TY - GEN
T1 - PlanarONoC
T2 - 55th Annual Design Automation Conference, DAC 2018
AU - Chuang, Yu Kai
AU - Chen, Kuan Jung
AU - Lin, Kun Lin
AU - Fang, Shao Yun
AU - Li, Bing
AU - Schlichtmann, Ulf
N1 - Publisher Copyright:
© 2018 Association for Computing Machinery.
PY - 2018/6/24
Y1 - 2018/6/24
N2 - Optical networks-on-chips (ONoCs) have become a promising solution for the on-chip communication of multi-and many-core systems to provide superior communication bandwidths, efficiency in power consumption, and latency performance compared to electronic NoCs. Serving as the critical part of ONoCs, an optical router composed of waveguides and photonic switching elements (PSEs) routes signals between two hubs or between a hub and a memory controller. Many studies focus on developing efficient architectures of optical routers, while their physical implementation that can seriously deteriorate the quality of the architectures is rarely addressed. The existing automatic place-and-route tools suffer from considerable insertion loss due to many waveguide crossings outside of PSEs, which leads to huge power consumption of laser sources. By observing that the logic schemes of most optical routers are actually planar, we develop a concurrent PSE placement and waveguide routing flow, called PlanarONoC, that guarantees optimal solutions in terms of crossings for planar logic schemes. Experimental results show that the proposed flow reduces the maximum insertion loss by 37% on average, guarantees no waveguide crossing outside of PSEs, and performs much more efficient compared to the state-of-the-art work.
AB - Optical networks-on-chips (ONoCs) have become a promising solution for the on-chip communication of multi-and many-core systems to provide superior communication bandwidths, efficiency in power consumption, and latency performance compared to electronic NoCs. Serving as the critical part of ONoCs, an optical router composed of waveguides and photonic switching elements (PSEs) routes signals between two hubs or between a hub and a memory controller. Many studies focus on developing efficient architectures of optical routers, while their physical implementation that can seriously deteriorate the quality of the architectures is rarely addressed. The existing automatic place-and-route tools suffer from considerable insertion loss due to many waveguide crossings outside of PSEs, which leads to huge power consumption of laser sources. By observing that the logic schemes of most optical routers are actually planar, we develop a concurrent PSE placement and waveguide routing flow, called PlanarONoC, that guarantees optimal solutions in terms of crossings for planar logic schemes. Experimental results show that the proposed flow reduces the maximum insertion loss by 37% on average, guarantees no waveguide crossing outside of PSEs, and performs much more efficient compared to the state-of-the-art work.
KW - Optical networks-on-chip
KW - Optical router
KW - PSE placement and routing
UR - http://www.scopus.com/inward/record.url?scp=85053664500&partnerID=8YFLogxK
U2 - 10.1145/3195970.3196093
DO - 10.1145/3195970.3196093
M3 - Conference contribution
AN - SCOPUS:85053664500
SN - 9781450357005
T3 - Proceedings - Design Automation Conference
BT - Proceedings of the 55th Annual Design Automation Conference, DAC 2018
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 24 June 2018 through 29 June 2018
ER -