.
var buffer = new BufferBlock<int>();

// Create the action block.  Since there's not a non-generic
// version, make it object, and pass null to signal, or
// make T the type that takes the input to the action
// and pass that.
var actionBlock = new ActionBlock<int>(o => {
    // Do work.
});

// Link the action block to the buffer block.
// NOTE: An IDisposable is returned here, you might want to dispose
// of it, although not totally necessary if everything works, but
// still, good housekeeping.
using (link = buffer.LinkTo(actionBlock, 
    // Want to propagate completion state to the action block.
    new DataflowLinkOptions {
        PropagateCompletion = true,
    },
    // Can filter on items flowing through if you want.
    i => true)
{ 
    // Post 100 times to the *buffer*
    foreach (int i in Enumerable.Range(0, 100)) buffer.Post(i);

    // Signal complete, this doesn't actually stop
    // the block, but says that everything is done when the currently
    // posted items are completed.
    actionBlock.Complete();

    // Wait for everything to complete, the Completion property
    // exposes a Task which can be waited on.
    actionBlock.Completion.Wait();
}

根据您需要执行的操作，TPL 数据流库将成为一个更具吸引力的选项，因为它可以处理链接在一起的所有任务之间的并发性，并且它允许您非常具体地说明您希望每个部分的并行程度，同时保持每个块的关注点适当分离。

Answer 3

您的解决方法不正确。原因是，如果最后一个工作项导致 threadCount 在之前变为零，Set 和 WaitOne 可能会发生争用。 > 排队线程必须有机会对所有工作项进行排队。修复方法很简单。将排队线程视为一个工作项本身。将 threadCount 初始化为 1，并在排队完成时递减并发出信号。

int threadCount = 1;
ManualResetEvent finished = new ManualResetEvent(false);
...
Interlocked.Increment(ref threadCount); 
ThreadPool.QueueUserWorkItem(delegate 
{ 
    try 
    { 
         // do work 
    } 
    finally 
    { 
        if (Interlocked.Decrement(ref threadCount) == 0) 
        { 
             finished.Set(); 
        } 
    } 
}); 
... 
if (Interlocked.Decrement(ref threadCount) == 0)
{
  finished.Set();
}
finished.WaitOne(); 

作为个人喜好，我喜欢使用 CountdownEvent 类来为我进行计数。

var finished = new CountdownEvent(1);
...
finished.AddCount();
ThreadPool.QueueUserWorkItem(delegate 
{ 
    try 
    { 
         // do work 
    } 
    finally 
    { 
      finished.Signal();
    } 
}); 
... 
finished.Signal();
finished.Wait(); 

Answer 4

添加到 dtb 的答案中，您可以将其包装到一个漂亮的简单类中。

public class Countdown : IDisposable
{
    private readonly ManualResetEvent done;
    private readonly int total;
    private long current;

    public Countdown(int total)
    {
        this.total = total;
        current = total;
        done = new ManualResetEvent(false);
    }

    public void Signal()
    {
        if (Interlocked.Decrement(ref current) == 0)
        {
            done.Set();
        }
    }

    public void Wait()
    {
        done.WaitOne();
    }

    public void Dispose()
    {
        ((IDisposable)done).Dispose();
    }
}

Answer 5

当我们想要回调时添加到 dtb 的答案。

using System;
using System.Runtime.Remoting.Messaging;
using System.Threading;

class Program
{
    static void Main(string[] args)
    {
        Main m = new Main();
        m.TestMRE();
        Console.ReadKey();

    }
}

class Main
{
    CalHandler handler = new CalHandler();
    int numberofTasks =0;
    public void TestMRE()
    {

        for (int j = 0; j <= 3; j++)
        {
            Console.WriteLine("Outer Loop is :" + j.ToString());
            ManualResetEvent signal = new ManualResetEvent(false);
            numberofTasks = 4;
            for (int i = 0; i <= 3; i++)
            {
                CalHandler.count caller = new CalHandler.count(handler.messageHandler);
                caller.BeginInvoke(i, new AsyncCallback(NumberCallback),signal);
            }
            signal.WaitOne();
        }

    }

    private void NumberCallback(IAsyncResult result)
    {
        AsyncResult asyncResult = (AsyncResult)result;

        CalHandler.count caller = (CalHandler.count)asyncResult.AsyncDelegate;

        int num = caller.EndInvoke(asyncResult);

        Console.WriteLine("Number is :"+ num.ToString());

        ManualResetEvent mre = (ManualResetEvent)asyncResult.AsyncState;
        if (Interlocked.Decrement(ref numberofTasks) == 0)
        {
            mre.Set();
        }
    }

}
public class CalHandler
{
    public delegate int count(int number);

    public int messageHandler ( int number )
    {
        return number;
    }

}

Answer 6

protected void WaitAllExt(WaitHandle[] waitHandles)
{
    //workaround for limitation of WaitHandle.WaitAll by <=64 wait handles
    const int waitAllArrayLimit = 64;
    var prevEndInd = -1;
    while (prevEndInd < waitHandles.Length - 1)
    {
        var stInd = prevEndInd + 1;
        var eInd = stInd + waitAllArrayLimit - 1;
        if (eInd > waitHandles.Length - 1)
        {
            eInd = waitHandles.Length - 1;
        }
        prevEndInd = eInd;

        //do wait
        var whSubarray = waitHandles.Skip(stInd).Take(eInd - stInd + 1).ToArray();
        WaitHandle.WaitAll(whSubarray);
    }

}

Answer 7

我确实通过简单地对要等待的事件数量进行分页来解决这个问题，而没有损失太多性能，并且它在生产环境中运行得很好。遵循代码：

        var events = new List<ManualResetEvent>();

        // code omited

        var newEvent = new ManualResetEvent(false);
        events.Add(newEvent);
        ThreadPool.QueueUserWorkItem(c => {

            //thread code
            newEvent.Set();
        });

        // code omited

        var wait = true;
        while (wait)
        {
            WaitHandle.WaitAll(events.Take(60).ToArray());
            events.RemoveRange(0, events.Count > 59 ? 60 : events.Count);
            wait = events.Any();

        }

Answer 8

这是另一个解决方案。这里的“事件”是ManualResetEvent 的列表。列表的大小可以大于 64 (MAX_EVENTS_NO)。

int len = events.Count;
if (len <= MAX_EVENTS_NO)
    {
        WaitHandle.WaitAll(events.ToArray());
    } else {
        int start = 0;
        int num = MAX_EVENTS_NO;
        while (true)
            {
                if(start + num > len)
                {
                   num = len - start;
                }
                List<ManualResetEvent> sublist = events.GetRange(start, num);
                WaitHandle.WaitAll(sublist.ToArray());
                start += num;
                if (start >= len)
                   break;
           }
   }