I'm currently trying to implement a real-time multithreading software in C#. I need 3 threads. Every thread execution has to be finished before a deadline (500µs / 100µs / 50µs). The threads must run parallel during the whole runtime (until the user shuts down the program).
Is there a mecanism that can guarantee that the thread execution will not pass the deadline?
Here is my code :
static void Main(string[] args)
{
Thread thread1 = new Thread(FirstThread);
Thread thread2 = new Thread(SecondThread);
Thread thread3 = new Thread(ThirdThread);
thread1.start();
thread2.start();
thread3.start();
}
static void FirstThread()
{
while(true)
{
SleepMicroSec(500);
}
}
static void SecondThread()
{
while(true)
{
SleepMicroSec(100);
}
}
static void ThirdThread()
{
while(true)
{
SleepMicroSec(50);
}
}
private static void SleepMicroSec(long microSec)
{
var sw = Stopwatch.StartNew();
while (sw.ElapsedTicks / (Stopwatch.Frequency / (1000L * 1000L)) < microSec)
{
}
}
I expect the scheduler to be able to perform the context switching if the Task deadline is reached.
Thanks in advance for your answers !
Here is a method that invokes repeatedly an action in a background thread, aborting and restarting the thread every time the deadline is passed. It also accepts a CancellationToken to allow for premature cancellation of the procedure (before the end of the program).
private static void RepeatInBackgroundThread(Action action, int timeout,
CancellationToken cancellationToken)
{
var timer = new System.Timers.Timer(timeout);
timer.AutoReset = false; // to raise the Elapsed event only once
var thread = new Thread(() =>
{
while (true)
{
if (cancellationToken.IsCancellationRequested) return;
timer.Start();
action();
timer.Stop();
}
});
timer.Elapsed += (sender, e) =>
{
thread.Abort();
thread.Join(); // Wait for the thread to die
if (cancellationToken.IsCancellationRequested) return;
RepeatInBackgroundThread(action, timeout, cancellationToken);
};
thread.IsBackground = true;
thread.Start();
}
Usage example:
var random = new ThreadLocal<Random>(() => new Random());
var cts = new CancellationTokenSource();
RepeatInBackgroundThread(() => Thread.Sleep(random.Value.Next(0, 1000)), 500, cts.Token);
RepeatInBackgroundThread(() => Thread.Sleep(random.Value.Next(0, 200)), 100, cts.Token);
RepeatInBackgroundThread(() => Thread.Sleep(random.Value.Next(0, 100)), 50, cts.Token);
//cts.CancelAfter(10000);
It should be noted that aborting threads is not a good practice in general.
Related
How would you synchronise multiple threads to be at a certain place in their code at a particular time?
I am not looking for WaitAll, await etc. As in the example code, the threads do not return, only one thread is created per worker.
I have made this monster...
It works for simple and slow cases, but there is a race condition: when a fast thread locks _sync before the synchroniser. As a result sometimes that thread will execute more than once per cycle and break any assumptions of what is ready for use by other threads.
static readonly Stopwatch timer = new Stopwatch();
static readonly object _locker1 = new object();
static readonly object _locker2 = new object();
...
static readonly object _sync = new object();
static void Main()
{
Task.Run(Synchroniser);
Task.Run(DoWork1);
Task.Run(DoWork2);
...
}
static void Synchroniser()
{
while (true)
{
lock (_sync)
{
timer.Restart();
while (timer.ElapseMilliseconds < 16) ;
lock (_locker1) ;
lock (_locker2) ;
...
}
}
}
static void DoWork1()
{
while (true)
{
lock (_locker1)
{
//all worker threads continue from here at the same time
...
}
lock (_sync) ;
}
}
...
It's hard to tell what kind of behavior you're after, but I think this is what you're trying to achieve. Run it and see how it behaves...
