Traditionally, parallel programming was always implemented with a focus on threads and data sharing. The only support we programmers got from the operating system and the language runtime libraries were thread—and synchronization—related functions. We were able to create a thread, maybe set some thread parameters (like thread priority), and kill a thread. We were also able to create some synchronization mechanisms—a critical section, mutex, or a semaphore. But that was all.

As you are not skipping ahead and you read the previous two chapters, you already know that being able to start a new thread and do the locking is not nearly enough. Writing parallel code that way is a slow, error-prone process. That's why in the last decade the focus in parallel code has shifted from threads to tasks and patterns. Everyone is doing it—Microsoft with the .NET Task Parallel Library, Intel with Thread Building Blocks, Embarcadero with the  Parallel Programming Library and third-party extensions...