FINCH (Fertile Darwinian Bytecode Harvester) is a system designed to evolutionarily improve actual extant software, which was not intentionally written for the purpose of serving as a GP representation in particular, nor for evolution in general. The only requirement is that the software source code be either written in Java or can be compiled to Java bytecode. The following chapter provides an overview of FINCH, ending with a prΓ©cis of results. Additional information can be found in [6,7].
Java compilers typically do not produce machine code directly, but instead compile source-code files to platformindependent bytecode, to be interpreted in software or, rarely, to be executed in hardware by a Java Virtual Machine (JVM) [4]. The JVM is free to apply its own optimization techniques, such as Just-in-Time (JIT) on-demand compilation Java compilation to native machine code-a process that is transparent to the user. The JVM implements a stackbased architecture with high-level language features such as object management and garbage collection, virtual function calls, and strong typing. The bytecode language itself is a well-designed assembly-like language with a limited yet powerful instruction set [3, 4]. Figure 1 shows a recursive Java program for computing the factorial of a number, and its corresponding bytecode.