[ This question needs to be reimagined. One of my thread queues MUST run on an STA thread, and the code below does not accommodate that. In particular it seems Task<> chooses its own thread and that just is not going to work for me. ]
I have a task queue (BlockingCollection) that I'm running through on a dedicated thread. That queue receives a series of Task<> objects that it runs sequentially within that thread via a while loop.
I need a means of Cancelling that series of tasks, and a means of knowing that the tasks are all complete. I have not been able to figure out how to do this.
Here's a fragment of my queuing class. ProcessQueue is run on a separate thread from main. QueueJob calls occur on the main thread.
using Job = Tuple<Task<bool>, string>;
public class JobProcessor
{
private readonly BlockingCollection<Job> m_queue = new BlockingCollection<Job>();
volatile bool cancel_queue = false;
private bool ProcessQueue()
{
while (true)
{
if (m_queue.IsAddingCompleted)
break;
Job tuple;
if (!m_queue.TryTake(out tuple, Timeout.Infinite))
break;
var task = tuple.Item1;
var taskName = tuple.Item2;
try
{
Console.WriteLine("Task {0}::{1} starting", this.name, taskName);
task.RunSynchronously();
Console.WriteLine("Task {0}::{1} completed", this.name, taskName);
}
catch (Exception e)
{
string message = e.Message;
}
if (cancel_queue) // CANCEL BY ERASING TASKS AND NOT RUNNING.
{
while (m_queue.TryTake(out tuple))
{
}
}
} // while(true)
return true;
}
public Task<bool> QueueJob(Func<bool> input)
{
var task = new Task<bool>(input);
try
{
m_queue.Add(Tuple.Create(task, input.Method.Name));
}
catch (InvalidOperationException)
{
Task<bool> dummy = new Task<bool>(() => false);
dummy.Start();
return dummy;
}
return task;
}
Here are the functions that trouble me:
public void ClearQueue()
{
cancel_queue = true;
// wait for queue to become empty. HOW?
cancel_queue = false;
}
public void WaitForCompletion()
{
// wait for all tasks to be completed.
// not sufficient to wait for empty queue because the last task
// must also execute and finish. HOW?
}
}
Here is some usage:
class SomeClass
{
void Test()
{
JobProcessor jp = new JobProcessor();
// launch Processor loop on separate thread... code not shown.
// send a bunch of jobs via QueueJob... code not show.
// launch dialog... code not shown.
if (dialog_result == Result.Cancel)
jp.ClearQueue();
if (dialog_result == Result.Proceed)
jp.WaitForCompletion();
}
}
The idea is after the work is completed or cancelled, new work may be posted. In general though, new work may come in asynchronously. WaitForCompletion might in fact be "when all work is done, inform the user and then do other stuff", so it doesn't strictly have to be a synchronous function call like above, but I can't figure how to make these happen.
(One further complication, I expect to have several queues that interact. While I am careful to keep things parallelized in a way to prevent deadlocks, I am not confident what happens when cancellation is introduced into the mix, but this is probably beyond scope for this question.)
WaitForCompletion() sounds easy enough. Create a semaphore or event, create a task whose only action is to signal the semaphore, queue up the task, wait on the semaphore.
When the thread finishes the last 'real' task, the semaphore task will be run and so the thread that called WaitForCompletion will become ready/running:)
Would not a similar approach work for cancellation? Have a very high priority thread that you create/signal that drains the queue of all pending jobs, disposing them, queueing up the semaphore task and waiting for the 'last task done' signal?
I found this example in the book. And it's written that this code leads to deadlock. But I don't get it and don't understand why it should be?
I always get
“Locked B and A”
“Locked A and B”
static void Main()
{
object lockA = new object();
object lockB = new object();
var up = Task.Run(() =>
{
lock (lockA)
{
Thread.Sleep(1000);
lock (lockB)
{
Console.WriteLine(“Locked A and B”);
}
}
});
lock (lockB)
{
lock (lockA)
{
Console.WriteLine(“Locked B and A”);
}
}
up.Wait();
Console.ReadLine();
}
The code isn't guaranteed to deadlock. There is a race condition as to whether or not it will deadlock. If the newly started task manages to grab the first lock before the main thread goes on to grab that same lock, then the code deadlocks, if (as is no doubt happening for you) the first thread manages to get both locks before the worker thread does anything, then there is no deadlock.
