Intel Threading Building Blocks

Chapter 2. Thinking Parallel

This chapter is about how to "Think Parallel.” It is a brief introduction to the mental discipline that helps you make a program parallelizable in a safe and scalable manner. Even though Intel Threading Building Blocks does much of the work that traditional APIs require the programmer to do, this kind of thinking is a fundamental requirement to writing a good parallel program.

This is the place in the book that defines terms such as scalability and provides the motivation for focusing on these concepts. The topics covered in this chapter are decomposition, scaling, threads, correctness, abstraction, and patterns.

If you don’t already approach every computer problem with parallelism in your thoughts, this chapter should be the start of a new way of thinking. If you are already deeply into parallel programming, this chapter can still show how Threading Building Blocks deals with the concepts.

How should we think about parallel programming? This is now a common question because, after decades in a world where most computers had only one central processing unit (CPU), we are now in a world where only “old” computers have one CPU. Multi-core processors are now the norm. Parallel computers are the norm. Therefore, every software developer needs to Think Parallel.

Today, when developers take on a programming job, they generally think about the best approach before programming. We already think about selecting the best algorithm, implementation language, and so ...