Optimizing Storage Space on Embedded Systems Using Superinstructions

Abstract

Embedded systems are gaining popularity as the devices get smarter and customized with time. But still, embedded systems have limitations in their available resources such as storage and processing capability. This study investigates the role of superinstructions in optimizing the storage space of Java-based applications for embedded systems. JVM must also be optimized to meet embedded systems’ limited resources. This research introduces the JEOpt framework (Java Embedded Optimization framework) to systematically analyze Java bytecode, identify basic blocks, compute effectiveness of superinstructions, and subsequently modify the bytecode with superinstructions based on the effectiveness computed by the metric ESI. Java applications including xml, sunflow, scimark, mpegaudio, derby, crypto, compress, and compiler available in SPECjvm2008 benchmark were examined in this study. Prototype results show that the proposed approach achieves significant reductions in bytecode size up to 31.13% and effectiveness up to 32.3 for the Java applications executing on embedded systems.