Dynamic Data Type Optimization and Memory Assignment Methodologies

Abstract

It is evident that most new computing platforms are becoming more and more complex encapsulating multiple cores and reconfigurable elements. This offers the designers a multitude of resources. It is their responsibility to exploit the available resources in such a way to efficiently implement their applications. Furthermore, the complexity of the applications that run on such platforms is increasing as well. Applications running on such platforms need to cope with dynamic events and have their resource requirements vary during the execution time. In order to cope with this dynamism applications rely in the usage of dynamic data. Applications use containers such as dynamic data types in order to store and retrieve these dynamic data. In this work a set of methodologies that is able to optimize the containers holding the dynamic data and efficiently assign them on the available memory resources is presented. The proposed approach is evaluated in a scheduler for an IEEE802.16-based broadband wireless telecom system and a 3D game application, achieving reductions in the memory energy consumption of 32% and 51% respectively.