Java Bytecode Compression for Embedded Systems

A program executing on an embedded system or similar environment faces limited memory resources and fixed time constrains. We demonstrate how factorization of common instruction sequences can be automatically applied to Java bytecode programs. Based on a series of experiments, we argue that program size is reduced by 30% on the average, typically with an execution time penalty of less than 30%. The one-time, minor modifications necessary to make a standard Java interpreter compatible with this factorized code are presented on the Harissa virtual machine, together with an algorithm for performing the factorization of Java bytecode.