Introduction

This chapter introduces the book by first giving an account of the historic events that pushed heterogeneous parallel computing into the mainstream. It distinguishes between latency-oriented multicore CPUs and throughput oriented many-thread GPUs, the two main styles of computing devices in modern heterogeneous computing systems. The chapter then goes into the main reasons for applications to demand more speed and the Amdahl’s law, an important observation about the parallelization of real application. It then explains how the book addresses the main challenges in parallel algorithms and parallel programming and how the skills that have been learned from the book based on CUDA, the language of choice for programming examples, and exercises in this book can be generalized into other parallel programming interfaces. Finally, the chapter highlights the overarching goals for the book and the contents and rationales of the rest of the chapters.