I'm learning multithreading and have come to a producer/consumer problem.
Here is the sample implementation of a producer/consumer queue taken from the Albahari's website (authors of the "C# in a Nutshell" book series):
using System;
using System.Threading;
using System.Collections.Generic;
class ProducerConsumerQueue : IDisposable
{
EventWaitHandle _wh = new AutoResetEvent (false);
Thread _worker;
readonly object _locker = new object();
Queue<string> _tasks = new Queue<string>();
public ProducerConsumerQueue()
{
_worker = new Thread (Work);
_worker.Start();
}
public void EnqueueTask (string task)
{
lock (_locker) // <---------------------------------------------- 1
_tasks.Enqueue (task);
_wh.Set();
}
public void Dispose()
{
EnqueueTask (null); // Signal the consumer to exit.
_worker.Join(); // Wait for the consumer's thread to finish.
_wh.Close(); // Release any OS resources.
}
void Work()
{
while (true)
{
string task = null;
lock (_locker) // <---------------------------------------------- 2
if (_tasks.Count > 0)
{
task = _tasks.Dequeue();
if (task == null) return;
}
if (task != null)
{
Console.WriteLine ("Performing task: " + task);
Thread.Sleep (1000); // simulate work...
}
else
_wh.WaitOne(); // No more tasks - wait for a signal
}
}
}
I understand the code, but my question is what happens if I remove locks in rows commented with "1" and "2"?
I tried to imagine different concurrent scenarios and couldn't have found one which would lead to a problem.
If you decide to answer my question please show a detailed step-by-step sequence of actions which would lead to a problem.
Queue<T> is not thread-safe.
If you write to it from multiple threads, it will break.
(probably when two threads try to resize the buffer at once)
Also, even if it was thread-safe, removing the second lock would allow a second thread to remove the last item while a first thread is inside the if, making the first thread try to read an empty queue.
I have two threads which use two different functions. First one to search from start to end and the second one to search from end to start.
Now I'm using Thread.Sleep(10) for synchronisation, but it takes too much time, and testing is not possible in such condition.
Any idea how can I sync two threads with different functions?
It depends slightly on what you want to do.
If you have two threads and you just want to exit one when the other reaches "success" (or n threads and you want to exit them all when one reaches "success" first) you just need to periodically check for success on each thread.
Use Interlocked to do this without locks, or some other mechanism (see below)
Use cancellable Task objects
If you need to do your search in phases, where each thread does something and then waits for the other to catch up, you need a different approach.
Use Barrier
Given that you are doing an A*-search you likely need a combination of all two/three anyway:
Barrier to coordinate the steps and update the open set between steps
Success signalling to work out when to exit threads if another thread succeeded
Task objects with CancellationToken to allow callers to cancel the search.
Another answer suggested Semaphore - this is not really suitable for your needs (see comments below).
Barrier can be used for searches such as this by:
enter step 0 of the algorithm
n threads split the current level into equal parts and work on each half, when each completes then it signals and waits for the other thread
when all threads are ready, proceed to the next step and repeat the search
Simple check for exit - Interlocked
The first part is checking for success. If you want to stay "lockless", you can use Interlocked to do this, the general pattern is:
// global success indicator
private const int NotDone = 0;
private const int AllDone = 1;
private int _allDone = NotDone;
private GeneralSearchFunction(bool directionForward) {
bool iFoundIt = false;
... do some search operations that won't take much time
if (iFoundIt) {
// set _allDone to AllDone!
