A survey on energy-efficient methodologies and architectures of network-on-chip

Integration of large number of electronic components on a single chip has resulted in complete and complex systems on a single chip. The energy efficiency in the System-on-Chip (SoC) and its communication subset, the Network-on-Chip (NoC), is a key challenge, due to the fact that these systems are typically battery-powered. We present a survey that provides a broad picture of the state-of-the-art energy-efficient NoC architectures and techniques, such as the routing algorithms, buffered and bufferless router architectures, fault tolerance, switching techniques, voltage islands, and voltage-frequency scaling. The objective of the survey is to educate the readers with the latest design-improvements that are carried out in reducing the power consumption in the NoCs.     

Hardware spiking neural network prototyping and application

EMBRACE has been proposed as a scalable, reconfigurable, mixed signal, embedded hardware Spiking Neural Network (SNN) device. EMBRACE, which is yet to be realised, targets the issues of area, power and scalability through the use of a low area, low power analogue neuron/synapse cell, and a digital packet-based Network on Chip (NoC) communication architecture. The paper describes the implementation and testing of EMBRACE-FPGA, an FPGA-based hardware SNN prototype. The operation of the NoC inter-neuron communication approach and its ability to support large scale, reconfigurable, highly interconnected SNNs is illustrated. The paper describes an integrated training and configuration platform and an on-chip fitness function, which supports GA-based evolution of SNN parameters. The practicalities of using the SNN development platform and SNN configuration toolset are described. The paper considers the impact of latency jitter noise introduced by the NoC router and the EMBRACE-FPGA processor-based neuron/synapse model on SNN accuracy and evolution time. Benchmark SNN applications are described and results demonstrate the evolution of high quality and robust solutions in the presence of noise. The reconfigurable EMBRACE architecture enables future investigation of adaptive hardware applications and self repair in evolvable hardware.     

Application of deterministic and stochastic Petri-Nets for performance modeling of NoC architectures

The design of appropriate communication architectures for complex Systems-on-Chip (SoC) is a challenging task. One promising alternative to solve these problems are Networks-on-Chip (NoCs). Recently, the application of deterministic and stochastic Petri-Nets (DSPNs) to model on-chip communication has been proven to be an attractive method to evaluate and explore different communication aspects. In this contribution the modeling of basic NoC communication scenarios featuring different processor cores, network topologies and communication schemes is presented. In order to provide a testbed for the verification of modeling results a state-of-the-art FPGA-platform has been utilized. This platform allows to instantiate a soft-core processor network which can be adapted in terms of communication network topologies and communication schemes. It will be shown that DSPN modeling yields good communication performance prediction results at low modeling effort. Different DSPN modeling aspects in terms of accuracy and computational effort are discussed.     

A multi-application approach for synthesizing custom network-on-chips

Execution of multiple applications on Multi-Processor System-on-Chips (MPSoCs) significantly boosts performance and energy efficiency. Although various researchers have suggested Network-on-Chip (NoC) architectures for MPSoCs, the problem still needs more investigations for the case of multi-application MPSoCs. In this paper, we propose a fully automated synthesis flow in five steps for the design of custom NoC fabrics for multi-application MPSoCs. The steps include: preprocessing, core to router allocation, voltage island merging, floorplanning, and router to router connection. The proposed flow finds design solutions that satisfy the performance, bandwidth, and power constraints of all input applications. If the user decides, the proposed synthesis adds network-level reconfiguration to improve the efficiency of the obtained design solutions. With the reconfiguration option, the proposed flow comes up with adaptive NoC architectures that satisfy each application’s communication requirements while power-gate idle resources, e.g., router ports and links. If reconfiguration option is not set by the user, the proposed flow considers the top communication requirements among the applications in finding design solutions. We have used the proposed synthesis flow to design custom NoCs for several combined graphs of real-world applications and synthetic graphs. Results show that the reconfiguration option can save up to 98% in the energy-delay product (EDP) of the ultimate designs.

     

Power- and performance-efficient cluster-based network-on-chip with reconfigurable topology

Topology is widely known as the most important characteristic of networks-on-chip (NoC), since it highly affects overall network performance, cost, and power consumption. In this paper, we propose a reconfigurable architecture and design flow for NoCs on which a customized topology for any target application can be implemented. In this structure, the nodes are grouped into some clusters interconnected by a reconfigurable communication infrastructure. The nodes inside a cluster are connected by a mesh to benefit from the interesting characteristics of the mesh topology, i.e. regular structure and efficient handling of local traffic. A reconfigurable inter-cluster topology then eliminates the major shortcoming of the mesh by providing short paths between remotely located nodes. We then present a design flow that maps the frequently communicating tasks of a given application onto the same cluster and exploits the reconfigurable infrastructure to set up appropriate inter-cluster connections. The experimental results show that by efficiently handling local and long-distance traffic flows, this structure is scalable, and power- and performance-efficient.     

