LBNoC

Abstract

An FPGA-based Network-on-Chip (NoC) using a low-latency router with a look-ahead bypass (LBNoC) is discussed in this article. The proposed design targets the optimized area with improved network performance. The techniques such as single-cycle router bypass, adaptive routing module, parallel Virtual Channel (VC), and Switch allocation, combined virtual cut through and wormhole switching, have been employed in the design of the LBNoC router. The LBNoC router is parameterizable with the network topology, traffic patterns, routing algorithms, buffer depth, buffer width, number of VCs, and I/O ports being configurable. A table-based routing algorithm has been employed to support the design of custom topologies. The input buffer modules of NoC router have been mapped on the FPGA Block RAM hard blocks to utilize resources efficiently. The LBNoC architecture consumes 4.5% and 27.1% fewer hardware resources than the ProNoC and CONNECT NoC architectures. The average packet latency of the LBNoC NoC architecture is 30% and 15% lower than the CONNECT and ProNoC architectures. The LBNoC architecture is 1.15× and 1.18× faster than the ProNoC and CONNECT NoC frameworks.