Amortised optimisation of non-functional properties in production environments

Abstract

Search Based Software Engineering has high potential for optimising non-functional properties such as execution time or power consumption. However, many non-functional properties are dependent not only on the software system under consideration but also the environment that surrounds the system. This necessitates a support for online, in situ optimisation. This paper introduces the novel concept of amortised optimisation which allows such online optimisation. The paper also presents two case studies: one that seeks to optimise JIT compilation, and another to optimise a hardware dependent algorithm. The results show that, by using the open source libraries we provide, developers can improve the speed of their Python script by up to 8.6% with virtually no extra effort, and adapt a hardware dependent algorithm automatically for unseen CPUs.