Java is both a compiled and an interpreted language. Java source code is turned into simple binary instructions, much like ordinary microprocessor machine code. However, whereas C or C++ source is reduced to native instructions for a particular model of processor, Java source is compiled into a universal format—instructions for a virtual machine.
Compiled Java bytecode is executed by a Java runtime interpreter. The runtime system performs all the normal activities of a hardware processor, but it does so in a safe, virtual environment. It executes a stack-based instruction set and manages memory like an operating system. It creates and manipulates primitive data types and loads and invokes newly referenced blocks of code. Most importantly, it does all this in accordance with a strictly defined open specification that can be implemented by anyone who wants to produce a Java-compliant virtual machine. Together, the virtual machine and language definition provide a complete specification. There are no features of the base Java language left undefined or implementation-dependent. For example, Java specifies the sizes and mathematical properties of all its primitive data types rather than leaving it up to the platform implementation.
The Java interpreter is relatively lightweight and small; it can be implemented in whatever form is desirable for a particular platform. The interpreter may be run as a separate application or it can be embedded in another piece of software, such as a web browser. Put together, this means that Java code is implicitly portable. The same Java application bytecode can run on any platform that provides a Java runtime environment, as shown in Figure 1-1. You don’t have to produce alternative versions of your application for different platforms, and you don’t have to distribute source code to end users.
The fundamental unit of Java code is the class. As in other object-oriented languages, classes are application components that hold executable code and data. Compiled Java classes are distributed in a universal binary format that contains Java bytecode and other class information. Classes can be maintained discretely and stored in files or archives locally or on a network server. Classes are located and loaded dynamically at runtime as they are needed by an application.
In addition to the platform-specific runtime system, Java has a number of fundamental classes that contain architecture-dependent methods. These native methods serve as the gateway between the Java virtual machine and the real world. They are implemented in a natively compiled language on the host platform and provide low-level access to resources such as the network, the windowing system, and the host filesystem. The vast majority of Java, however, is written in Java itself—bootstrapped from these basic primitives—and is therefore portable. This includes fundamental Java tools such as the Java compiler, networking, and GUI libraries, which are also written in Java and are therefore available on all Java platforms in exactly the same way without porting.
Historically, interpreters have been considered slow, but Java is not a traditional interpreted language. In addition to compiling source code down to portable bytecode, Java has also been carefully designed so that software implementations of the runtime system can further optimize their performance by compiling bytecode to native machine code on the fly. This is called just-in-time (JIT) or dynamic compilation. With JIT compilation, Java code can execute as fast as native code and maintain its transportability and security.
This is an often misunderstood point among those who want to compare language performance. There is only one intrinsic performance penalty that compiled Java code suffers at runtime for the sake of security and virtual machine design—array bounds checking. Everything else can be optimized to native code just as it can with a statically compiled language. Going beyond that, the Java language includes more structural information than many other languages, providing for more types of optimizations. Also remember that these optimizations can be made at runtime, taking into account the actual application behavior and characteristics. What can be done at compile time that can’t be done better at runtime? Well, there is a tradeoff: time.
The problem with a traditional JIT compilation is that optimizing code takes time. So a JIT compiler can produce decent results, but may suffer a significant latency when the application starts up. This is generally not a problem for long-running server-side applications, but is a serious problem for client-side software and applications that run on smaller devices with limited capabilities. To address this, Java’s compiler technology, called HotSpot, uses a trick called adaptive compilation. If you look at what programs actually spend their time doing, it turns out that they spend almost all their time executing a relatively small part of the code again and again. The chunk of code that is executed repeatedly may be only a small fraction of the total program, but its behavior determines the program’s overall performance. Adaptive compilation also allows the Java runtime to take advantage of new kinds of optimizations that simply can’t be done in a statically compiled language, hence the claim that Java code can run faster than C/C++ in some cases.
To take advantage of this fact, HotSpot starts out as a normal Java bytecode interpreter, but with a difference: it measures (profiles) the code as it is executing to see what parts are being executed repeatedly. Once it knows which parts of the code are crucial to performance, HotSpot compiles those sections into optimal native machine code. Since it compiles only a small portion of the program into machine code, it can afford to take the time necessary to optimize those portions. The rest of the program may not need to be compiled at all—just interpreted—saving memory and time. In fact, the Java VM can run in one of two modes: client and server, which determine whether it emphasizes quick startup time and memory conservation or flat out performance.
A natural question to ask at this point is, Why throw away all this good profiling information each time an application shuts down? Well, Sun partially broached this topic with the release of Java 5.0 through the use of shared, read-only classes that are stored persistently in an optimized form. This significantly reduced both the startup time and overhead of running many Java applications on a given machine. The technology for doing this is complex, but the idea is simple: optimize the parts of the program that need to go fast and don’t worry about the rest.