A network-on-chip construction method, device, medium and product

Abstract

The application discloses a network-on-chip construction method, device, medium and product, and relates to the technical field of network-on-chip design. The method comprises the following steps: establishing a bottom-layer cellular network based on a cellular topology structure; each node in the bottom-layer network comprises a router and a processing unit, and performs data interaction with adjacent nodes; taking a center node of the bottom-layer network as an origin, the network is symmetrically divided into a plurality of geometric sectors with the same coverage angle, and at least one node is selected from each sector as a cross-layer node; and the cross-layer node is connected to a corresponding node in a top-layer switching network through an additional channel, and the top-layer network node only comprises a router and is fully interconnected through the additional channel. The long-distance communication is transferred to the top-layer switching network through the hierarchical structure, the data transmission hop count and signal delay are effectively reduced, the hardware resource consumption is reduced, and the network expansibility and communication efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of network-on-chip design, and particularly relates to a network-on-chip construction method, device, medium and product. BACKGROUND

[0002] With the improvement of chip integration, the traditional two-dimensional network-on-chip (NoC) is difficult to meet the demand of large-scale integrated circuits due to the problems of physical wiring length and signal delay. In the prior art, although the three-dimensional NoC shortens the network diameter, there are problems of through-silicon via (TSV) communication bottleneck and structural complexity; the hierarchical interconnection structure optimizes long-distance communication by introducing a top-level switching network, but the Mesh network node has a high connection degree, resulting in large link overhead and wiring complexity. The cellular topology network is concerned due to its low node connection degree and less hardware resource consumption, but the traditional cellular network lacks hierarchical design and is difficult to balance local communication efficiency and global transmission performance.

[0003] Therefore, there is an urgent need to improve the network-on-chip construction method to solve the problems of large hardware overhead, high wiring complexity and insufficient scalability in the traditional scheme. SUMMARY

[0004] The present application provides a network-on-chip construction method, device, medium and product to solve the problems of large hardware overhead, high wiring complexity and insufficient scalability in the traditional scheme.

[0005] The present application provides a network-on-chip construction method, which comprises the following steps:

[0006] A bottom cellular network is established based on a cellular topology structure; each node in the bottom cellular network comprises a router and a processing unit, and each node in the bottom cellular network interacts with adjacent nodes;

[0007] At least one node in each region of the bottom cellular network is selected as a cross-layer node; each region is a geometric sector obtained by symmetrically dividing the bottom cellular network with the center node of the bottom cellular network as the origin, and the coverage angle of each sector is the same;

[0008] The cross-layer node is connected to a corresponding node in a top-level switching network through an additional channel to obtain a hierarchical cellular network-on-chip; each node in the top-level switching network comprises a router, and full interconnection is realized through the additional channel.

[0009] The present application also provides an electronic device comprising a memory for storing a computer program and a processor for executing the computer program to realize the steps of any of the network-on-chip construction methods.

[0010] The application further provides a computer readable storage medium, and the computer readable storage medium stores a computer program.

[0011] The application further provides a computer program product, comprising a computer program, and the computer program is executed by a processor to implement the steps of any one of the network-on-chip construction methods.

[0012] Through the application, a bottom layer network is established based on a cellular topology structure, a cross-layer node is selected through symmetrical division of a geometric sector, and an additional channel is connected to a top layer switching network of full interconnection to form a hierarchical communication architecture. The bottom layer cellular network adopts a hexagonal cellular topology structure, and each node only needs to be connected with 3 adjacent nodes (compared with 4-degree connection of a traditional two-dimensional Mesh network), which significantly reduces the number of physical links, greatly reduces the wiring area and power consumption; at the same time, the symmetrical division of the geometric sector ensures that the cross-layer nodes are uniformly distributed in the bottom layer network, avoids the wiring congestion problem caused by the excessively high node density in the local area, and improves the overall link utilization rate. The cross-layer node is directly connected to the top layer switching network of full interconnection through the additional channel, a high-speed transmission channel is constructed, and the cross-sector and long-distance data packet does not need to traverse the bottom layer cellular node hop by hop, but directly reaches the target area through the top layer network. The geometric symmetry of the bottom layer cellular topology is closely combined with the sector division rule, supports flexible expansion of the network, and newly added nodes only need to expand outward according to the hexagonal ring layer and follow the symmetrical sector layout, without the need to redesign the routing logic or adjust the top layer architecture; in addition, the division of labor between the bottom layer and the top layer in the hierarchical architecture (the bottom layer is responsible for adjacent communication, and the top layer is responsible for cross-area transmission) is compatible with the existing network-on-chip protocol, reduces the integration and migration cost of the traditional Mesh or NoC system. The symmetrical sector division ensures the uniform distribution of the cross-layer nodes in the bottom layer network, in combination with the path diversity of the full interconnection top layer network, so that the communication traffic can be dynamically allocated to different sectors and top layer links, avoiding the congestion problem of the edge node caused by the single path in the traditional Mesh network; at the same time, the full interconnection characteristic of the top layer network provides redundant channels for the key path, which can automatically switch the route when the local link fails, and improves the system reliability. The application complements the hardware efficiency advantage of the cellular topology and the global transmission capacity of the hierarchical architecture, and solves the trade-off problem between the scalability, delay and resource consumption of the traditional network-on-chip. BRIEF DESCRIPTION OF DRAWINGS

