Towards Scalable Symbolic Routing for Multi-Objective Networked Embedded System Design and Optimization

Beitrag bei einer Tagung


Details zur Publikation

Autor(en): Graf S, Reimann F, Glaß M, Teich J
Titel Sammelwerk: 2014 International Conference on Hardware/Software Codesign and System Synthesis, CODES+ISSS 2014
Verlag: IEEE Press
Verlagsort: New York, NY, USA
Jahr der Veröffentlichung: 2014
Tagungsband: Proc. of the International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS)
Seitenbereich: 2:1-2:10


Abstract


Symbolic encoding for resource allocation, task binding, and message routing during multi-objective design space exploration (DSE) has gained significant attention in recent years. To determine the message routing, existing symbolic approaches typically rely on an explicit encoding of routing hops which results in a huge number of required variables and/or constraints. As a result, these approaches fail in case of large network diameters and/or a huge number of messages or resources and even for smaller problems, the convergence of the involved optimization process suffers. To tackle this shortcomings, this work proposes three novel symbolic routing encoding strategies that all avoid to encode hops explicitly, but are based on an encoding of individual links or complete sender-receiver paths, but still cover the same design space. The result is a more compact problem representation with less constraints and, in particular, less variables; the latter eliminates ineffective degrees of freedom from the search space and significantly enhances the optimization quality of a multi-objective optimization with even non-linear objectives. In an extensive test-suite, three major classes of wired networked embedded systems are considered: (a) hierarchical stars as in MPSoCs or automotive, (b) redundant backbone buses as common in rail systems or avionics, and (c) mesh-based architectures that often occur in NoC-based MPSoCs. For all three classes, the proposed approaches significantly outperform existing techniques in both scalability and optimization quality and, thus, considerably enlarge the field of application of a multi-objective DSE for the networked embedded system design. Copyright is held by the owner/author(s).



FAU-Autoren / FAU-Herausgeber

Glaß, Michael Prof. Dr.-Ing.
Lehrstuhl für Informatik 12 (Hardware-Software-Co-Design)
Graf, Sebastian
Lehrstuhl für Informatik 12 (Hardware-Software-Co-Design)
Reimann, Felix
Lehrstuhl für Informatik 12 (Hardware-Software-Co-Design)
Teich, Jürgen Prof. Dr.-Ing.
Lehrstuhl für Informatik 12 (Hardware-Software-Co-Design)


Zitierweisen

APA:
Graf, S., Reimann, F., Glaß, M., & Teich, J. (2014). Towards Scalable Symbolic Routing for Multi-Objective Networked Embedded System Design and Optimization. In Proc. of the International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS) (pp. 2:1-2:10). New Delhi, IN: New York, NY, USA: IEEE Press.

MLA:
Graf, Sebastian, et al. "Towards Scalable Symbolic Routing for Multi-Objective Networked Embedded System Design and Optimization." Proceedings of the International Conference on Hardware/Software Codesign and System Synthesis (CODES+ISSS), New Delhi New York, NY, USA: IEEE Press, 2014. 2:1-2:10.

BibTeX: 

Zuletzt aktualisiert 2018-09-08 um 23:27