Efficient Workload Balancing on Heterogeneous GPUs using Mixed- Integer Non-Linear Programming

Main Article Content

Chih-Sheng Lin
Chih-Wei Hsieh
Hsi-Ya Chang
Pao-Ann Hsiung

Abstract

Recently, heterogeneous system architectures are becoming mainstream for achieving high performance and powerefficiency. In particular, many-core graphics processing units (GPUs) now play an important role for computing inheterogeneous architectures. However, for application designers, computational workload still needs to be distributedto heterogeneous GPUs manually and remains inefficient. In this paper, we propose a mixed integer non-linearprogramming (MINLP) based method for efficient workload distribution on heterogeneous GPUs by consideringasymmetric capabilities of GPUs for various applications. Compared to the previous methods, the experimental resultsshow that our proposed method improves performance and balance up to 33% and 116%, respectively. Moreover, ourmethod only requires a few overhead while achieving high performance and load balancing.

 

PLUM ANALYTICS

Article Details

How to Cite
Lin, C.-S., Hsieh, C.-W., Chang, H.-Y., & Hsiung, P.-A. (2014). Efficient Workload Balancing on Heterogeneous GPUs using Mixed- Integer Non-Linear Programming. Journal of Applied Research and Technology, 12(6). https://doi.org/10.1016/S1665-6423(14)71676-1