[Zurück]

E. Shamsa, A. Kanduri, A. Rahmani, P. Liljeberg, A. Jantsch, N. Dutt:
"Goal-driven autonomy for efficient on-chip resource management: Transforming objectives to goals.";
in: "2019 Design, Automation & Test in Europe Conference & Exhibition (DATE)", 2019 Design, Automation & Test in Europe Conference & Exhibition (DATE); IEEE (Hrg.); herausgegeben von: IEEE; IEEE, Florence, Italy, 2019, ISBN: 978-3-9819263-2-3, S. 1 - 6.

-Run-time resource allocation of heterogeneousmulti-core systems is challenging with varying workloads andlimited power and energy budgets. User interaction within thesesystems changes the performance requirements, often conflictingwith concurrent applications´ objective and system constraints.Current resource allocation approaches focus on optimizing fixedobjective, ignoring the variation in system and applications´objective at run-time. For an efficient resource allocation, thesystem has to operate autonomously by formulating a hier-archy of goals. We present goal-driven autonomy (GDA) foron-chip resource allocation decisions, which allows systems togenerate and prioritize goals in response to the workload andsystem dynamic variation. We implemented a proof-of-conceptresource management framework that integrates the proposedgoal management control to meet power, performance and userrequirements simultaneously. Experimental results on an Exynosplatform containing ARM´s big.LITTLE-based heterogeneousmulti-processor (HMP) show the effectiveness of GDA in efficientresource allocation in comparison with existing fixed objectivepolicies.

Goal-Driven Autonomy, Autonomous and Self-Aware Systems, On-chip Resource allocation, HeterogeneousMulti-core Systems

http://dx.doi.org/10.23919/DATE.2019.8715134