The code makes use of the ManualResetEvent class and the WaitAll function:
class Program
{
static readonly ManualResetEvent startWorkers = new ManualResetEvent(false); // workers initially blocked
static readonly ManualResetEvent worker1 = new ManualResetEvent(false);
static readonly ManualResetEvent worker2 = new ManualResetEvent(false);
static readonly Random R = new Random();
static void Main()
{
Task.Run(new Action(Synchroniser));
Task.Run(new Action(DoWork1));
Task.Run(new Action(DoWork2));
Console.ReadLine(); // keep program from closing
}
static void Synchroniser()
{
ManualResetEvent[] workers = new ManualResetEvent[] { worker1, worker2 };
while (true)
{
Console.WriteLine("Pausing...");
Thread.Sleep(5000); // long pause so we can see what is happening
Console.WriteLine("Signalling workers.");
startWorkers.Set(); // allow workers to run
startWorkers.Reset(); // workers can only run ONCE until we set again
Console.WriteLine("Waiting for all workers to be done...");
WaitHandle.WaitAll(workers); // wait for all the workers to be done
Console.WriteLine("All workers are done.");
worker1.Reset();
worker2.Reset();
}
}
static void DoWork1()
{
while (true)
{
Console.WriteLine("Worker1 waiting.");
startWorkers.WaitOne(); // wait here until signalled
// ... do something in here ...
Console.WriteLine("Worker1 processing...");
Thread.Sleep(R.Next(3000, 10000)); // random amount of work
Console.WriteLine("Worker1 done.");
worker1.Set(); // let Synchroniser know we are done
}
}
static void DoWork2()
{
while (true)
{
Console.WriteLine("Worker2 waiting.");
startWorkers.WaitOne(); // wait here until signalled
// ... do something in here ...
Console.WriteLine("Worker2 processing...");
Thread.Sleep(R.Next(3000, 10000)); // random amount of work
Console.WriteLine("Worker2 done.");
worker2.Set(); // let Synchroniser know we are done
}
}
}
Sample output:
Worker1 waiting.
Pausing...
Worker2 waiting.
Signalling workers.
Waiting for all workers to be done...
Worker2 processing...
Worker1 processing...
Worker1 done.
Worker1 waiting.
Worker2 done.
Worker2 waiting.
All workers are done.
Pausing...
Signalling workers.
Waiting for all workers to be done...
Worker2 processing...
Worker1 processing...
Worker1 done.
Worker1 waiting.
Worker2 done.
Worker2 waiting.
All workers are done.
Pausing...
If somehow the worker thread finishes before startWorkers.Reset();
I think it's unlikely to happen, but here's a two gate system that should prevent that from happening. After signalling that the thread is done, it will sit and wait until all threads have completed before being allowed to continue:
class Program
{
static readonly ManualResetEvent startWorkers = new ManualResetEvent(false); // workers initially blocked
static readonly ManualResetEvent releaseWorkers = new ManualResetEvent(false); // workers initially blocked
static readonly ManualResetEvent worker1 = new ManualResetEvent(false);
static readonly ManualResetEvent worker2 = new ManualResetEvent(false);
static readonly Random R = new Random();
static void Main()
{
Task.Run(new Action(Synchroniser));
Task.Run(new Action(DoWork1));
Task.Run(new Action(DoWork2));
Console.ReadLine(); // keep program from closing
}
static void Synchroniser()
{
ManualResetEvent[] workers = new ManualResetEvent[] { worker1, worker2 };
while (true)
{
Console.WriteLine("Pausing...");
Thread.Sleep(5000); // long pause so we can see what is happening
Console.WriteLine("Signalling workers.");
startWorkers.Set(); // allow workers to run, automatically reset
startWorkers.Reset();
Console.WriteLine("Waiting for all workers to be done...");
WaitHandle.WaitAll(workers); // wait for all the workers to be done
Console.WriteLine("All workers are done.");
Console.WriteLine("Resetting workers...");
// Reset all workers
foreach(ManualResetEvent mre in workers)
{
mre.Reset();
}
// release the workers to wait at the top of their loops
Console.WriteLine("Releasing workers...");
releaseWorkers.Set();
releaseWorkers.Reset();
}
}
static void DoWork1()
{
while (true)
{
Console.WriteLine("Worker1 waiting.");
startWorkers.WaitOne(); // wait here until signalled
// ... do something in here ...
Console.WriteLine("Worker1 processing...");
Thread.Sleep(R.Next(3000, 10000)); // random amount of work
Console.WriteLine("Worker1 done.");
worker1.Set(); // let Synchroniser know we are done
Console.WriteLine("Worker1 waiting for release...");
releaseWorkers.WaitOne(); // wait for all clear
}
}
static void DoWork2()
{
while (true)
{
Console.WriteLine("Worker2 waiting.");
startWorkers.WaitOne(); // wait here until signalled
// ... do something in here ...
