微系统芯片(System-on-Chip,SoC)发展到今天,集成密度指数增长和芯片面积的急剧膨胀使得全局连线的延时上升,可靠性下降,成为集成电路的设计瓶颈.片上网络(Network-on-Chip,NoC)是解决整个芯片上数据有效传输的结构之一,以片上网络为基...微系统芯片(System-on-Chip,SoC)发展到今天,集成密度指数增长和芯片面积的急剧膨胀使得全局连线的延时上升,可靠性下降,成为集成电路的设计瓶颈.片上网络(Network-on-Chip,NoC)是解决整个芯片上数据有效传输的结构之一,以片上网络为基础通信架构的微系统芯片称为片上网上系统芯片(System-on-Network-on-Chip,SoNoC).微系统芯片内通信模式兼有随机性和确定性,应该根据特定应用的通信特征设计片上网络.本文在确定SoNoC设计流程的基础上,根据SoNoC的通信特征,选择了合适的离散平面结构,对SoNoC的运算及控制等模块进行布局、对模块间的通信依赖关系进行布线,发展出FRoD(Floor-plan and Routing on Discrete Plane)算法,以自动生成片上网络的拓扑结构.该算法定义了离散平面的一般表示方法,并在四种典型的离散平面上使用不同规模的随机系统完成了系列实验.为了处理系统和网络之间的耦合关系,逐点分裂的布局算法可以逐步学习和适应系统的通信需求,同时优化系统的执行时间和通信能量,在运行随机任务流图的模拟系统上与随机布局结果相比可以节省30%左右的通信能量,20%左右的系统通信时间.串行、并行和串并混合的布线算法使用最短路径把通信关系分布在离散平面的通道上,使不同的通信关系尽量复用网络通道,与全连接网络相比可以节省10%到30%的面积代价.展开更多
片上网络为具有多个处理单元的高速并行片上系统提供一种结构化的片上通信与互连的方法。当前丰富多样的通信实体的选择、建模和仿真,对于精确评估和优化片上网络的整体性能非常重要。本文提出了一种基于SystemC的片上网络仿真和评估构...片上网络为具有多个处理单元的高速并行片上系统提供一种结构化的片上通信与互连的方法。当前丰富多样的通信实体的选择、建模和仿真,对于精确评估和优化片上网络的整体性能非常重要。本文提出了一种基于SystemC的片上网络仿真和评估构架,以结构化、自动化的方式,支持基于当前通信实体的N o C仿真和设计流程。展开更多
First-Input-First-Output (FIFO) buffers are extensively used in contemporary digital processors and System-on-Chips (SoC). There are synchronous FIFOs and asycnrhonous FIFOs. And different sized FIFOs should be implem...First-Input-First-Output (FIFO) buffers are extensively used in contemporary digital processors and System-on-Chips (SoC). There are synchronous FIFOs and asycnrhonous FIFOs. And different sized FIFOs should be implemented in different ways. FIFOs are used not only for the pipeline design within a processor, for the inter-processor communication networks, for example Network-on-Chips (NoCs), but also for the peripherals and the clock domain crossing at the whole SoC level. In this paper, we review the interface, the circuit implementation, and the various usages of FIFOs in various levels of the digital design. We can find that the usage of FIFOs could greatly facilitate the signal storage, signal decoupling, signal transfer, power domain separation and power domain crossing in digital systems. We hope that more attentions are paid to the usages of synchronous and asynchronous FIFOs and more sophististicated usages are discovered by the digital design communities.展开更多
Dataflow architecture has shown its advantages in many high-performance computing cases. In dataflow computing, a large amount of data are frequently transferred among processing elements through the network-on-chip ...Dataflow architecture has shown its advantages in many high-performance computing cases. In dataflow computing, a large amount of data are frequently transferred among processing elements through the network-on-chip （NoC）. Thus the router design has a significant impact on the performance of dataflow architecture. Common routers are designed for control-flow multi-core architecture and we find they are not suitable for dataflow architecture. In this work, we analyze and extract the features of data transfers in NoCs of dataflow architecture： multiple destinations, high injection rate, and performance sensitive to delay. Based on the three features, we propose a novel and efficient NoC router for dataflow architecture. The proposed router supports multi-destination; thus it can transfer data with multiple destinations in a single transfer. Moreover, the router adopts output buffer to maximize throughput and adopts non-flit packets to minimize transfer delay. Experimental results show that the proposed router can improve the performance of dataflow architecture by 3.6x over a state-of-the-art router.展开更多
