Using Machine Learning to Estimate Latency and Delivered Packets in Hybrids NoCs

Hybrid Networks-on-Chip broaden paths for constrained optimization. The design space exploration (DSE) approach searches optimized network configurations to comply with those constraints. However, when DSE relies solely on a simulation approach, it may take a considerable time to obtain the desired...

Full description

Saved in:
Bibliographic Details
Main Authors: da Silva Oliveira, Samuel, Pereira, Monica Magalhaes, Kreutz, Marcio Eduardo, de Carvalho, Bruno Motta
Format: Conference Proceeding
Language:English
Subjects:
Computational modeling
DSE
Error analysis
Hybrid Network
Machine learning
Network-on-Chip
Optimization
Predictive models
Space exploration
Wireless communication
Wireless Network-on-Chip
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Hybrid Networks-on-Chip broaden paths for constrained optimization. The design space exploration (DSE) approach searches optimized network configurations to comply with those constraints. However, when DSE relies solely on a simulation approach, it may take a considerable time to obtain the desired results, thus shortening the number of possible scenarios to be tested. This paper proposes using Machine Learning (ML) to optimize the DSE step, making the process faster than using simulation-based approaches. Experimental results demonstrate that the proposed model was able to predict the latency values with an error rate of 2% and predict the delivered packets with an error rate of 1%.
ISSN:2324-7894
DOI:10.1109/SBESC53686.2021.9628340