Energy-Efficiency Prediction of Multithreaded Workloads on Heterogeneous Composite Cores Architectures using Machine Learning Techniques

TL;DR

This paper presents a machine learning-based method to predict optimal configurations for multithreaded workloads on heterogeneous composite cores architectures, aiming to maximize energy-efficiency through concurrent tuning of multiple parameters.

Contribution

It introduces a systematic, machine learning-driven approach for predicting energy-efficient configurations for multithreaded applications on CCAs, considering multiple tuning parameters simultaneously.

Findings

ML models accurately predict optimal configurations

Significant energy savings achieved with the approach

Effective at runtime scheduling for multithreaded workloads

Abstract

Heterogeneous architectures have emerged as a promising alternative for homogeneous architectures to improve the energy-efficiency of computer systems. Composite Cores Architecture (CCA), a class of dynamic heterogeneous architectures enabling the computer system to construct the right core at run-time for each application by composing cores together to build larger core or decomposing a large core into multiple smaller cores. While this architecture provides more flexibility for the running application to find the best run-time settings to maximize energy-efficiency, due to the interdependence of various tuning parameters such as the type of the core, run-time voltage and frequency and the number of threads, it makes it more challenging for scheduling. Prior studies mainly addressed the scheduling problem in CCAs by looking at one or two of these tuning parameters. However, as we will show in this paper, it is important to concurrently optimize and fine-tune these parameters. In addition, most previous works on CCA mainly study traditional single threaded CPU applications. This paper describes a systematic approach to predict the right configurations for running multithreaded workloads on CCAs. It achieves this by developing a machine learning-based approach to predict core type, voltage and frequency setting to maximize the energy-efficiency. Our predictor learns offline from an extensive set of training multithreaded workloads. It is then applied to predict the optimal processor configuration at run-time by taking into account the multithreaded application characteristics and the optimization objective. For this purpose, five well-known machine learning models are implemented for energy-efficiency optimization and precisely compared in terms of accuracy and hardware overhead to guide the scheduling decisions in a CCA.