No abstract

No abstract

Although most Network-on-Chip (NoC) designs are based on Packet Switching (PS), the importance of Circuit Switching (CS) should not be underestimated. Many MPSoC executing real-time applications require an underlying communication backbone that can relay messages from one node to another with guaranteed throughput. Compared to PS, CS can provide guaranteed throughput with lower area and power overheads. It is also highly suited for applications where nodes transfer long messages. Spatial Division Multiplexing (SDM) can allow more efficient use of available network resources by dividing them among multiple simultaneous transactions. The network developed by Vali [1] has three design variations based on the number of sub-channels, has a predictable connection setup time, and uses CS to provide guaranteed throughput once a connection is established. In this paper we use this network as a basis to study the effect of flexibility based on SDM, on the performance of a CS networks. A network evaluation platform has been developed to configure and evaluate networks with a maximum of 8 sub-networks, with each subnetwork comprising of 1, 2 or 4 sub-channels. We show that under uniform traffic pattern with requests of uniform random bandwidth (BW) requirement, a less flexible network outperforms a network with higher flexibility due to a phenomenon we call 'stray requests'. We conclude this paper by showing that under high network traffic, performance of our flexible networks can be as much as 113% better than HAGAR [2] and Liu's [3] network.

No abstract

This article introduces DDRNoC, an on-chip interconnection network capable of routing packets at Dual Data Rate. The cycle time of current 2D-mesh Network-on-Chip routers is limited by their control as opposed to the datapath (switch and link traversal), which exhibits significant slack. DDRNoC capitalizes on this observation, allowing two flits per cycle to share the same datapath. Thereby, DDRNoC achieves higher throughput than a Single Data Rate (SDR) network. Alternatively, using lower voltage circuits, the above slack can be exploited to reduce power consumption while matching the SDR network throughput. In addition, DDRNoC exhibits reduced clock distribution power, improving energy efficiency, as it needs a slower clock than a SDR network that routes packets at the same rate. Post place and route results in 28nm technology show that, compared to an iso-voltage (1.1V) SDR network, DDRNoC improves throughput proportionally to the SDR datapath slack. Moreover, a low-voltage (0.95V) DDRNoC implementation converts that slack to power reduction offering the 1.1V SDR throughput at a substantially lower energy cost. CCS Concepts: • Hardware → Network on chip; Buses and high-speed links; Interconnect; System on a chip; Application specific integrated circuits; • Computer systems organization → Interconnection architectures;