文摘微系统芯片(System-on-Chip,SoC)发展到今天,集成密度指数增长和芯片面积的急剧膨胀使得全局连线的延时上升,可靠性下降,成为集成电路的设计瓶颈.片上网络(Network-on-Chip,NoC)是解决整个芯片上数据有效传输的结构之一,以片上网络为基础通信架构的微系统芯片称为片上网上系统芯片(System-on-Network-on-Chip,SoNoC).微系统芯片内通信模式兼有随机性和确定性,应该根据特定应用的通信特征设计片上网络.本文在确定SoNoC设计流程的基础上,根据SoNoC的通信特征,选择了合适的离散平面结构,对SoNoC的运算及控制等模块进行布局、对模块间的通信依赖关系进行布线,发展出FRoD(Floor-plan and Routing on Discrete Plane)算法,以自动生成片上网络的拓扑结构.该算法定义了离散平面的一般表示方法,并在四种典型的离散平面上使用不同规模的随机系统完成了系列实验.为了处理系统和网络之间的耦合关系,逐点分裂的布局算法可以逐步学习和适应系统的通信需求,同时优化系统的执行时间和通信能量,在运行随机任务流图的模拟系统上与随机布局结果相比可以节省30%左右的通信能量,20%左右的系统通信时间.串行、并行和串并混合的布线算法使用最短路径把通信关系分布在离散平面的通道上,使不同的通信关系尽量复用网络通道,与全连接网络相比可以节省10%到30%的面积代价.
文摘片上网络为具有多个处理单元的高速并行片上系统提供一种结构化的片上通信与互连的方法。当前丰富多样的通信实体的选择、建模和仿真,对于精确评估和优化片上网络的整体性能非常重要。本文提出了一种基于SystemC的片上网络仿真和评估构架,以结构化、自动化的方式,支持基于当前通信实体的N o C仿真和设计流程。
文摘First-Input-First-Output (FIFO) buffers are extensively used in contemporary digital processors and System-on-Chips (SoC). There are synchronous FIFOs and asycnrhonous FIFOs. And different sized FIFOs should be implemented in different ways. FIFOs are used not only for the pipeline design within a processor, for the inter-processor communication networks, for example Network-on-Chips (NoCs), but also for the peripherals and the clock domain crossing at the whole SoC level. In this paper, we review the interface, the circuit implementation, and the various usages of FIFOs in various levels of the digital design. We can find that the usage of FIFOs could greatly facilitate the signal storage, signal decoupling, signal transfer, power domain separation and power domain crossing in digital systems. We hope that more attentions are paid to the usages of synchronous and asynchronous FIFOs and more sophististicated usages are discovered by the digital design communities.
基金This work was supported by the National High Technology Research and Development 863 Program of China under Grant No. 2015AA01A301, the National Natural Science Foundation of China under Grant No. 61332009, the National HeGaoJi Project of China under Grant No. 2013ZX0102-8001-001-001, and the Beijing Municipal Science and Technology Commission under Grant Nos. Z15010101009 and Z151100003615006.
文摘Dataflow architecture has shown its advantages in many high-performance computing cases. In dataflow computing, a large amount of data are frequently transferred among processing elements through the network-on-chip （NoC）. Thus the router design has a significant impact on the performance of dataflow architecture. Common routers are designed for control-flow multi-core architecture and we find they are not suitable for dataflow architecture. In this work, we analyze and extract the features of data transfers in NoCs of dataflow architecture： multiple destinations, high injection rate, and performance sensitive to delay. Based on the three features, we propose a novel and efficient NoC router for dataflow architecture. The proposed router supports multi-destination; thus it can transfer data with multiple destinations in a single transfer. Moreover, the router adopts output buffer to maximize throughput and adopts non-flit packets to minimize transfer delay. Experimental results show that the proposed router can improve the performance of dataflow architecture by 3.6x over a state-of-the-art router.
