Energy consumption analysis of parallel sorting algorithms running on multicore systems

Sorting algorithms have been ubiquitously used in numerous applications nowadays. As the data size scales up exponentially, energy-efficiency is gradually becoming equally important as performance for sorting algorithms that process large scale data. Unfortunately, the energy consumption behaviour of various sorting algorithms is not fully explored, although the performance of sorting algorithms have been well studied. Will the sorting algorithms consume more energy when they are parallelized and executed on multicore computers? How to improve the energy-efficient performance of sorting algorithms? In this paper, we comprehensively analyze the performance and energy consumption of three traditional sorting algorithms, Odd-Even Sort, ShellSort and QuickSort, on a multicore system. Our experimental results show that algorithms with better performance tend to conserve energy as well, when evaluated on a computer with eight AMD cores. In addition, for the same algorithm, we observe that more energy savings may be achieved when task granularity is properly selected.