Java bytecode compilation for high-performance, platform-independent logical inference.

Abstract

Automated reasoning systems are powerful computer programs capable of solving complex problems. They are characterized as computationally intensive having high performance requirements. Very few reasoning systems have been implemented in Java so far; its performance is regarded as an impediment to its use as a programming language for computationally intensive applications such as automated reasoning. In this thesis we discuss techniques that motivate the use of Java as the underlying platform to design a framework for high-performance logical inference. The techniques are centered around the idea of using a specialized compiler that can generate Java classes which contain Java bytecodes customized for performing reasoning efficiently. The benefit of generating bytecodes customized for logical inference is reflected in the improved performance observed from the experiments conducted.