Study to achieve environment adaptive software

TL;DR

This paper proposes environment adaptive software that automatically optimizes application code for heterogeneous hardware like GPUs, FPGAs, and IoT devices, simplifying development and deployment.

Contribution

It introduces a framework for automatically converting application code and configuring settings to adapt to various hardware environments.

Findings

Proposed processing flow for environment adaptation

Elemental technologies enabling automatic code conversion

Demonstrated high performance across heterogeneous hardware

Abstract

Recently, heterogeneous hardware such as GPU and FPGA is used in many systems and also IoT devices are increased repidly. However, to utilize heterogeneous hardware, the hurdles are high because of much technical skills. In order to break down such a situation, we think it is required in the future that application programmers only need to write logics to be processed, then software will adapt to the environments with heterogeneous hardware, to make it easy to utilize heterogeneous hardware and IoT devices. Therefore, in this paper, we propose environment adaptive software to operate an once written application with high performance by automatically converting the code and configuring setting so that we can utilize GPU, FPGA and IoT devices in the location to be deployed. We explain a proceesing flow and elemental technologies to achieve environment adaptive software.