Console.WriteLine("Worker2 processing...");
Thread.Sleep(R.Next(3000, 10000)); // random amount of work
Console.WriteLine("Worker2 done.");
worker2.Set(); // let Synchroniser know we are done
Console.WriteLine("Worker2 waiting for release...");
releaseWorkers.WaitOne(); // wait for all clear
}
}
}
You should have a look at the wait() function and such for the Task class.
It provides a way to continue execution of the current thread when a Task or a given array of tasks finishes execution.
For reference: https://learn.microsoft.com/en-gb/dotnet/api/system.threading.tasks.task.wait?view=netcore-3.1
So for example:
static void Main()
{
Task t1 = Task.Run(() => DoStuff_1("Task"));
Task t2 = Task.Run(() => DoStuff_2("Task"));
Task.WaitAll({t1, t2}); // Blocks the current thread till the given tasks finish execution
}
I have seen plenty of examples (here and elsewhere) of creating a non-reentrant timer by stopping the timer when the elapsed handler method is called and starting it again at the end of the elapsed handler method. This seems to be the recommended approach. The problem with this approach is that you will have a gap in time while the Elapsed Handler Method is running. You could end up with timing that is off by quite a lot within a short period of time.
So I was thinking about a better approach and I can up with the idea to use a bool to determine the state of the Timer, and whether the Elapsed Handler is currently running or not, it is is running then the call to the Elapsed Handler is returned immediately and the rest is not executed.
Below is the basic Idea
volatile bool _IsProcessingElapsedMethod = false;
private void _timer_Elapsed(object sender, ElapsedEventArgs e)
{
if (_IsProcessingElapsedMethod)
{
Console.WriteLine("Warning: Re-Entrance was attempted and Ignored.");
return;
}
_IsProcessingElapsedMethod = true;
//** DO Something here
_IsProcessingElapsedMethod = false;
}
There has to be a reason I have never seen anyone do this. Am I missing some obvious Gotcha? It seems like a pretty easy solution.
Below is a compilable example.
using System;
using System.Threading.Tasks;
using System.Timers;
namespace QuestionNon_ReEntrantTimer
{
class Program
{
static private int Timer1_ElapsedCount = 1;
static void Main(string[] args)
{
NonReEntrantTimer timer1 = new NonReEntrantTimer(500);
timer1.Elapsed += Timer1_Elapsed;
timer1.Start();
Console.WriteLine("Press Any key to Exit");
Console.ReadKey();
}
private static void Timer1_Elapsed(object sender, ElapsedEventArgs e)
{
int delayTime;
if(Timer1_ElapsedCount < 10)
{
delayTime = 300 * Timer1_ElapsedCount++;
}
else
{
Timer1_ElapsedCount++;
delayTime = 400;
}
Console.WriteLine($"Timer1_Elapsed Call Count is {Timer1_ElapsedCount} Waiting for {delayTime} ms");
Task.Delay(delayTime).Wait();
}
}
public class NonReEntrantTimer : IDisposable
{
Timer _timer = new Timer();
public event ElapsedEventHandler Elapsed;
volatile bool _IsProcessingElapsedMethod = false;
public NonReEntrantTimer(double interval)
{
_timer = new Timer(interval);
_timer.Elapsed += _timer_Elapsed;
}
public void Start() => _timer.Start();
public void Stop() => _timer.Stop();
public void Close() => _timer.Close();
private void _timer_Elapsed(object sender, ElapsedEventArgs e)
{
if (_IsProcessingElapsedMethod)
{
Console.WriteLine("Warning: Re-Entrance was attempted and Ignored.");
return;
}
_IsProcessingElapsedMethod = true;
Elapsed?.Invoke(sender, e);
_IsProcessingElapsedMethod = false;
}
public void Dispose()
{
_timer.Dispose();
}
}
}
I would propose this simple async pattern. It executes the given Action every ts, but starts countdown to the next execution before starting the current iteration. If the execution takes more time than ts, the next iteration is postponed till after the previous one finishes.
async Task ExecuteEvery(TimeSpan ts, Action a, CancellationToken ct)
{
try
{
var currentDelay = Task.Delay(ts, ct);
while (!ct.IsCancellationRequested)
{
await currentDelay; // waiting for the timeout
currentDelay = Task.Delay(ts, ct); // timeout finished, starting next wait
a(); // executing action in the meanwhile
}
}
catch (OperationCanceledException) when (ct.IsCancellationRequested)
{
// if we are cancelled, nothing to do, just exit
}
}
You can stop the iterations by cancelling the token. You can offload the action execution to the thread pool by starting the operation with Task.Run.
