Transportation-network-inspired network-on-chip

Abstract

A cost-efficient network-on-chip is needed in a scalable many-core systems. Recent multicore processors have leveraged a ring topology and hierarchical ring can increase scalability but presents different challenges, including higher hop count and global ring bottleneck. In this work, we describe a hierarchical ring topology that we refer to as a transportationnetwork-inspired network-on-chip (tNoC) that leverages principles from transportation network systems. In particular, we propose a novel hybrid flow control for hierarchical ring topology to scale the topology efficiently. The flow control is hybrid in that the channels are allocated on flit granularity while the buffers are allocated on packet granularity. The hybrid flow control enables a simplified router microarchitecture (to minimize per-hop latency) as router input buffers are minimized and buffers are pushed to the edges, either at the output ports or at the hub routers that interconnect the local rings to the global ring -while still supporting virtual channels to avoid protocol deadlock. We also describe a packet-quota-system (PQS) and a separate credit network that provide congestion management, support prioritized arbitration in the network, and provide support for multiflit packets. A detailed evaluation of a 64-core CMP shows that the tNoC improves performance by up to 21% compared with a baseline, buffered hierarchical ring topology while reducing NoC energy by 51%.

Multiple-combinational-channel: A network architecture for workload balance and deadlock free

“…In [16], adding a small side buffer was proved to be more efficient than the technology of bufferless. Kim et al [17] presented a hybrid packet-flit flow control and a packet quota system, which enable a lightweight router with support for VCs while simplifying the switch arbitration. However, packet-switching NoCs with VCs are inefficient for small or medium-scale CMPs and various messages with different lengths in CC-NUMA.…”

Design and Evaluation of Hierarchical Rings with Deflection Routing

“…RELATED WORK To our knowledge, HiRD is the first hierarchical ring design that uses simple, deflection-based ring transfers to eliminate the need for buffering within rings while guaranteeing end-to-end packet delivery. Hierarchical Interconnects: Hierarchical ring-based interconnect was proposed in a previous line of work [43,51,21,44,19,27]. We have already extensively compared to past hierarchical ring proposals qualitatively and quantitatively.…”

“…In contrast, all of these previous works use routers with in-ring buffering, wormhole switching and flow control. Concurrent works by Kim et al propose tNoCs, hybrid packet-flit credit-based flow control [27] and Clumsy Flow Control (CFC) [26]. However, these two designs add additional complexity because tNoCs [27] requires an additional credit network to guarantee forward progress while CFC requires coordination between cores and memory controllers.…”

“…Concurrent works by Kim et al propose tNoCs, hybrid packet-flit credit-based flow control [27] and Clumsy Flow Control (CFC) [26]. However, these two designs add additional complexity because tNoCs [27] requires an additional credit network to guarantee forward progress while CFC requires coordination between cores and memory controllers. In contrast, flow control in HiRD is lightweight (with deflection based flow control, the Retransmit-Once mechanism, and simpler local-to-global and global-to-local buffers).…”

“…In contrast, flow control in HiRD is lightweight (with deflection based flow control, the Retransmit-Once mechanism, and simpler local-to-global and global-to-local buffers). Additionally, throttling decisions in HiRD can be made locally in each local ring as opposed to global decisions in CFC [26] and tNoCs [27].…”