The dataflow pattern is a generalized pattern with a one-to-many and a many-to-one relationship. For example, the following diagram represents two tasks, Task 1 and Task 2, that execute in parallel, and a third task, Task 3, that will only start when both of the first two tasks are completed. Once Task 3 is completed, Task 4 and Task 5 will be executed in parallel:

We can implement the preceding example using the following code:

static void Main(string[] args) 
{ 
  //Creating two tasks t1 and t2 and starting them at the same //time
  Task<int> t1 = Task.Factory.StartNew(() => { return Task1(); }); 
  Task<int> t2 = Task.Factory.StartNew(() => { return Task2(); }); 

  //Creating task 3 and used ContinueWhenAll that runs when both the 
  //tasks T1 and T2 will be completed
  Task<int> t3 = Task.Factory.ContinueWhenAll(
  new[] { t1, t2 }, (tasks) => { return Task3(); }); 

  //Task 4 and Task 5 will be started when Task 3 will be completed. 
  //ContinueWith actually creates a continuation of executing tasks 
  //T4 and T5 asynchronously when the task T3 is completed
  Task<int> t4 = t3.ContinueWith((antecendent) => { return Task4(); }); 
  Task<int> t5 = t3.ContinueWith((antecendent) => { return Task5(); }); 
  Console.Read(); 
} 
//Implementation of Task1
public static int Task1() 
{ 
  Thread.Sleep(1000); 
  Console.WriteLine("Task 1 is executed"); 
  return 1; 
} 

//Implementation of Task2 
public static int Task2() 
{ 
  Thread.Sleep(1000); 
  Console.WriteLine("Task 2 is executed"); 
  return 1; 
} 
//Implementation of Task3 
public static int Task3() 
{ 
  Thread.Sleep(1000); 
  Console.WriteLine("Task 3 is executed"); 
  return 1; 
} 
Implementation of Task4
public static int Task4() 
{ 
  Thread.Sleep(1000); 
  Console.WriteLine("Task 4 is executed"); 
  return 1; 
} 
 
//Implementation of Task5
public static int Task5() 
{ 
  Thread.Sleep(1000); 
  Console.WriteLine("Task 5 is executed"); 
  return 1; 
}