Self-optimisation using runtime code generation for wireless sensor networks

International audienceThis paper addresses the use of runtime code specialisation in resource-constrained embedded systems such as nodes of a Wireless Sensor Network (WSN), in order to improve software efficiency, hence the lifetime of WSN nodes. In our approach, runtime code specialisation is achieved with in-place runtime code generation. We present a self-optimising system using runtime code generation. Our system is able to automatically make the decision to generate specialised code and use it each time an improvement is observed in application performance. In the Internet of Things (IoT), devices usually have limited precision; our system adapts to theses devices decreasing precision in order to increase performance. We evaluate our system on floating point multiplication using the WisMote platform, where the specialised code executes more than 7 times faster than generic code, all overheads included. To the best of our knowledge, it is the first time that a runtime code generation system is used to automatically optimise code in such constrained devices as WSN nodes