Run-time mapping for reliable many-cores based on energy/performance trade-offs

Research output: Contribution to book/conference proceedings/anthology/reportConference contributionContributedpeer-review

Contributors

  • C. Bolchini - , Polytechnic University of Milan (Author)
  • M. Carminati - , Polytechnic University of Milan (Author)
  • A. Miele - , Polytechnic University of Milan (Author)
  • A. Das - , National University of Singapore (Author)
  • A. Kumar - , National University of Singapore (Author)
  • B. Veeravalli - , National University of Singapore (Author)

Abstract

This paper presents a run-time resource manager for NoC-based many-core architectures that dynamically determines the most effective mapping of tasks on the processing nodes of the architecture to optimize system reliability while leveraging on performance and communication energy. An adaptive engine is exploited to pursue the given optimization goal taking into account various metrics. Experimental results for a set of benchmarks show that the adaptive engine (with MTTF optimization only) achieves 16% improvement in MTTF with respect to static pre-computed mapping at a reasonable communication energy overhead of 10%. Further, with MTTF and energy optimization engine turned-on, the adaptive engine is able to minimize communication energy consumption by 50% while exploiting the available MTTF slack.

Details

Original languageEnglish
Title of host publication2013 IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFTS)
Place of PublicationNew York
PublisherIEEE Xplore
Pages58-64
Number of pages7
ISBN (electronic)978-1-4799-1585-9, 978-1-4799-1583-5
Publication statusPublished - 2013
Peer-reviewedYes
Externally publishedYes

Publication series

SeriesIEEE International Symposium on Defect and Fault Tolerance in VLSI Systems
ISSN1550-5774

Conference

Title2013 26th IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems, DFTS 2013
Duration2 - 4 October 2013
CityNew York City, NY
CountryUnited States of America

Keywords

Research priority areas of TU Dresden

Sustainable Development Goals

ASJC Scopus subject areas

Library keywords