Interlocked.Exchange(ref _allDone, AllDone);
return;
}
... do more work
// after one or a few iterations, if this thread is still going
// see if another thread has set _allDone to AllDone
if (Interlocked.CompareExchange(ref _allDone, NotDone, NotDone) == AllDone) {
return; // if they did, then exit
}
... loop to the top and carry on working
}
// main thread:
Thread t1 = new Thread(() => GeneralSearchFunction(true));
Thread t2 = new Thread(() => GeneralSearchFunction(false));
t1.Start(); t2.Start(); // start both
t1.Join(); t2.Join();
// when this gets to here, one of them will have succeeded
This is the general pattern for any kind of success or cancellation token:
do some work
if you succeed, set a signal every other thread checks periodically
if you haven't yet succeeded then in the middle of that work, either every iteration, or every few iterations, check to see if this thread should exit
So an implementation would look like:
class Program
{
// global success indicator
private const int NotDone = 0;
private const int AllDone = 1;
private static int _allDone = NotDone;
private static int _forwardsCount = 0; // counters to simulate a "find"
private static int _backwardsCount = 0; // counters to simulate a "find"
static void Main(string[] args) {
var searchItem = "foo";
Thread t1 = new Thread(() => DoSearchWithBarrier(SearchForwards, searchItem));
Thread t2 = new Thread(() => DoSearchWithBarrier(SearchBackwards, searchItem));
t1.Start(); t2.Start();
t1.Join(); t2.Join();
Console.WriteLine("all done");
}
private static void DoSearchWithBarrier(Func<string, bool> searchMethod, string searchItem) {
while (!searchMethod(searchItem)) {
// after one or a few iterations, if this thread is still going
// see if another thread has set _allDone to AllDone
if (Interlocked.CompareExchange(ref _allDone, NotDone, NotDone) == AllDone) {
return; // if they did, then exit
}
}
Interlocked.Exchange(ref _allDone, AllDone);
}
public static bool SearchForwards(string item) {
// return true if we "found it", false if not
return (Interlocked.Increment(ref _forwardsCount) == 10);
}
public static bool SearchBackwards(string item) {
// return true if we "found it", false if not
return (Interlocked.Increment(ref _backwardsCount) == 20); // make this less than 10 to find it backwards first
}
}
Using Tasks to the same end
Of course, this wouldn't be .NET 4.5 without using Task:
class Program
{
private static int _forwardsCount = 0; // counters to simulate a "find"
private static int _backwardsCount = 0; // counters to simulate a "find"
static void Main(string[] args) {
var searchItem = "foo";
var tokenSource = new CancellationTokenSource();
var allDone = tokenSource.Token;
Task t1 = Task.Factory.StartNew(() => DoSearchWithBarrier(SearchForwards, searchItem, tokenSource, allDone), allDone);
Task t2 = Task.Factory.StartNew(() => DoSearchWithBarrier(SearchBackwards, searchItem, tokenSource, allDone), allDone);
Task.WaitAll(new[] {t2, t2});
Console.WriteLine("all done");
}
private static void DoSearchWithBarrier(Func<string, bool> searchMethod, string searchItem, CancellationTokenSource tokenSource, CancellationToken allDone) {
while (!searchMethod(searchItem)) {
if (allDone.IsCancellationRequested) {
return;
}
}
tokenSource.Cancel();
}
...
}
However, now you have used the CancellationToken for the wrong things - really this should be kept for the caller of the search to cancel the search, so you should use CancellationToken to check for a requested cancellation (only the caller needs tokenSource then), and a different success synchronisation (such as the Interlocked sample above) to exit.
Phase/step synchronisation
This gets harder for many reasons, but there is a simple approach. Using Barrier (new to .NET 4) in conjunction with an exit signal you can:
Perform the assigned thread's work for the current step, and then wait for the other thread to catch up before doing another iteration
Exit both threads when one succeeds
There are many different approaches for thread sync, depending on exactly what you want to achieve. Some are:
Barrier: This is probably the most suitable if you are intending for both your forwards and backwards searches to run at the same time. It also screams out your intent, i.e. "all threads can't go on until they everyone reaches a barrier"
ManualResetEvent - when one thread releases a signal, all others can proceed until it is set again. AutoResetEvent is similar, except it only allows one thread to proceed before blocking again.
Interlocked - in combination with SpinWait this is a viable lockless solution
Semaphore - possible to use, but not really suited for your scenario
I have only provided a full sample for Barrier here as it seems the most suitable in your case. Barrier is one of the most performant, second only to ManualResetEventSlim (ref. albahari), but using ManualResetEvent will need more complex code.
Other techniques to look at, if none of the above work for you are Monitor.Wait and Monitor.Pulse (now you're using locking) and Task Continuations. The latter is more used for passing data from one async operation to another, but it could be used for your scenario. And, as with the samples at the top of the answer, you are more likely to combine Task with Barrier than use one instead of the other. Task Continuations could be used to do the post-step revision of the open set in the A*-search, but you can just as easily use Barrier for that anyway.
