Performance Analysis of ARM big.LITTLE Architecture Based Mobile Processor with Multi-Thread Face Detection

Abstract

Because of the power consumption properties of the heterogeneous processors, they have also started to be used with the high level mobile devices. However, working characteristics of the heterogeneous architectures equipped with different core clusters can not be easily guessed when more than one cores are activated. In this study, the multi-core performance of an ARM big.LITTLE architecture based mobile processor was investigated by solving face detection problem. The results of the experimental studies showed that fluctuations of the average core frequency are limited when the number of utilized cores is less than the number of cores in the selected cluster. In addition to this, it was concluded that the total required time for completion of the execution decreases when the number of utilized cores exceeds the number of cores in the selected cluster but fluctuations of the average core frequency are on the increasing trend.