Update: if you want the timer to try catching up after the slow action, you can do it with some minor changes:
async Task ExecuteEvery(TimeSpan ts, Action a, CancellationToken ct)
{
try
{
for (var targetTime = DateTime.Now + ts; !ct.IsCancellationRequested; targetTime += ts)
{
var timeToWait = targetTime - DateTime.Now;
if (timeToWait > TimeSpan.Zero)
await Task.Delay(timeToWait, ct);
a();
}
}
catch (OperationCanceledException) when (ct.IsCancellationRequested)
{
// if we are cancelled, nothing to do, just exit
}
}
Consider the following:
//base stuff
private readonly ConcurrentQueue<message> queue = new ConcurrentQueue<message>();
private readonly MyCacheData _cache = new MyCacheData ();
//setuo
timer = new Timer { Interval = 60_000, AutoReset = true };
timer.Elapsed += OnTimedEvent;
httpClient.Timeout = new TimeSpan(0, 0, 60); // 60 seconds too
//
// each 60 seconds
private async void OnTimedEvent(object sender, ElapsedEventArgs e)
{
if (cache 30 minutes old)
{
//Fire and Forget GetWebDataAsync()
// and continue executing next stuff
// if I await it will wait 60 seconds worst case
// until going to the queue and by this time another
// timed even fires
}
// this always should execute each 60 seconds
if (queue isnt empty)
{
process queue
}
}
// heavy cache update each 10-30 minutes
private async Task GetWebDataAsync()
{
if (Semaphore.WaitAsync(1000))
{
try
{
//fetch WebData update cache
//populate Queue if needed
}
catch (Exception)
{
}
finally
{
release Semaphore
}
}
}
Colored: https://ghostbin.com/paste/6edov
Because I cheat and use the cheap ConcurrentQueue solution I don't really care much about what happens during GetWebDataAsync(), I just want to fire it and do its job, while I instantly go to process queue because it always must be done each 60 seconds or timer resolution.
How do I correctly do that, avoid much overhead or unnecessary thread spawning?
EDIT: got an answer for my case elsewhere
private async void OnTimedEvent(object sender, ElapsedEventArgs e)
{
async void DoGetWebData() => await GetWebDataAsync()
if (condition)
{
DoGetWebData(); // Fire&Forget and continue, exceptions handled inside
}
//no (a)waiting for the GetWebDataAsync(), we already here
if (queue isnt empty)
{
//process queue
}
}
private async Task GetWebDataAsync()
{
if (Semaphore.WaitAsync(1000))
{
try
{
//fetch WebData update cache
//populate Queue if needed
}
catch (Exception)
{
//log stuff
}
finally
{
///always release lock
}
}
}
Task.Run(...);
ThreadPool.QueueUserItem(...);
Anything wrong with these?...
How about something like that:
ManualResetEvent mre = new ManualResetEvent(false);
void Foo()
{
new Thread(() =>
{
while (mre.WaitOne())
{
/*process queue item*/
if (/*queue is empty*/)
{
mre.Reset();
}
}
}) { IsBackground = true }.Start();
}
void AddItem()
{
/*queue add item*/
mre.Set();
}
Call an async method from another async method without await statement
JS's setInterval and setTimeOut is really convenient. And I want to ask how to implement the same thing in C#.
