Investigating Software Deployment on Heterogeneous Platforms

Devices with computing capabilities are everywhere. Physical components that once were restricted to its forms and mechanics to provide functionality have now gained an enormous number of possibilities through the implementation of Embedded Systems (ES). However, together with the advantages of ES come the challenges of dealing with both the increasing complexity of software and the different constraints of hardware. Moreover, it is becoming a common practice to deploy software on computation units of different types (i.e. CPUs, GPUs and FPGAs) in order to optimize data processing and resource utilization. This paper reports on a PhD research project in the area of software deployment on heterogeneous platforms, focusing on how different deployment strategies contribute to achieve certain non-functional properties, such as performance and energy efficiency. As it is important to understand previous practices and acknowledge possible gaps in the research area, a systematic literature review is currently being conducted, and its initial/planning stages are described. Further, recent contributions in the context of data marshalling are also discussed, as well as the planned next steps and the impact of this project in an industrial context.