Customized And Runtime Adaptive, Multi-threaded Dynamic Memory Management For Embedded Systems

Abstract

The appropriate dynamic memory manager selection has a huge impact on metrics, such as the memory fragmentation, accesses and footprint. These metrics are related with the energy consumption and the overall system performance, which are significant factors for embedded systems. Furthermore, the advance in the embedded computing domain, breaks new ground in the development of applications with main characteristics the extremely complex source code and the intense requirement and constrain variations at runtime. Therefore, emerges as an urgent need the adaptation ability of the memory manager, being either being user driven or system driven, via system signals that it may emit.The purpose of the current diploma thesis is the development of a software library along with the appropriate automation tools for the generation of dynamic memory managers for embedded systems. The programming language that was preferred is Ansi-C, because it iswidespread in the domain of embedded computing. The dynamic memory manager design space was developed in a tunable way so that various but distinct solutions could be explored at design time. In addition, the library was enriched with management mechanisms that allow the memory managers to be runtime adaptive. Thus, the proposed design methodology produces application specific dynamic memory managers that take into account both the static and the dynamicapplication behavior.