Optimise web browsing on heterogeneous mobile platforms: A machine learning based approach

摘要: Web browsing is an activity that billions of mobile users perform on a daily basis. Battery life primary concern to many who often find their phone has died at most inconvenient times. The heterogeneous multi-core architecture solution for energy-efficient processing. However, the current web browsers rely operating system exploit underlying hardware, which no knowledge individual contents and leads poor energy efficiency. This paper describes automatic approach render workloads performance It achieves this by developing machine learning based predict processor use run rendering engine what frequencies processors should operate. Our predictor learns offline from set training workloads. built then integrated into browser optimal configuration runtime, taking account workload characteristics optimisation goal: whether it load time, consumption or trade-off between them. We evaluate our representative ARM big.LITTLE using hottest 500 webpages. 80% delivered ideal predictor. obtain, average, 45%, 63.5% 81% improvement respectively delay product, when compared Linux multi-processing scheduler.