METRO: A Software-Hardware Co-Design of Interconnections for Spatial DNN Accelerators

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Zhao, Zhu, Jingchen, Zhou, Zhe, Sun, Guangyu
Natura: Preprint
Pubblicazione: 2021
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929718324887552
author Wang, Zhao
Zhu, Jingchen
Zhou, Zhe
Sun, Guangyu
author_facet Wang, Zhao
Zhu, Jingchen
Zhou, Zhe
Sun, Guangyu
contents Tiled spatial architectures have proved to be an effective solution to build large-scale DNN accelerators. In particular, interconnections between tiles are critical for high performance in these tile-based architectures. In this work, we identify the inefficiency of the widely used traditional on-chip networks and the opportunity of software-hardware co-design. We propose METRO with the basic idea of decoupling the traffic scheduling policies from hardware fabrics and moving them to the software level. METRO contains two modules working in synergy: METRO software scheduling framework to coordinate the traffics and METRO hardware facilities to deliver the data based on software configurations. We evaluate the co-design using different flit sizes for synthetic study, illustrating its effectiveness under various hardware resource constraints, in addition to a wide range of DNN models selected from real-world workloads. The results show that METRO achieves 56.3% communication speedup on average and up to 73.6% overall processing time reduction compared with traditional on-chip network designs.
format Preprint
id arxiv_https___arxiv_org_abs_2108_10570
institution arXiv
publishDate 2021
record_format arxiv
spellingShingle METRO: A Software-Hardware Co-Design of Interconnections for Spatial DNN Accelerators
Wang, Zhao
Zhu, Jingchen
Zhou, Zhe
Sun, Guangyu
Hardware Architecture
Tiled spatial architectures have proved to be an effective solution to build large-scale DNN accelerators. In particular, interconnections between tiles are critical for high performance in these tile-based architectures. In this work, we identify the inefficiency of the widely used traditional on-chip networks and the opportunity of software-hardware co-design. We propose METRO with the basic idea of decoupling the traffic scheduling policies from hardware fabrics and moving them to the software level. METRO contains two modules working in synergy: METRO software scheduling framework to coordinate the traffics and METRO hardware facilities to deliver the data based on software configurations. We evaluate the co-design using different flit sizes for synthetic study, illustrating its effectiveness under various hardware resource constraints, in addition to a wide range of DNN models selected from real-world workloads. The results show that METRO achieves 56.3% communication speedup on average and up to 73.6% overall processing time reduction compared with traditional on-chip network designs.
title METRO: A Software-Hardware Co-Design of Interconnections for Spatial DNN Accelerators
topic Hardware Architecture
url https://arxiv.org/abs/2108.10570