A Java virtual machine (JVM) is a virtual machine that enables a computer to run Java programs as well as programs written in other languages, referred to as JVM languages, that are also compiled to Java bytecode. The JVM is detailed by a specification that formally describes what is required in a JVM implementation. Having a specification ensures interoperability of Java programs across different implementations so that program authors using the Java Development Kit (JDK) need not worry about idiosyncrasies of the underlying hardware platform.

The JVM reference implementation is developed by the OpenJDK project as open source code and includes a JIT compiler called HotSpot. The commercially supported Java releases available from Oracle are based on the OpenJDK runtime. Eclipse OpenJ9 is another open source JVM for OpenJDK.

JVM specification The Java virtual machine is an abstract (virtual) computer defined by a specification. It is a part of the Java Runtime Environment. The garbage collection algorithm used and any internal optimization of the Java virtual machine instructions (their translation into machine code) are not specified. The main reason for this omission is to not unnecessarily constrain implementers. Any Java application can be run only inside some concrete implementation of the abstract specification of the Java virtual machine.Bill Venners, Inside the Java Virtual Machine Chapter 5

Starting with Java Platform, Standard Edition (J2SE) 5.0, changes to the JVM specification have been developed under the Java Community Process as JSR 924. , changes to the specification to support changes proposed to the class file format (JSR 202) are being done as a maintenance release of JSR 924. The specification for the JVM was published as the blue book,The Java Virtual Machine Specification (the first and second editions are also available online). whose preface states:

The most commonly used Java virtual machine is Oracle's HotSpot.

Oracle owns the Java trademark and may allow its use to certify implementation suites as fully compatible with Oracle's specification.

Garbage collectors

Garbage collectors available in Java OpenJDKs virtual machine (JVM) include: Serial Parallel CMS (Concurrent Mark Sweep) G1 (Garbage-First) ZGC (Z Garbage Collector) Epsilon Shenandoah GenZGC (Generational ZGC) GenShen (Generational Shenandoah) IBM Metronome (only in IBM OpenJDK) SAP (only in SAP OpenJDK) Azul C4 (Continuously Concurrent Compacting Collector) (only in Azul Systems OpenJDK)

+ Java versions and their Garbage Collectors Version Default GC Available GCs 6u14 Serial /Parallel (MP) Serial, Parallel, CMS, G1 (E) 7u4–8 Serial, Parallel, CMS, G1 9–10 Serial / G1 11 Serial, Parallel, CMS, G1, Epsilon (E), ZGC (E) 12–13 Serial, Parallel, CMS, G1, Epsilon (E), ZGC (E), Shenandoah (E) 14 Serial, Parallel, G1, Epsilon (E), ZGC (E), Shenandoah (E) 15–20 Serial, Parallel, G1, Epsilon (E), ZGC, Shenandoah 21–22 Serial, Parallel, G1, Epsilon (E), ZGC, Shenandoah, GenZGC (E) 23 Serial, Parallel, G1, Epsilon (E), ZGC, Shenandoah, GenZGC (default ZGC) 24 Serial, Parallel, G1, Epsilon (E), Shenandoah, GenZGC, GenShen (E) 25-26 Serial, Parallel, G1, Epsilon (E), Shenandoah, GenZGC, GenShen 27 G1 Serial, Parallel, G1, Epsilon (E), Shenandoah, GenZGC, GenShen 28 Serial, Parallel, G1, Epsilon (E), Shenandoah, GenZGC, GenShen (default Shenandoah) (E) = experimental

Java's design philosophy revolves around the assumption of a garbage collector. Unlike languages such as C++ and Rust, deterministic memory management through a delete keyword (as in C++) is not possible. Even introducing such a feature is not possible, due to a lack of ownership, aside from using or through to allocate/deallocate memory outside the Java heap. Due to Java primarily using heap-based allocation, objects are stored as references, and deletion would result in dangling pointers. Foo a = new Foo(); Foo b = a;

delete a; System.out.println(b);

Virtual machine architecture

The JVM operates on specific types of data as specified in Java Virtual Machine specifications. The data types can be divided into primitive types (integer and floating-point values) and reference types. long and double types, which are 64-bits, are supported natively, but consume two units of storage in a frame's local variables or operand stack, since each unit is 32 bits. boolean, byte, short, and char types are all sign-extended (except char which is zero-extended) and operated on as 32-bit integers, the same as int types. The smaller types only have a few type-specific instructions for loading, storing, and type conversion. boolean is operated on as 8-bit byte values, with 0 representing false and 1 representing true. (Although boolean has been treated as a type since The Java Virtual Machine Specification, Second Edition clarified this issue, in compiled and executed code there is little difference between a boolean and a byte except for name mangling in method signatures and the type of Boolean arrays. booleans in method signatures are mangled as Z while bytes are mangled as B. Boolean arrays carry the type boolean[] but use 8 bits per element, and the JVM has no built-in capability to pack booleans into a bit array, so except for the type they perform and behave the same as byte arrays. In all other uses, the boolean type is effectively unknown to the JVM as all instructions to operate on Booleans are also used to operate on bytes.)

