Parallel Programming and
Multithreading in C#
Filip Ekberg
Principal Consultant & CEO
@fekberg fekberg.com
Using the Task from the Task Parallel Library
Task.Run(() => {
// This code will execute on a different context
});
The Task from the Task Parallel Library
Only executes on one thread
Task.Run(() => {
});
Break down a large problem
and compute each
piece independently
Task Parallel Library
await Task.Run(() => {
// I’m an asynchronous operation that is awaited
});
Parallel.Invoke(
() => { /* Parallel Thread 1 */ },
() => { /* Parallel Thread 2 */ },
() => { /* Parallel Thread 3 */ },
() => { /* Parallel Thread 4 */ },
);
Running Work on Another Thread
Task.Run(() => {
var msft = Calculate(“MSFT”);
var googl = Calculate(“GOOGL”);
var ps = Calculate(“PS”);
var amaz = Calculate(“AMAZ”);
return new [] { msft, googl, ps, amaz };
});
Running Work on Another Thread
Task.Run(() => {
var msft = Calculate(“MSFT”);
Goal is to run
var googl = Calculate(“GOOGL”); these on 4
var ps = Calculate(“PS”); different threads
var amaz = Calculate(“AMAZ”);
return new [] { msft, googl, ps, amaz };
});
Get help from the
framework to optimize the
parallel operation
Parallel Programming in .NET
Thread Task Parallel Library
Task Parallel Library
and its Task should be the
preferred way to introduce
parallel programming
Task.Run(() => {
});
No need to care about lower-level threads
Work may be scheduled on a new, or reused thread.
Parallel Programming with Task Parallel Library
Parallel.Invoke(
() => {},
() => {},
() => {}
);
Parallel.For(0, 10, (index) => {});
Parallel.ForEach(source, (element) => {});
Parallel Programming with Task Parallel Library
Parallel.Invoke(
() => {},
() => {},
() => {}
);
Parallel.For(0, 10, (index) => {});
Parallel.ForEach(source, (element) => {});
Parallel Programming with Task Parallel Library
Parallel.Invoke(
() => {},
() => {},
() => {}
);
Parallel.For(0, 10, (index) => {});
Parallel.ForEach(source, (element) => {});
Task Parallel Library
provides a way to write
Parallel LINQ (PLINQ)
Parallel (Extensions)
Built on-top of the Task in the Task Parallel Library
A Problem to Solve in Parallel
Parallel will ensure that work
is distributed efficiently on
the system that runs the
application
When to Use Parallel Programming
CPU bound operations Independent chunks of data
Use the Parallel Methods
Task.Run(() => { });
Task.Run(() => { });
Task.Run(() => { });
Task.Run(() => { });
Parallel.For(0, 4, (i) => {});
Automatically distribute the work
There’s no guarantee that the
operations will run in parallel
Your First Parallel Operation
The Parallel Methods Blocks the Calling Thread
Parallel.Invoke(...);
Block the calling thread until all the
parallel operations completed
Parallel.For(...);
Parallel.ForEach(...);
Deadlocking with the Parallel Class
Parallel.For(0, 4, (index) => {
Dispatcher.Invoke(() => {
// Run on the UI Thread
});
}); This causes a deadlock!
By default calling these
Parallel methods will
consume as much computer
power as possible
Parallel Invoke
Parallel.Invoke(
new ParallelOptions { MaxDegreeOfParallelism = 2
}
() => { /* Parallel Thread 1 */ },
() => { /* Parallel Thread 2 */ },
() => { /* Parallel Thread 3 */ },
() => { /* Parallel Thread 4 */ }
);
Parallel Invoke with Max Degree of Parallelism
Parallel.Invoke(
new ParallelOptions { MaxDegreeOfParallelism = 2
}
() => { /* Parallel Thread 1 */ },
() => { /* Parallel Thread 2 */ },
() => { /* Parallel Thread 1 */ },
() => { /* Parallel Thread 2 */ }
);
Misusing Parallel in ASP.NET
can cause bad performance
for all users!
Next: Using Parallel and Asynchronous
Principles Together
Using Parallel and Asynchronous
Principles Together
Task Parallel Library
await Task.Run(() => {
// I’m an asynchronous operation that is awaited
});
Parallel.Invoke(
() => { /* Parallel Thread 1 */ },
() => { /* Parallel Thread 2 */ },
() => { /* Parallel Thread 3 */ },
() => { /* Parallel Thread 4 */ },
);
Don’t reinvent this,
use the Task Parallel
Library!
Next: Handling Exceptions
Handling Exceptions
Handling Exceptions
Parallel.Invoke(...);
Automatically validates the
parallel operations.
Parallel.For(...);
Parallel.ForEach(...);
This Will Throw an Aggregate Exception
Parallel.Invoke(
() => { throw new Exception(“1”); },
() => { throw new Exception(“2”); },
() => { throw new Exception(“3”); },
() => { throw new Exception(“4”); },
);
Not yet executed parallel
operations will not be
cancelled just because one
operation fails
Next: Processing a Collection of Data
in Parallel
Processing a Collection of Data in Parallel
Normal Foreach vs Parallel.ForEach
foreach(var element in source)
{
// Execute sequentially
}
Parallel.ForEach(source, (element) => {
// Execute in parallel
});
Normal Foreach vs Parallel.ForEach
foreach(var element in source)
{
// Execute sequentially
}
Parallel.ForEach(source, (element) => {
// Execute in parallel
});
Automatically distributed
work that runs in parallel
The performance benefits
will be more obvious with
larger collections to
process
Break won’t automatically
stop
running operations
Example: ParallelLoopState.Break()
Parallel.For(0, 100, (i, state) => {
if(i == 50) Scheduled iterations for
{ indices lower than 50 will
state.Break(); still start!
}
Only operations for indices
over 50 won’t be
});
scheduled to start
Normal For vs Parallel.For
for(int i = 0; i < 10; i++)
{
// Execute sequentially
}
Parallel.For(0, 10, (i) => {
// Execute in parallel
});
Automatically distributed
work that runs in parallel
Example: Parallel.For
Parallel.For(0, 10, (i, state) => {
});
Example: Parallel.For
Inclusive
Parallel.For(0, 10, (i, state) => {
Exclusive
});
Creating Parallel Operations
Parallel.Invoke(...);
Parallel.For(...);
Parallel.ForEach(...);
Parallel.ForEach
Parallel.ForEach(source, (element) => {
// Execute in parallel
});
Automatically distributed
work that runs in parallel
The Parallel Methods Blocks the Calling Thread
Parallel.Invoke(...);
Block the calling thread until all the
parallel operations completed
Parallel.For(...);
Parallel.ForEach(...);
Parallel + Asynchronous
await Task.Run(() => { Parallel.Invoke(...); });
await Task.Run(() => { Parallel.For(...); });
await Task.Run(() => { Parallel.ForEach(...); });
Implications of parallelism
Difference and similarities between parallel
and asynchronous programming
Summary Builds on-top of the Task in the Task
Parallel Library
Works in any C# and .NET application
Every problem and machine won’t benefit
from parallelism
Break down a problem in small pieces and
solve them independently
Use thread-safe collections like
ConcurrentBag<T>
