The Integration of GPU-based and Heterogeneous Devices Using HLA

Abstract

The use of distributed components may offer many advantages in the most types of applications. It includes, for example, the easy use of disperse resources in a network, allowing a potential gain on the processing power and data loading. But, the major challenge is interoperability and communication management. In this aspect, the High Level Architecture (HLA) establishes specific policies for data exchanging and time management of communication in the integration of the simulation components. Thus, this paper aims to explain and analyse the feasibility to integrate devices based on heterogeneous architectures using distributed and parallel computing, multi-core CPU and GPU, with OpenCL technology. For that, the use of the IEEE 1516 standard, called High Level Architecture, is proposed targeting this integration and allowing the use of all computing power from available devices in a distributed system. In this study experiments also were done to demonstrate the approach of running a simple parallel image processing application under the conceived architecture.