NSGA-II-Based Design Space Exploration for Energy and Throughput Aware Multicore Architectures. Issue 6 (3rd October 2017)

• Record Type: Journal Article
• Authors: Hussain, Ishfaq; Parveen, Abida; Ahmad, Ayaz; Qadri, Muhammad Yasir; Qadri, Nadia N.; Ahmed, Jameel
• Is Part Of: Cybernetics and systems. Volume 48:Issue 6/7(2017)
• Page Start: 536
• Page End: 550
• Publication Date: 2017-10-03
• Subjects: Design space exploration -- energy -- multicore -- optimization -- throughput; Cybernetics -- Periodicals; System theory -- Periodicals; 003.5
• DOI: 10.1080/01969722.2017.1402433
• Languages: English
• ISSNs: 0196-9722

Abstract

ABSTRACT: Multicore architectures are mainstream due to ever increasing demand of throughput by modern applications. However, the suboptimal utilization of available resources in these architectures may imply an inevitable energy overhead. This energy overhead can only be avoided if the multicore systems support reconfiguration of available resources as per application demand. To achieve the target objectives (i.e., Energy efficiency with Throughput maximization) in multicore systems, many decision variables need to be optimized or analyzed to find the better trade-off. Heuristic-based approaches are aimed to provide a good-enough solution instead of a lengthy exhaustive search. This paper presents an Evolutionary Algorithm (EA)-based approach, i.e., Nondominated Sorting Genetic Algorithm-II (NSGA-II). Three decision variables, i.e., number of cores, cache size and frequency are used to find best solution. The proposed approach is validated over a set of parallel benchmarks using a cycle accurate simulator. The results show a significant amount of energy saving along with minimal impact on the throughput of the system.