You can just do a Task.Delay within a Task.Run, try out:
var task = Task.Run(async () => {
for(;;)
{
await Task.Delay(10000)
Console.WriteLine("Hello World after 10 seconds")
}
});
Then You could even wrap this up in to your own SetInterval method that takes in an action
class Program
{
static void Main(string[] args)
{
SetInterval(() => Console.WriteLine("Hello World"), TimeSpan.FromSeconds(2));
SetInterval(() => Console.WriteLine("Hello Stackoverflow"), TimeSpan.FromSeconds(4));
Thread.Sleep(TimeSpan.FromMinutes(1));
}
public static async Task SetInterval(Action action, TimeSpan timeout)
{
await Task.Delay(timeout).ConfigureAwait(false);
action();
SetInterval(action, timeout);
}
}
or you could just use the built in Timer class which practically does the same thing
static void Main(string[] args)
{
var timer1 = new Timer(_ => Console.WriteLine("Hello World"), null, 0, 2000);
var timer2 = new Timer(_ => Console.WriteLine("Hello Stackoverflow"), null, 0, 4000);
Thread.Sleep(TimeSpan.FromMinutes(1));
}
Just make sure you're timers don't go out of scope and get disposed.
.NET 6 Update
.NET 6 introduced a new type called PeriodicTimer this simplifies the above, and can be used like the following:
var timer = new PeriodicTimer(TimeSpan.FromSeconds(10));
while (await timer.WaitForNextTickAsync())
{
Console.WriteLine("Hello World after 10 seconds")
}
If you need to be able to cancel the timer the WaitForNextTickAsync function has an overload for
a cancellation token.
Its simply like this, you define an static System.Timers.Timer; then call the function that binds the timer.Elapsed event to your interval function that will be called each X miliseconds.
public class StaticCache {
private static System.Timers.Timer syncTimer;
StaticCache(){
SetSyncTimer();
}
private void SetSyncTimer(){
// Create a timer with a five second interval.
syncTimer = new System.Timers.Timer(5000);
// Hook up the Elapsed event for the timer.
syncTimer.Elapsed += SynchronizeCache;
syncTimer.AutoReset = true;
syncTimer.Enabled = true;
}
private static void SynchronizeCache(Object source, ElapsedEventArgs e)
{
// do this stuff each 5 seconds
}
}
I need a method to run accurately every 5 minutes. I can't use Timer because I noticed it will slowly become out of sync (i.e. it will eventually run at 00:01, 00:06, 00:11, 00:16, and so on).
Although it needs to be accurate, I don't need it to be too precise. Every 5 minutes +/- 1 second will be okay, just as long as after days of running, it will still tick accurately on the 5 minute marks.
What I have thought of so far is creating a Timer with an Interval of 1 second that constantly checks DateTime.Now to see if the next 5 minute mark is passed. I am wondering if there is a more elegant solution or something in the C# libraries that I have missed.
Edit: I have the following template now, which is working to my requirements.
public class ThreadTest
{
private Thread thread;
private long nextExecutionTime;
private long interval;
public void StartThread(long intervalInMillis)
{
interval = intervalInMillis * TimeSpan.TicksPerMillisecond;
nextExecutionTime = DateTime.Now.Ticks;
thread = new Thread(Run);
thread.Start();
}
private void Run()
{
while (true)
{
if (DateTime.Now.Ticks >= nextExecutionTime)
{
nextExecutionTime += interval;
// do stuff
}
}
}
}
if you are not happy with Timer?
then you can try to make your thread sleep for 5 mintues, instead of using Timer
have a look this, hope it helps
using System;
using System.Threading;
public class Worker
{
// This method will be called when the thread is started.
public void DoWork()
{
while (!_shouldStop)
{
Task.Factory.Start(() =>
{
// do you task async
})
Thread.Sleep(300000);
}
}
public void DoWork2()
{
var watch = new Stopwatch();
while (!_shouldStop)
{
watch.Start();
Task.Factory.Start(() =>
{
// do you task async
})
while(watch.Elapsed.ElapsedMilliseconds < 300000);
watch.Stop();
watch.Reset();
}
}
public void RequestStop()
{
_shouldStop = true;
}
private volatile bool _shouldStop;
}
public class WorkerThreadExample
{
static void Main()
{
// Create the thread object. This does not start the thread.
Worker workerObject = new Worker();
Thread workerThread = new Thread(workerObject.DoWork);
// Start the worker thread.
workerThread.Start();
// Loop until worker thread activates.
while (!workerThread.IsAlive);
while (true)
{
//do something to make it break
}
// Request that the worker thread stop itself:
workerObject.RequestStop();
workerThread.Join();
}
}
or you can try this: