Work in progress: analysis and evaluation of the impact of the code approximation for IoT applications




Approximate Computing, Energy Efficiency, Internet of Things, Software Approximations


Due to the need to improve resource management for computer systems in many levels and applications (mainly for embedded systems and energy consumption), how can we enhance the energy efficiency of computational methods? One approach is through approximate computing, which intentionally introduces controlled errors to save resources such as energy, area, or time. This research aims to empirically measure the impact of introducing approximations in an embedded system by conducting a controlled experiment. To focus on evaluating the impact of the approximations themselves rather than the best methods of implementing them, the approximations will be manually incorporated into the code. The benchmark chosen for evaluation is MiBench due to its widespread usage. All the codes can be recompiled to run on the MIPS architecture of the NodeMCU-ESP8266. A second NodeMCU-ESP8266 will be utilized, connected in series to measure the actual power consumption of the first board. The analysis of results will involve hypothesis tests, where the experiment hypotheses will be statistically evaluated at a specific significance level. By directly comparing variations and experiment data, the proposal's validity will be effectively demonstrated. Since this paper is a work in progress, we will explain the experiment planned to be run.




Como Citar

Medeiros Cruz, D. e Almeida, T. 2023. Work in progress: analysis and evaluation of the impact of the code approximation for IoT applications. Academic Journal on Computing, Engineering and Applied Mathematics. 4, 2 (out. 2023), 17–20. DOI:



Edição Especial