< Back to previous page

Project

Optimized Dynamic Memory Management Techniques on Heterogeneous Memory Systems

In this PhD thesis, the basic goal is to implement and evaluate Dynamic Memory Management (DMM) frameworks and techniques in embedded devices and systems. More specifically, firstly, we give emphasis on the understanding of how memory management systems work and behave. DMM techniques help embedded devices with limited capabilities to perform indirectly complex calculations that may be requested by users. Additionally, the complete comprehension of applications and how these affect dynamic memory behavior is of major importance. The first stage of the work is an analysis of the existing literature in order to identify possible deficiencies or optimizations that could be applied for memory management resources in embedded environments and software and hardware assumptions which influence that. Then we will move on to an implementation of a dynamic memory manager. The role of that manager will be to decide dynamically how the data of the executed application will be handled, aiming at a multiparametric optimization of different metrics, such as memory footprint, memory accesses, battery lifetime of devices, total energy consumption and overall system performance. Finally, after the manager is finished, we will go further to optimize application execution through dynamic memory management and evaluate manager techniques on a variety of different innovative hardware platform technologies and memories. The DMM techniques will help us to accelerate the execution of applications as well as minimize consumed energy on built-in devices.

Date:1 Jan 2021 →  Today
Keywords:Memory Management, Memory Optimizations
Disciplines:Memory management
Project type:PhD project