Optimizing Java

Chapter 9. Code Execution on the JVM

The two main services that any JVM provides are memory management and an easy-to-use container for execution of application code. We covered garbage collection in some depth in Chapters 6 through 8, and in this chapter we turn to code execution.

Note

Recall that the Java Virtual Machine specification, usually referred to as the VMSpec, describes how a conforming Java implementation needs to execute code.

The VMSpec defines execution of Java bytecode in terms of an interpreter. However, broadly speaking, interpreted environments have unfavorable performance as compared to programming environments that execute machine code directly. Most production-grade modern Java environments solve this problem by providing dynamic compilation capability.

As we discussed in Chapter 2, this ability is otherwise known as Just-in-Time compilation, or just JIT compilation. It is a mechanism by which the JVM monitors which methods are being executed in order to determine whether individual methods are eligible for compilation to directly executable code.

In this chapter we start by providing a brief overview of bytecode interpretation and why HotSpot is different from other interpreters that you may be familiar with. We then turn to the basic concepts of profile-guided optimization. We discuss the code cache and then introduce the basics of HotSpot’s compilation subsystem.

In the following chapter we will explain the mechanics behind some of HotSpot’s most common ...