A new test scheduling algorithm based on Networks-on-Chip as Test Access Mechanisms

Networks-on-Chip (NoCs) can be used for test data transportation during manufacturing tests. On one hand, NoC can avoid dedicated Test Access Mechanisms (TAMs), reducing long global wires, and potentially simplifying the layout. On the other hand, (a) it is not known how much wiring is saved by reusing NoCs as TAMs, (b) the impact of reuse-based approaches on test time is not clear, and (c) a computer aided test tool must be able to support different types of NoC designs. This paper presents a test environment where the designer can quickly evaluate wiring and test time for different test architectures. Moreover, this paper presents a new test scheduling algorithm for NoC TAMs which does not require any NoC timing detail and it can easily model NoCs of different topologies. The experimental results evaluate the proposed algorithm for NoC TAMs with an exiting algorithm for dedicated TAMs. The results demonstrate that, on average, 24% (up to 58%) of the total global wires can be eliminated if dedicated TAMs are not used. Considering the reduced amount of dedicated test resources with NoC TAM, the test time of NoC TAM is only, on average, 3.88% longer compared to dedicated TAMs.     

Efficient routing techniques in heterogeneous 3D Networks-on-Chip

Three-dimensional Networks-on-Chips (3D NoCs) have recently been proposed to address the on-chip communication demands of future highly dense 3D multi-core systems. Homogeneous 3D NoC topologies have many Through Silicon Vias (TSVs) which have a costly and complex manufacturing process. Also, 3D routers use more memory and are more power hungry than conventional 2D routers. Alternatively, heterogeneous 3D NoCs combine both the area and performance benefits of 2D and 3D static router architectures by using a limited number of TSVs. To improve the performance of heterogeneous 3D NoCs, we propose an adaptive router architecture which balances the traffic in such NoCs. Particularly, experimental results show that our proposed architecture significantly improves the performance up to 75% by replacing 2D static routers with adaptive 2D routers in heterogeneous 3D NoCs, while keeping the maximum clock frequency, power and energy consumption of the adaptive router nearly at the same level as the static router.     

State observer controller design for packets flow control in networks-on-chip

Packet-switched networks-on-chips (NoCs) are efficient communication architectures for future multiprocessors system-on-chip (MP-SoC) platforms. However the run-time management of their communication, especially flow control from individual intellectual property (IP) in an NoC which contains large number of IPs, is a challenging task. This paper proposes a state space model for NoC with state observer controller in its feedback path. It is seen that controlling the input and output flow rates alone is not sufficient to stabilize the network, but it is also important to monitor the intermediate flow rates from the on-chip routers. This is possible through a state space model for the NoC. The state observer observes the flow rates from each on-chip router which are then treated as state space variables. These variables can be controlled by the poles placement in the feedback controller. The proposed mathematical model can also observe the required intermediate flow rates which cannot be measured directly (reduced state observer). With these observed states we can attach a state controller. With this controller the network can be stabilized by controlling the flow rates at the intermediate level.     

Bit-accurate energy estimation for Networks-on-Chip

Networks-on-Chip (NoCs) are recognized as the solution to address the communication bottleneck in a Multi-processor System-on-Chip (MPSoC). As NoCs represent a significant part of system consumption, MPSoC designers expect accurate power models in order to produce energy efficient systems. Nowadays, NoC simulators rely on power models that integrate link models without crosstalk modeling. In this study, we present Noxim-XT, a NoC simulator based on Noxim that embeds a link power model with crosstalk modeling. We show that the crosstalk effect has a deep impact on NoC energy consumption since our results demonstrate that classical models generate errors up to 45.5% on the whole NoC energy consumption estimation. In addition, this tool is able to run application-based traffic and we show that under application-based traffics, the energy estimation made by classical models overestimates the NoC energy consumption by up to 50%.     

An analytical model of broadcast in QoS-aware wormhole-routed NoCs

Networks-on-Chip (NoC) emerged to address the technological and design issues related to development of large systems-on-chip (SoCs). Due to diversity of the application's performance requirements, most NoC architectures offer supports for quality of service (QoS). Also, to utilize the available bandwidth efficiently, they might implement mechanisms for delivering collective communication operations. This paper presents an analytical model to predict the average latency of wormhole-routed prioritized broadcast communication in NoCs. The model assumes that the network uses all-port routers scheme and offers differentiated services-based QoS. To verify the analysis, the model predictions are compared against the results obtained from a discrete-event simulator developed using OMNET++.     

