Async & TPL

• Async : Responsiveness (backend)
• Parallel : Performance (divide workload)
• Risk:
  • Race Conditions
  • Dependency -> ensure that you wait for the dependent task to have the required data

Task

• Creating a Task in parallel (depends on # of cores also)
  • Single core: threads are shared - Still best for responsiveness (ex: main thread is for UI)
  • Multi Core: threads run in parallel
    • Only the main thread/ UI thread can update UI

using System.Threading.Tasks;

Task t = new Task( () => { 
  // code
});
t.Start();

// another way to create and start task
task tnew = task.Factory.StartNew( _code_ );


// To Wait for the task completion
tnew.Wait();

• Continue With

Task t = new Task( () => { 
  // code
});

Task t2 = t.ContinueWith( (antecedent) => { } ); // starts after T is finished
t.Start();

• Safe Way (typesafe, reduce the race condition issues)

Task<int> t = Task.Factory.StartNew( () => 
{
  // ...
  return 1;
});
int r = t.Result;      // implicitlly calls t.Wait()

• WAIT

Task t = new Task( () => { 
  // code
});
t.Start();
t.Wait();

int r = t.Result; // implicitlly calls t.Wait()


// Multiple tasks
Task t1 = Task.Factory.startNew( code );
Task t2 = Task.Factory.startNew( code );
Task t3 = Task.Factory.startNew( code );

Task[] tasks = {t1, t2, t3 };
Task.WaitAll( tasks ); // wait for ALL to finish

int index = Task.WaitAny( tasks );  // wait for FIRST to finish
Task completedTask = tasks[index];

// Wait All One By One - (when some may fail | hide latency)
List<task<TResult> tasks = new List<task<TResult>();
for (int i =0 ....)
  tasks.Add( Task.Factory.StartNew( code ) );

while (tasks.Count > 0)
{
  int index = Task.WaitAny( tasks.ToArray() );
  // ... process tasks[index].Result
  tasks.removeAt(index);
}