This code, using Barrier works. In essence, DoSearchWithBarrier is the only bit doing the synchronisation - all the rest is setup and teardown code.
class Program {
...
private static int _forwardsCount = 0; // counters to simulate a "find"
private static int _backwardsCount = 0; // counters to simulate a "find"
static void Main(string[] args) {
Barrier barrier = new Barrier(numThreads,
b => Console.WriteLine("Completed search iteration {0}", b.CurrentPhaseNumber));
var searchItem = "foo";
Thread t1 = new Thread(() => DoSearchWithBarrier(SearchForwards, searchItem, barrier));
Thread t2 = new Thread(() => DoSearchWithBarrier(SearchBackwards, searchItem, barrier));
t1.Start(); Console.WriteLine("Started t1");
t2.Start(); Console.WriteLine("Started t2");
t1.Join(); Console.WriteLine("t1 done");
t2.Join(); Console.WriteLine("t2 done");
Console.WriteLine("all done");
}
private static void DoSearchWithBarrier(Func<string, bool> searchMethod, string searchItem, Barrier barrier) {
while (!searchMethod(searchItem)) {
// while we haven't found it, wait for the other thread to catch up
barrier.SignalAndWait(); // check for the other thread AFTER the barrier
if (Interlocked.CompareExchange(ref _allDone, NotDone, NotDone) == AllDone) {
return;
}
}
// set success signal on this thread BEFORE the barrier
Interlocked.Exchange(ref _allDone, AllDone);
// wait for the other thread, and then exit (and it will too)
barrier.SignalAndWait();
}
...
}
There are two things going on here:
Barrier is used to synchronise the two threads so they can't do their next step until the other has caught up
The exit signal uses Interlocked, as I first described.
Implementing this for A* searches is very similar to the above sample. Once all threads reach the barrier and therefore continue you could use a ManualResetEvent or a simple lock to then let one (and only one) revise the open set.
A note on Semaphore
This is probably not what you want as it's most often used when you have a limited pool of resources, with more resource users requiring access than you have resources.
Think of the PlayStation with CoD on it in the corner of the work canteen - 4 controllers, 20 people waiting (WaitOne) to use it, as soon as your character dies you Release the controller and someone else takes your place. No particular FIFO/LIFO ordering is enforced, and in fact Release can be called by the bouncer you employ to prevent the inevitable fights (i.e. thread identity is not enforced).
Simple check for exit - other approaches
Use of lock for simple success indication
You can achieve the same with locking. Both Interlocked and lock ensure you don't see any memory cache issues with reading a common variable between threads:
private readonly object _syncAllDone = new object();
...
if (iFoundIt) {
lock (_syncAllDone) { _allDone = AllDone };
return;
}
...
// see if another thread has set _allDone to AllDone
lock (_syncAllDone) {
if (_allDone == AllDone) {
return; // if they did, then exit
}
}
The disadvantage of this is that locking may well be slower, but you need to test for your situation. The advantage is that if you are using lock anyway to do other things such as writing out results from your thread, you don't have any extra overhead.
Use of ManualResetEvent for simple success indication
This is not really the intended use of reset events, but it can work. (If using .NET 4 or later, use ManualResetEventSlim instead of ManualResetEvent):
private ManualResetEvent _mreAllDone = new ManualResetEvent(true); // will not block a thread
...
if (iFoundIt) {
_mreAllDone.Reset(); // stop other threads proceeding
return;
}
...
// see if another thread has reset _mreAllDone by testing with a 0 timeout
if (!_mreAllDone.WaitOne(0)) {
return; // if they did, then exit
}
Phase synchronisation - other approaches
All of the other approaches get a lot more complex, as you have to do two-way continuation checks to prevent race conditions and permanently blocked threads. I don't recommend them, so I won't provide a sample here (it would be long and complicated).
References:
Interlocked
ManualResetEvent
MSDN - ManualResetEvent and ManualResetEventSlim
Barrier
MSDN - Continuation Tasks
MSDN - Task Cancellation
Semaphore
thread.Join()
is possibly what your after. This will make your current thread block until the other thread ends.
It's possible to Join multiple threads there by syncing all of them to one point.
List<Thread> threads = new List<Thread>();
threads.Add(new Thread(new ThreadStart(<Actual method here>)));
threads.Add(new Thread(new ThreadStart(<Another method here>)));
threads.Add(new Thread(new ThreadStart(<Another method here>)));
foreach(Thread thread in threads)
{
thread.Start();
}
//All your threads are now running
foreach(Thread thread in threads)
{
thread.Join();
}
//You wont get here until all those threads have finished
In some cases You can use AutoResetEvent to wait some result from thread.
