Java Threads, 3rd Edition

Chapter 15. Parallelizing Loops for Multiprocessor Machines

In previous chapters, we examined threading as a technique that allows us to simplify programming: we used threading to achieve asynchronous behavior and perform independent tasks. Although we discussed how threads are scheduled on machines with multiple processors, by and large the techniques that we’ve shown so far are not affected by a machine with multiple processors nor do they exploit the number of processors on a machine to make the program run faster.

Multithreaded programs have a special bond with multiprocessor systems. The separation of threads provides a clear and simple separation for the multiprocessor machine. Since the operating system can place different threads on different processors, the program runs faster.

In this chapter, we’ll look at how to parallelize Java programs so that they run faster on a machine with multiple CPUs. The processes that we examine are beneficial not only to newly developed Java programs but also to existing Java programs that have a CPU-intensive loop, allowing us to improve the performance of those programs on a multiprocessor system.

How does the Java threading system behave in a multiprocessor system? There are no conceptual differences between a program running on a machine with one processor and a machine with two or more processors; threads behave exactly the same in either case. The real difference is that the threads actually do execute simultaneously. In Chapter 2, we ...