Threading
Remarks#
A thread is a part of a program that can execute independently of other parts. It can perform tasks simultaneously with other threads. Multithreading is a feature that enables programs to perform concurrent processing so that more than one operation can be done at a time.
For example, you can use threading to update a timer or counter in the background while simultaneously performing other tasks in the foreground.
Multithreaded applications are more responsive to user input and are also easily scalable, because the developer can add threads as and when the workload increases.
By default, a C# program has one thread - the main program thread. However, secondary threads can be created and used to execute code in parallel with the primary thread. Such threads are called worker threads.
To control the operation of a thread, the CLR delegates a function to the operating system known as Thread Scheduler. A thread scheduler assures that all the threads are allocated proper execution time. It also checks that the threads that are blocked or locked do not consume much of the CPU time.
The .NET Framework System.Threading namespace makes using threads easier. System.Threading enables multithreading by providing a number of classes and interfaces. Apart from providing types and classes for a particular thread, it also defines types to hold a collection of threads, timer class and so on. It also provides its support by allowing synchronized access to shared data.
Thread is the main class in the System.Threading namespace. Other classes include AutoResetEvent, Interlocked, Monitor, Mutex, and ThreadPool.
Some of the delegates that are present in the System.Threading namespace include
ThreadStart, TimerCallback, and WaitCallback.
Enumerations in System.Threading namespace include ThreadPriority, ThreadState,
and EventResetMode.
In .NET Framework 4 and later versions, multithreaded programming is made easier and simpler through the System.Threading.Tasks.Parallel and System.Threading.Tasks.Task classes, Parallel LINQ (PLINQ), new concurrent collection classes in the System.Collections.Concurrent namespace, and a new task-based programming model.
Simple Complete Threading Demo
class Program
{
static void Main(string[] args)
{
// Create 2 thread objects. We're using delegates because we need to pass
// parameters to the threads.
var thread1 = new Thread(new ThreadStart(() => PerformAction(1)));
var thread2 = new Thread(new ThreadStart(() => PerformAction(2)));
// Start the threads running
thread1.Start();
// NB: as soon as the above line kicks off the thread, the next line starts;
// even if thread1 is still processing.
thread2.Start();
// Wait for thread1 to complete before continuing
thread1.Join();
// Wait for thread2 to complete before continuing
thread2.Join();
Console.WriteLine("Done");
Console.ReadKey();
}
// Simple method to help demonstrate the threads running in parallel.
static void PerformAction(int id)
{
var rnd = new Random(id);
for (int i = 0; i < 100; i++)
{
Console.WriteLine("Thread: {0}: {1}", id, i);
Thread.Sleep(rnd.Next(0, 1000));
}
}
}Simple Complete Threading Demo using Tasks
class Program
{
static void Main(string[] args)
{
// Run 2 Tasks.
var task1 = Task.Run(() => PerformAction(1)));
var task2 = Task.Run(() => PerformAction(2)));
// Wait (i.e. block this thread) until both Tasks are complete.
Task.WaitAll(new [] { task1, task2 });
Console.WriteLine("Done");
Console.ReadKey();
}
// Simple method to help demonstrate the threads running in parallel.
static void PerformAction(int id)
{
var rnd = new Random(id);
for (int i = 0; i < 100; i++)
{
Console.WriteLine("Task: {0}: {1}", id, i);
Thread.Sleep(rnd.Next(0, 1000));
}
}
}Explicit Task Parallism
private static void explicitTaskParallism()
{
Thread.CurrentThread.Name = "Main";
// Create a task and supply a user delegate by using a lambda expression.
Task taskA = new Task(() => Console.WriteLine($"Hello from task {nameof(taskA)}."));
Task taskB = new Task(() => Console.WriteLine($"Hello from task {nameof(taskB)}."));
// Start the task.
taskA.Start();
taskB.Start();
// Output a message from the calling thread.
Console.WriteLine("Hello from thread '{0}'.",
Thread.CurrentThread.Name);
taskA.Wait();
taskB.Wait();
Console.Read();
}Implicit Task Parallelism
private static void Main(string[] args)
{
var a = new A();
var b = new B();
//implicit task parallelism
Parallel.Invoke(
() => a.DoSomeWork(),
() => b.DoSomeOtherWork()
);
}Creating and Starting a Second Thread
If you’re doing multiple long calculations, you can run them at the same time on different threads on your computer. To do this, we make a new Thread and have it point to a different method.
using System.Threading;
class MainClass {
static void Main() {
var thread = new Thread(Secondary);
thread.Start();
}
static void Secondary() {
System.Console.WriteLine("Hello World!");
}
}Starting a thread with parameters
using System.Threading;
class MainClass {
static void Main() {
var thread = new Thread(Secondary);
thread.Start("SecondThread");
}
static void Secondary(object threadName) {
System.Console.WriteLine("Hello World from thread: " + threadName);
}
}Creating One Thread Per Processor
Environment.ProcessorCountGets the number of logical processors on the current machine.