You can use Task's for start/stop/wait result of some workers.
You can use Producer/Consumer pattern with BlockingCollection in case your functions eat some data and returns collection of something.
I am creating a thread A and in that thread creating a new thread B.
So how is the thread hierarchy? Thread B is child of Thread A? Or the threads are created as peers?
I want to abort the parent thread A which in turn kills/aborts its child threads.
How is that possible in C#?
Threads should ideally never be aborted. It simply isn't safe. Consider this as a way of putting down an already sick process. Otherwise, avoid like the plague.
The more correct way of doing this is to have something that the code can periodically check, and itself decide to exit.
An example of stopping threads the polite way:
using System;
using System.Threading;
namespace Treading
{
class Program
{
static void Main(string[] args)
{
Console.WriteLine("Main program starts");
Thread firstThread = new Thread(A);
ThreadStateMessage messageToA = new ThreadStateMessage(){YouShouldStopNow = false};
firstThread.Start(messageToA);
Thread.Sleep(50); //Let other threads do their thing for 0.05 seconds
Console.WriteLine("Sending stop signal from main program!");
messageToA.YouShouldStopNow = true;
firstThread.Join();
Console.WriteLine("Main program ends - press any key to exit");
Console.Read();//
}
private class ThreadStateMessage
{
public bool YouShouldStopNow = false; //this assignment is not really needed, since default value is false
}
public static void A(object param)
{
ThreadStateMessage myMessage = (ThreadStateMessage)param;
Console.WriteLine("Hello from A");
ThreadStateMessage messageToB = new ThreadStateMessage();
Thread secondThread = new Thread(B);
secondThread.Start(messageToB);
while (!myMessage.YouShouldStopNow)
{
Thread.Sleep(10);
Console.WriteLine("A is still running");
}
Console.WriteLine("Sending stop signal from A!");
messageToB.YouShouldStopNow = true;
secondThread.Join();
Console.WriteLine("Goodbye from A");
}
public static void B(object param)
{
ThreadStateMessage myMessage = (ThreadStateMessage)param;
Console.WriteLine("Hello from B");
while(!myMessage.YouShouldStopNow)
{
Thread.Sleep(10);
Console.WriteLine("B is still running");
}
Console.WriteLine("Goodbye from B");
}
}
}
Using Thread.Abort(); causes an exception to be thrown if your thread is in a waiting state of any kind. This is sort of annoying to handle, since there are quite a number of ways that a thread can be waiting. As others have said, you should generally avoid doing it.
Thread.Abort will do what you want, but it is not recommended to abort thread, better choose is to think a way for finishing threads correctly by Thread synchronization mechanism
Here's yet another way to politely signal a thread to die:
Note that this fashion favors finite state automatons where the slave periodically checks for permission to live, then performs a task if allowed. Tasks are not interrupted and are 'atomic'. This works great with simple loops or with command queues. Also this makes sure the thread doesn't spin 100% cpu by giving the slave thread a rest period, set this one to 0 if you don't want any rest in your slave.
var dieEvent = new AutoResetEvent(false);
int slaveRestPeriod = 20;// let's not hog the CPU with an endless loop
var master = new Thread(() =>
{
doStuffAMasterDoes(); // long running operation
dieEvent.Set(); // kill the slave
});
var slave = new Thread(() =>
{
while (!dieEvent.WaitOne(restPeriod))
{
doStuffASlaveDoes();
}
});
slave.Start();
master.Start();
Threads are created as peers, obtain a handle to Thread A and then call ThreadA.Abort()
to forcefully end it. It's better to check a boolean in the thread and if it evaluates to false exit the thread.
public class MyClass
{
public static Thread ThreadA;
public static Thread ThreadB;
private void RunThings()
{
ThreadA = new Thread(new ThreadStart(ThreadAWork));
ThreadB = new Thread(new ThreadStart(ThreadBWork));
ThreadA.Start();
ThreadB.Start();
}
static void ThreadAWork()
{
// do some stuff
// thread A will close now, all work is done.
}
static void ThreadBWork()
{
// do some stuff
ThreadA.Abort(); // close thread A
// thread B will close now, all work is done.
}
}