| Person in charge of the project: |
Scenario Based Resource Management of Processing Elements.|
As the dynamism of applications to be executed and the complexity of the hardware platforms increase, the run-time resource management for guaranteed application timing becomes more challenging. It has been well demonstrated that the complexity of cost-effective resource management requires mixed design-time and run-time approach interfaced with Pareto surfaces, e.g. IMEC's TCM (Task Concurrency Management) and MPSoC Run-time Management.|
The layered approach of handling platform complexity, such as allocation of memory and communication resources before processing elements, introduces more timing and precedence constraints that need to be handled by the PE resource management.
Existing resource management algorithms are very basic, in the sense they do not handle such realistic timing and precedence constraints among tasks and do not consider the operating point switching overhead. Also there has been no quantitative evaluation of different heuristics for pareto selection and alternative strategies to handle shared resource introduced interferences.
Along with all these aspects, the application dynamism makes this problem even hard, and the existing system scenario based methodology can be applied, but it requires systematic scenarios identification, detection, and exploitation mechanisms to handle the dynamism for heterogeneous platform mapping.
In this work we focus on aforementioned aspects to improve TCM further to make it more practically usable, and propose a systematic system scenario methodology to handle the resource management issues for dynamic applications.
Duration of the project:
26.11.2007 - 26.11.2011