Collaborative Research: A Multi-Element and Multi- Objective
Optimization Approach for Allocating Tasks to Multi-Core Processors
(funded by the National Science Foundation)
The objective of the proposed research is to design innovative algorithms and tools for energy-aware scheduling and mapping of tasks onto homogeneous and heterogeneous multi-core processor (HeMP) architectures.
We propose to develop a new theoretical and experimental framework called multi-element and multi-objective (MEMO) optimization that will simultaneously and flexibly optimize the goals of energy minimization and performance maximization while taking into account constraints due to multiple architectural elements such as cores, caches, etc. of current and emerging multi-core processors.
We will develop CorePac, a toolkit that will provide a flexible and friendly environment to schedule task-parallel applications on HeMPs under various performance/energy trade-offs.
To demonstrate the usefulness of the algorithms and CorePac, we will benchmark our algorithms on a diverse suite of scientific, multimedia, and bioinformatics applications.
Through its production of new algorithms and software toolkit, this work will have a direct and immediate impact on a number of communities. At our home institutions, this project will have an educational impact by involving undergraduate and graduate students. This situation also presents excellent opportunities for interaction with postdoctoral researchers as well as with colleagues in academic, government and industry research labs.
The CorePac software toolkit will be the basis for subsequent development of production quality software for energy-performance tradeoffs. Developing means to manage energy consumption in computers is imperative from both environmental and economical perspectives. Reductions in energy consumption of multi-core processors will contribute to system-wide energy and cost savings.