The CLR will then schedule each thread to a logical processor, this theoretically could mean each thread on a different logical processor, all threads on a single logical processor or some other combination.
using System;
using System.Threading;
class MainClass {
static void Main() {
for (int i = 0; i < Environment.ProcessorCount; i++) {
var thread = new Thread(Secondary);
thread.Start(i);
}
}
static void Secondary(object threadNumber) {
System.Console.WriteLine("Hello World from thread: " + threadNumber);
}
}Avoiding Reading and Writing Data Simultaneously
Sometimes, you want your threads to simultaneously share data. When this happens it is important to be aware of the code and lock any parts that could go wrong. A simple example of two threads counting is shown below.
Here is some dangerous (incorrect) code:
using System.Threading;
class MainClass
{
static int count { get; set; }
static void Main()
{
for (int i = 1; i <= 2; i++)
{
var thread = new Thread(ThreadMethod);
thread.Start(i);
Thread.Sleep(500);
}
}
static void ThreadMethod(object threadNumber)
{
while (true)
{
var temp = count;
System.Console.WriteLine("Thread " + threadNumber + ": Reading the value of count.");
Thread.Sleep(1000);
count = temp + 1;
System.Console.WriteLine("Thread " + threadNumber + ": Incrementing the value of count to:" + count);
Thread.Sleep(1000);
}
}
}You’ll notice, instead of counting 1,2,3,4,5… we count 1,1,2,2,3…
To fix this problem, we need to lock the value of count, so that multiple different threads cannot read and write to it at the same time. With the addition of a lock and a key, we can prevent the threads from accessing the data simultaneously.
using System.Threading;
class MainClass
{
static int count { get; set; }
static readonly object key = new object();
static void Main()
{
for (int i = 1; i <= 2; i++)
{
var thread = new Thread(ThreadMethod);
thread.Start(i);
Thread.Sleep(500);
}
}
static void ThreadMethod(object threadNumber)
{
while (true)
{
lock (key)
{
var temp = count;
System.Console.WriteLine("Thread " + threadNumber + ": Reading the value of count.");
Thread.Sleep(1000);
count = temp + 1;
System.Console.WriteLine("Thread " + threadNumber + ": Incrementing the value of count to:" + count);
}
Thread.Sleep(1000);
}
}
}Parallel.ForEach Loop
If you have a foreach loop that you want to speed up and you don’t mind what order the output is in, you can convert it to a parallel foreach loop by doing the following:
using System;
using System.Threading;
using System.Threading.Tasks;
public class MainClass {
public static void Main() {
int[] Numbers = new int[] { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 };
// Single-threaded
Console.WriteLine("Normal foreach loop: ");
foreach (var number in Numbers) {
Console.WriteLine(longCalculation(number));
}
// This is the Parallel (Multi-threaded solution)
Console.WriteLine("Parallel foreach loop: ");
Parallel.ForEach(Numbers, number => {
Console.WriteLine(longCalculation(number));
});
}
private static int longCalculation(int number) {
Thread.Sleep(1000); // Sleep to simulate a long calculation
return number * number;
}
}Deadlocks (two threads waiting on eachother)
A deadlock is what occurs when two or more threads are waiting for eachother to complete or to release a resource in such a way that they wait forever.
A typical scenario of two threads waiting on eachother to complete is when a Windows Forms GUI thread waits for a worker thread and the worker thread attempts to invoke an object managed by the GUI thread. Observe that with this code exmaple, clicking button1 will cause the program to hang.
private void button1_Click(object sender, EventArgs e)
{
Thread workerthread= new Thread(dowork);
workerthread.Start();
workerthread.Join();
// Do something after
}
private void dowork()
{
// Do something before
textBox1.Invoke(new Action(() => textBox1.Text = "Some Text"));
// Do something after
}workerthread.Join() is a call that blocks the calling thread until workerthread completes.
textBox1.Invoke(invoke_delegate) is a call that blocks the calling thread until the GUI thread has processed invoke_delegate, but this call causes deadlocks if the GUI thread is already waiting for the calling thread to complete.
