Why GPGPU? What Is Heterogeneous Computing?

As developers, we are used to adjusting to a changing world. Our industry changes the world almost as a matter of routine. We learn new languages, adopt new methodologies, start using new user interface paradigms, and take for granted that it will always be possible to make our programs better. When it seems we will “hit a wall” following one path to making version n+1 better than version n, we find another path. The newest path some developers are about to follow is the path of heterogeneous computing.

In this chapter you’ll review some of the history of performance improvements to see what wall some developers are facing. You’ll learn the basic differences between a CPU and a GPU, two of the possible components of a heterogeneous computing solution, and what kinds of problems are suitable for acceleration using these parallel techniques. Then you’ll review the CPU and GPU parallel techniques in use today, followed by an introduction to the concepts behind C++ AMP, to lay the groundwork for the details in the subsequent chapters.