Stochastic communication for application-specific Networks-on-Chip

In this paper, we have developed analytical stochastic communication technique for inter and intra-Networks-on-Chip (NoC) communication. It not only separates the computation and communication in Networks-in-Package (NiP) but also predicts the communication performance. Moreover, it will help in tracking of the lost data packets and their exact location during the communication. Further, the proposed technique helps in building the Closed Donor Controlled Based Compartmental Model, which helps in building Stochastic Model of NoC and NiP. This model helps in computing the transition probabilities, latency, and data flow from one IP to other IP in a NoC and among NoCs in NiP. From the simulation results, it is observed that the transient and steady state response of transition probabilities give state of data flow latencies among the different IPs in NoC and among the compartments of NoCs in NiP. Furthermore, the proposed technique produces low latency as compared to the latencies being produced by the existing topologies.     

Performance evaluation and design tradeoffs of on-chip interconnect architectures

Network-on-Chip (NoC) has been proposed as an alternative to bus-based schemes to achieve high-performance and scalability in System-on-Chip (SoC) design. Performance analysis and evaluation of on-chip interconnect architectures are widely based on simulations, which become computationally expensive, especially for large-scale NoCs. In this paper, a Network Calculus-based methodology is presented to analyze and evaluate the performance and cost metrics, such as latency and energy consumption. The 2D Mesh, Spidergong, and WK-Recursive on-chip interconnect architectures are analyzed using this methodology and results are compared with those produced using simulations. The values obtained by simulations and by analysis show similar trends in the same order of magnitude. Furthermore, WK outperforms the other on-chip interconnects in all considered metrics.     

Floorplan-aware application-specific network-on-chip topology synthesis using genetic algorithm technique

Communication plays a critical role in the design and performance of multi-core systems-on-chip (SoCs). Networks-on-chip (NoCs) have been proposed as a promising solution to complex on-chip communication problems. As regular NoC topologies are infeasible to satisfy the performance demand for application-specific NoC, customized topology synthesis is therefore desirable. However, NoC topology synthesis problem is an NP-hard problem. In this paper, we propose a suboptimal genetic-algorithm based technique to synthesize application-specific NoC topology with system-level floorplan awareness. The method minimizes the power consumption and router resources while satisfying latency and bandwidth performance constraints. We have evaluated the proposed technique by running a number of representative benchmark applications and the results indicate that our method generates approximate optimal topologies effectively and efficiently for all benchmarks under consideration.     

Using non-preemptive regions and path modification to improve schedulability of real-time traffic over priority-based NoCs

Network-on-Chip (NoC) is a preferred communication medium for massively parallel platforms. Fixed-priority based scheduling using virtual-channels is one of the promising solutions to support real-time traffic in on-chip networks. Most of the existing works regarding priority-based NoCs use a flit-level preemptive scheduling. Under such a mechanism, preemptions can only happen between the transmissions of successive flits but not during the transmission of a single flit. In this paper, we present a modified framework where the non-preemptive region of each NoC packet increases from a single flit. Using the proposed approach, the response times of certain traffic flows can be reduced, which can thus improve the schedulability of the whole network. As a result, the utilization of NoCs can be improved by admitting more real-time traffic. Schedulability tests regarding the proposed framework are presented along with the proof of the correctness. Additionally, we also propose a path modification approach on top of the non-preemptive region based method to further improve schedulability. A number of experiments have been performed to evaluate the proposed solutions, where we can observe significant improvement on schedulability compared to the original flit-level preemptive NoCs.     

The 2D digraph-based NoCs: attractive alternatives to?the 2D mesh NoCs

This paper proposes two-dimensional directed graphs (or digraphs for short) as a promising alternative to the popular 2D mesh topology for networks-on-chip (NoCs). Mesh is the most popular topology for the NoCs, mainly due to its suitability for on-chip implementation and low cost. However, the fact that a digraph offers a lower diameter than its equivalent linear array of equal cost motivated us to evaluate digraphs as the underlying topology of NoCs. This paper introduces a family of NoC topologies based on three well-known digraphs, namely de Bruijn, shuffle-exchange, and Kautz. We study topological properties of the proposed topologies. We show that the proposed digraph-based topologies have several attractive features including constant node degree, low diameter and cost, and low zero load latency which result in superior performance over the mesh. We introduce a deadlock-free routing algorithm for the proposed NoC topologies and compare NoCs employing the proposed topologies and the mesh topology in terms of power consumption and performance. Simulation results also reveal that the proposed NoC topologies offer higher performance and consume lower power than the mesh NoC.     

