What is the difference between DVM and JVM?

DVM is Register based which is designed to run on low memory, uses its own byte code and runs .Dex file

JVM is Stack based which uses java byte code and runs .class file having JIT.

Java source code is compiled by the Java compiler into .class files. Then the dx (dexer) tool, part of the Android SDK processes the .class files into a file format called DEX that contains Dalvik bytecode. The dx tool eliminate all the redundant information that is present in the classes. In DEX all the classes of the application are packed into one file. DVM has been designed so that a device can run multiple instances of the VM efficiently.

stack-based machines must use instructions to load data on the stack and manipulate that data, and, thus, require more instructions than register machines to implement the same high level code, but the instructions in a register machine must encode the source and destination registers and, therefore, tend to be larger.

Conceptually, there is little difference from an application level between a DVM and a JVM. Architecturally, there is a major difference between the registerbased DVM and the stack-based JVM.

Both use a VM code model. However, the DVM uses registerbased opcodes that are comparable to the register-based bytecode instructions that most of the target platforms already execute. This includes architectures such as those available from ARM and MIPS and the x86-compatible architectures from Intel, AMD, and VIA Technologies.

Google developed Android and chose DVM for several reasons. First, there were licensing issues with most JVMs. Next, the DVM should be more efficient in terms of memory usage and performance on a register-based machine. DVM is also supposed to be more efficient when running multiple instances of the DVM. Applications are given their own instance. Hence, multiple active applications require multiple DVM instances. Like most Java implementations, the DVM has an automatic garbage collector.

More about it