The impact of varying memory region number on RTSJ execution time

Abstract

Developing a real time, embedded system in Java requires awareness of memory behaviour in addition to program logic. Region-based memory management has certain advantages over garbage collection in terms of predictability; however, the use of regions is in many ways arbitrary, since it is left to the developer to decide on the number of required regions prior to program execution. In theory, a memory model with a large number of regions will negatively impact the execution time of software since the run-time system has the overhead of managing the space in each of these regions. In this paper, we explore the effect of varying the number of regions on the performance of RTSJ execution times. RTSJ code was used to allocate varying numbers of objects (100 to 2500) into regions and then execution times were recorded. Results suggest that more regions do actually lead to increases in execution time. By applying a relatively simple refactoring to the original code, an improvement in execution times was achieved.