The JVM has a garbage-collected heap for storing objects and arrays. Code, constants, and other class data are stored in the "method area". The method area is logically part of the heap, but implementations may treat the method area separately from the heap, and for example might not garbage collect it. Each JVM thread also has its own call stack (called a "Java Virtual Machine stack" for clarity), which stores frames. A new frame is created each time a method is called, and the frame is destroyed when that method exits.

Each frame provides an "operand stack" and an array of "local variables". The operand stack is used for operands to run computations and for receiving the return value of a called method, while local variables serve the same purpose as registers and are also used to pass method arguments. Thus, the JVM is both a stack machine and a register machine. In practice, HotSpot eliminates every stack besides the native thread/call stack even when running in Interpreted mode, as its Templating Interpreter technically functions as a compiler.

The JVM uses references and stack/array indexes to address data; it does not use byte addressing like most physical machines do, so it does not neatly fit the usual categorization of 32-bit or 64-bit machines. In one sense, it could be classified as a 32-bit machine, since this is the size of the largest value it natively stores: a 32-bit integer or floating-point value or a 32-bit reference. Because a reference is 32 bits, each program is limited to at most 2 unique references and therefore at most 2 objects. However, each object can be more than one byte large, and potentially very large; the largest possible object is an array of long of length 2 - 1 which would consume 16 GiB of memory, and there could potentially be 2 of these if there were enough memory available. This results in upper bounds that are more comparable to a typical 64-bit byte-addressable machine. A JVM implementation can be designed to run on a processor that natively uses any bit width as long as it correctly implements the integer (8-, 16-, 32-, and 64-bit) and floating-point (32- and 64-bit) math that the JVM requires. Depending on the method used to implement references (native pointers, compressed pointers, or an indirection table), this can limit the number of objects to less than the theoretical maximum. An implementation of the JVM on a 64-bit platform has access to a much larger address space than one on a 32-bit platform, which allows for a much larger heap size and an increased maximum number of threads, which is needed for certain kinds of large applications; however, there can be a performance hit from using a 64-bit implementation compared to a 32-bit implementation.

JVM languages

A JVM language is any language with functions that can be expressed in a valid class file, which can be hosted by a JVM. A class file contains JVM instructions (JVM bytecode), a symbol table, and other ancillary information. The class file format is the hardware- and operating system-independent binary format used to represent compiled classes and interfaces.

There are several JVM languages, both older established languages ported to JVM, and much newer from-scratch languages. JRuby and Jython may be the best known ports of older languages, i.e., Ruby and Python respectively. Of the new from-scratch languages created to compile to Java bytecode, Clojure, Groovy, Scala and Kotlin may be the most popular. A notable feature of the JVM languages is their language interoperability, they are compatible with each other, so that, for example, Scala libraries can be used with Java programs and vice versa.

Java 7 JVM implements JSR 292: Supporting Dynamically Typed Languages on the Java Platform, a new feature which supports dynamically typed languages in the JVM. This feature is developed within the Da Vinci Machine project whose mission is to extend the JVM so that it supports languages other than Java.

Class loader

One of the organizational units of JVM byte code is a class. A class loader implementation must be able to recognize and load anything that conforms to the Java class file format. Any implementation is free to recognize other binary forms besides class files, but it must recognize class files.

The class loader performs three basic activities in this strict order:

Loading: finds and imports the binary data for a type Linking: performs verification, preparation, and (optionally) resolution Verification: ensures the correctness of the imported type Preparation: allocates memory for class variables and initializing the memory to default values Resolution: transforms symbolic references from the type into direct references. Initialization: invokes Java code that initializes class variables to their proper starting values.

In general, there are three types of class loader: bootstrap class loader, extension class loader and System / Application class loader.

Every Java virtual machine implementation must have a bootstrap class loader that is capable of loading trusted classes, as well as an extension class loader or application class loader. The Java virtual machine specification does not specify how a class loader should locate classes.

Bytecode instructions

The JVM has instructions for the following groups of tasks:

The aim is binary compatibility. Each particular host operating system needs its own implementation of the JVM and runtime. These JVMs interpret the bytecode semantically the same way, but the actual implementation may be different. More complex than just emulating bytecode is compatibly and efficiently implementing the Java core API that must be mapped to each host operating system.

These instructions operate on a set of common rather the native data types of any specific instruction set architecture.

Bytecode verifier