Automatic GPU Offloading Technology for Open IoT Environment

Abstract

Recently, Internet of Things (IoT) technologies have been progressed. To overcome of the high cost of developing IoT services by vertically integrating devices and services, open IoT enables various IoT services to be developed by integrating horizontally separated devices and services. For open IoT, we have proposed Tacit Computing technology to discover the devices that have data users need on demand and use them dynamically. However, existing Tacit Computing does not consider performance. Therefore, in this paper, we propose an automatic graphics processing unit (GPU) offloading technology as a new elementary technology of Tacit Computing that uses genetic algorithm to extract appropriate offloading areas from parallelizable loop statements automatically. This can improve performance of IoT applications. We evaluate our proposed GPU offloading technology by applying it to five C/C++ applications of image processing, matrix manipulation, and so on to verify its effectiveness and find that it can process them more than ten times as quickly as only using central processing units within 1 h tuning time.