High Performance Heterogeneous Multicore Architectures: A Study

Abstract

The significant increase in the need for high-performance and energy-efficient computing systems has introduced heterogenous computing. However, the incorporation of different architectures into one system complicates the distribution of the workload between architectures. To address this challenge while meeting the goals of high-performance computing systems, several research contributions have been made. This paper reviews some of the proposed workload partitioning approaches for GPU-based, DSP-based and FPGA-based heterogenous systems. This research also covers some comparison studies regarding the FPGA versus DSP and FPGA versus GPU debates, showing that sometimes collaboration between these architectures seems to be the key. The aim of this study is to provide academic and industrial researchers with an insight of techniques to achieve the workload balancing in heterogenous systems and motivate them for further research in the field.