Scalable network-on-chip architecture for configurable neural networks

Providing highly flexible connectivity is a major architectural challenge for hardware implementation of reconfigurable neural networks. We perform an analytical evaluation and comparison of different configurable interconnect architectures (mesh NoC, tree, shared bus and point-to-point) emulating variants of two neural network topologies (having full and random configurable connectivity). We derive analytical expressions and asymptotic limits for performance (in terms of bandwidth) and cost (in terms of area and power) of the interconnect architectures considering three communication methods (unicast, multicast and broadcast). It is shown that multicast mesh NoC provides the highest performance/cost ratio and consequently it is the most suitable interconnect architecture for configurable neural network implementation. Routing table size requirements and their impact on scalability were analyzed. Modular hierarchical architecture based on multicast mesh NoC is proposed to allow large scale neural networks emulation. Simulation results successfully validate the analytical models and the asymptotic behavior of the network as a function of its size.     

A framework for rapid evaluation of heterogeneous 3-D NoC architectures

The scalability of communication infrastructure in modern Integrated Circuits (ICs) becomes a challenging issue, which might be a significant bottleneck if not carefully addressed. Towards this direction, the usage of Networks-on-Chip (NoC) is a preferred solution. In this work, we propose a software-supported framework for quantifying the efficiency of heterogeneous 3-D NoC architectures. In contrast to existing approaches for NoC design, the introduced heterogeneous architecture consists of a mixture of 2-D and 3-D routers, which reduces the delay and power consumption with a slight impact on packet hops. More specifically, the experimental results with a number of DSP applications show the effectiveness of the introduced methodology, as we achieve on average 25% higher maximum operation frequency and 39% lower power consumption compared to the uniform 3-D NoCs.     

A study of 3D Network-on-Chip design for data parallel H.264 coding

In this paper, we implement, analyze and compare different Network-on-Chip (NoC) architectures aiming at higher efficiencies for MPEG-4/H.264 coding. Two-dimensional (2D) and three-dimensional (3D) NoCs based on Non-Uniform Cache Access (NUCA) are analyzed. We present results using a full system simulator with realistic workloads. Experiments show the average network latencies in two 3D NoCs are reduced by 28% and 34% respectively, comparing with 2D design. It is also shown that heat dissipation is a trade-off in improving performance of 3D chips. Our analysis and experiment results provide a guideline to design efficient 3D NoCs for data parallel H.264 coding applications.     

MoNoC: A monitored network on chip with path adaptation mechanism

Complex systems on chip containing dozens of processing resources with critical communication requirements usually rely on the use of networks on chip (NoCs) as communication infrastructure. NoCs provide significant advantages over simpler infrastructures such as shared busses or point to point communication, including higher scalability, more efficient energy management, higher bandwidth and lower average latency. Applications running on NoCs with more than 10% of bandwidth usage attest that the most significant portion of message latencies refers to buffered packets waiting to enter the NoC, whereas the latency portion that depends on the packet traversing the NoC is sometimes negligible. This work presents an adaptive routing architecture, named Monitored NoC (MoNoC), which is based on a traffic monitoring mechanism and the exchange of high priority control packets. This method enables to adapt paths by choosing less congested routes. Practical experiments show that the proposed path adaptation is a fast process, enabling to transmit packets with smaller latencies, up to 9 times smaller, by using non-congested NoC regions.     

3D floorplanning of low-power and area-efficient Network-on-Chip architecture

Network-on-Chip (NoC) architectures have been adopted by chip multi-processors (CMPs) as a flexible solution to the increasing delay in the deep sub-micron regime. However, the shrinking feature size limits the performance of NoCs due to power and area constraints. In this paper, we propose three 3D floorplanning methods for a Triplet-based Hierarchical Interconnection Network (THIN) which is a new high performance NoC. The proposed floorplanning methods use both Manhattan and Y-architecture routing architectures so as to improve the performance, reduce the power consumption and area requirement of THIN. A cycle accurate simulator was developed based on Noxim NoC simulator and ORION 2.0 energy model. The proposed floorplanning methods show up to 24.69% energy and 8.84% area reduction at best compared with 3D Mesh. Our analysis concludes that THIN is not only a feasible but also a low-power and area-efficient NoC at physical level.