Autonomous multi-processor-SoC optimization with distributed learning classifier systems XCS

Authors:
Andreas Bernauer;Gunnar Arndt;Oliver Bringmann;Wolfgang Rosenstiel
Affiliations:
Faculty of Science, University of Tübingen, Tübingen, Germany;Faculty of Science, University of Tübingen, Tübingen, Germany;Forschungszentrum Informatik, Karlsruhe, Germany;Faculty of Science, University of Tübingen, Tübingen, Germany
Venue:
Proceedings of the 8th ACM international conference on Autonomic computing
Year:
2011

Citing 14
Cited 1

A static power model for architects

Proceedings of the 33rd annual ACM/IEEE international symposium on Microarchitecture
An Algorithmic Description of XCS

IWLCS '00 Revised Papers from the Third International Workshop on Advances in Learning Classifier Systems
Parameter variations and impact on circuits and microarchitecture

Proceedings of the 40th annual Design Automation Conference
Temperature Influence on Power Consumption and Time Delay

DSD '03 Proceedings of the Euromicro Symposium on Digital Systems Design
Temperature-aware microarchitecture: Modeling and implementation

ACM Transactions on Architecture and Code Optimization (TACO)
Towards a Framework and a Design Methodology for Autonomic SoC

ICAC '05 Proceedings of the Second International Conference on Automatic Computing
Designing Reliable Systems from Unreliable Components: The Challenges of Transistor Variability and Degradation

IEEE Micro
Reliability Concerns in Embedded System Designs

Computer
High-performance CMOS variability in the 65-nm regime and beyond

IBM Journal of Research and Development - Advanced silicon technology
Collective behavior based hierarchical XCS

Proceedings of the 9th annual conference companion on Genetic and evolutionary computation
Classifier fitness based on accuracy

Evolutionary Computation
Generic Self-Adaptation to Reduce Design Effort for System-on-Chip

SASO '09 Proceedings of the 2009 Third IEEE International Conference on Self-Adaptive and Self-Organizing Systems
Toward a theory of generalization and learning in XCS

IEEE Transactions on Evolutionary Computation
Statistical clock skew analysis considering intradie-process variations

IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems

The subsumption mechanism for XCS using code fragmented conditions

Proceedings of the 15th annual conference companion on Genetic and evolutionary computation

Quantified Score

Hi-index	0.00

Visualization

Abstract

We present a methodology for autonomous optimization of multi-processor system-on-chips (MPSoC) using distributed learning classifier systems (LCS). The methodology promises to offer a high design reuse rate and to reduce design complexity. The distributed LCS autonomously optimize each core of an MPSoC and, by exchanging classifiers, help each other in achieving their individual goals, compensating reciprocal heating of the cores. We validate our methodology with a realistic simulation of an MPSoC inspired by the Cell processor, where the LCSs set the frequency and voltage of the cores to maximize each core's performance while minimizing temperature-dependent timing errors. In our example application, selecting emigrants by fitness, communicating along a complete graph, and deleting by prediction error results in the highest performance for each core that allows operation with almost no temperature-dependent timing errors but retains self-adaptation capabilities. The results show how distributed LCS autonomously optimize the cores of an MPSoC, greatly reducing design efforts through a reusable self-adaptation component and an according design methodology.