PETA-MC Probabilistic Energy and Timing Analysis of Data Flow Applications on Multi-Core Processors

Goal

The definition of high performance time-critical systems implies the design of highly integrated electronic devices that run sophisticated software. This trend leads to a sharp increase in complexity in embedded systems. Extensive analysis of the hardware/software architectures of embedded systems with respect to their timing behavior under realistic working scenarios is necessary to avoid costly design cycles. The analysis of the timing behavior of software running in parallel on complex multi-core platforms is difficult.

In the field of time response analysis of multi-core systems, simulation-based approaches and mathematical formal methods exist. Both approaches struggle with scaling problems. Our work aims to increase the scalability of the analysis methods by using probability-based simulation techniques.

Our goal in this project is to demonstrate potential improvements in the analyzability of architectures with respect to size (more cores) and complexity (heterogeneity and memory hierarchies) by using probability-based methods instead of existing analytical or state-based formal approaches.

Persons

External Leader

Prof. Dr. Sébastien Le Nours

Scientific Director

Publications
A Fast Yet Accurate Message-level Communication Bus Model for Timing Prediction of SDFGs on MPSoC

Hai-Dang Vu and Sebastien Le Nours and Sebastien Pillement and Ralf Stemmer and Kim Grüttner; 26th Asia and South Pacific Design Automation Conference (ASP-DAC) 2021; 001 / 2021

A Measurement-based Performance Evaluation Framework for Neural Networks on MPSoCs

Quentin Dariol, Sebastien Le Nours, Sebastien Pillement, Ralf Stemmer, Kim Grüttner, Domenik Helms; 15ème Colloque National du GDR SOC2; 0Jun / 2021

Partners
Prof. Dr. Sébastien Le Nours, Université de Nantes
www.ietr.fr/spip.php?auteur696

Duration

Start: 31.12.2018
End: 30.12.2020

Source of funding

Deutscher Akademischer Austauschdienst e.V.