To get around this, one can use a non-blocking way of invoking the textbox instead:
private void dowork()
{
// Do work
textBox1.BeginInvoke(new Action(() => textBox1.Text = "Some Text"));
// Do work that is not dependent on textBox1 being updated first
}However, this will cause trouble if you need to run code that is dependent on the textbox being updated first. In that case, run that as part of the invoke, but be aware that this will make it run on the GUI thread.
private void dowork()
{
// Do work
textBox1.BeginInvoke(new Action(() => {
textBox1.Text = "Some Text";
// Do work dependent on textBox1 being updated first,
// start another worker thread or raise an event
}));
// Do work that is not dependent on textBox1 being updated first
}Alternatively start af whole new thread and let that one do the waiting on the GUI thread, so that workerthread might complete.
private void dowork()
{
// Do work
Thread workerthread2 = new Thread(() =>
{
textBox1.Invoke(new Action(() => textBox1.Text = "Some Text"));
// Do work dependent on textBox1 being updated first,
// start another worker thread or raise an event
});
workerthread2.Start();
// Do work that is not dependent on textBox1 being updated first
}To minimize the risk of running into a deadlock of mutual waiting, always avoid circular references between threads when possible. A hierarchy of threads where lower-ranking threads only leave messages for higher-ranking threads and never waiting on them will not run into this kind of issue. However, it would still be vulnerable to deadlocks based on resource locking.
Deadlocks (hold resource and wait)
A deadlock is what occurs when two or more threads are waiting for eachother to complete or to release a resource in such a way that they wait forever.
If thread1 holds a lock on resource A and is waiting for resource B to be released while thread2 holds resource B and is waiting for resource A to be released, they are deadlocked.
Clicking button1 for the following example code will cause your application to get into aforementioned deadlocked state and hang
private void button_Click(object sender, EventArgs e)
{
DeadlockWorkers workers = new DeadlockWorkers();
workers.StartThreads();
textBox.Text = workers.GetResult();
}
private class DeadlockWorkers
{
Thread thread1, thread2;
object resourceA = new object();
object resourceB = new object();
string output;
public void StartThreads()
{
thread1 = new Thread(Thread1DoWork);
thread2 = new Thread(Thread2DoWork);
thread1.Start();
thread2.Start();
}
public string GetResult()
{
thread1.Join();
thread2.Join();
return output;
}
public void Thread1DoWork()
{
Thread.Sleep(100);
lock (resourceA)
{
Thread.Sleep(100);
lock (resourceB)
{
output += "T1#";
}
}
}
public void Thread2DoWork()
{
Thread.Sleep(100);
lock (resourceB)
{
Thread.Sleep(100);
lock (resourceA)
{
output += "T2#";
}
}
}
}To avoid being deadlocked this way, one can use Monitor.TryEnter(lock_object, timeout_in_milliseconds) to check if a lock is held on an object already. If Monitor.TryEnter does not succeed in acquiring a lock on lock_object before timeout_in_milliseconds, it returns false, giving the thread a chance to release other held resources and yielding, thus giving other threads a chance to complete as in this slightly modified version of the above:
private void button_Click(object sender, EventArgs e)
{
MonitorWorkers workers = new MonitorWorkers();
workers.StartThreads();
textBox.Text = workers.GetResult();
}
private class MonitorWorkers
{
Thread thread1, thread2;
object resourceA = new object();
object resourceB = new object();
string output;
public void StartThreads()
{
thread1 = new Thread(Thread1DoWork);
thread2 = new Thread(Thread2DoWork);
thread1.Start();
thread2.Start();
}
public string GetResult()
{
thread1.Join();
thread2.Join();
return output;
}
public void Thread1DoWork()
{
bool mustDoWork = true;
Thread.Sleep(100);
while (mustDoWork)
{
lock (resourceA)
{
Thread.Sleep(100);
if (Monitor.TryEnter(resourceB, 0))
{
output += "T1#";
mustDoWork = false;
Monitor.Exit(resourceB);
}
}
if (mustDoWork) Thread.Yield();
}
}
public void Thread2DoWork()
{
Thread.Sleep(100);
lock (resourceB)
{
Thread.Sleep(100);
lock (resourceA)
{
output += "T2#";
}
}
}
}Note that this workaround relies on thread2 being stubborn about its locks and thread1 being willing to yield, such that thread2 always take precedence. Also note that thread1 has to redo the work it did after locking resource A, when it yields. Therefore be careful when implementing this approach with more than one yielding thread, as you’ll then run the risk of entering a so-called livelock - a state which would occur if two threads kept doing the first bit of their work and then yield mutually, starting over repeatedly.