[0013] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the following will briefly introduce the drawings needed to be used in the specific embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0014] Figure 1 A conversion process diagram of a mesh network to a cellular network provided by an embodiment of the present application;

[0015] Figure 2 A method flowchart of a network-on-chip construction method provided by an embodiment of the present application;

[0016] Figure 3 A two-dimensional HM4 cellular topology network diagram provided by an embodiment of the present application;

[0017] Figure 4 An HM3 cellular network coordinate system diagram provided by an embodiment of the present application;

[0018] Figure 5 An HM3 cellular network area division diagram provided by an embodiment of the present application;

[0019] Figure 6 A cross-layer node diagram in an HM3 cellular network provided by an embodiment of the present application;

[0020] Figure 7 A type one construction diagram in a hierarchical HM2 network topology provided by an embodiment of the present application;

[0021] Figure 8 A type one construction diagram in a hierarchical HM3 network topology provided by an embodiment of the present application;

[0022] Figure 9 A type one construction diagram in a hierarchical HM4 network topology provided by an embodiment of the present application;

[0023] Figure 10 A type two construction diagram in a hierarchical HM2 network topology provided by an embodiment of the present application;

[0024] Figure 11 A type two construction diagram in a hierarchical HM3 network topology provided by an embodiment of the present application;

[0025] Figure 12 A type two construction diagram in a hierarchical HM4 network topology provided by an embodiment of the present application;

[0026] Figure 13 A structure diagram of a network-on-chip construction apparatus provided by an embodiment of the present application;

[0027] Figure 14 A structure diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0028] With reference to the drawings, the technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0029] It should be noted that, in the description of the present application, the terms "comprise", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. The terms "first", "second" and the like in the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence.

[0030] In order to enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the drawings and specific embodiments.

[0031] In combination with the specific application environment architecture or specific hardware architecture on which the execution of the network-on-chip construction method depends, the specific application environment architecture or specific hardware architecture is described here.

[0032] Network-on-chip (NoC) is a brand-new on-chip interconnection architecture, which transmits data packets through the network on the NoC chip. Although NoC optimizes the scalability and parallel operation of SoC, it does not solve the problem of shortening the physical connection and the signal delay caused by it. With the further improvement of chip integration, limited by the two-dimensional layout condition, 2D NoC is difficult to guarantee the adjacency of key components, and difficult to shorten the length of critical path and reduce signal delay. Although three-dimensional network solves the problems existing in two-dimensional network-on-chip, shortens the network diameter, and increases the path diversity, it has good scalability, but the three-dimensional structure has high complexity and TSV communication bottleneck.

[0033] Due to the simplicity of Mesh structure, it has high regularity, symmetry and scalability, and most of the current hierarchical network-on-chip is based on Mesh network, which not only retains the advantages of Mesh network, but also reduces the network delay.

[0034] The cellular mesh network (referred to as a cellular network) has a similar topology to the Mesh network. The Mesh network can be converted into a cellular network through appropriate changes, and the conversion process is shown in Figure 1. Since the connectivity of the cellular mesh node is only 3, which is 1 less than the node degree of the Mesh network, the hardware overhead of the cellular mesh is 25% less than the Mesh network, which is superior to the two-dimensional Mesh network. Since the cellular network has a similar topology to the Mesh network, similar properties, the cellular mesh can be better researched on the basis of the research of the Mesh network, and the network cost can be greatly reduced compared with the Mesh network.

[0035] Therefore, the embodiment of the present application provides a network-on-chip construction method, which is described in detail in combination with the execution process of the network-on-chip construction method.

[0036] Firstly, the terms involved in the present application are introduced.

[0037] NoC: Network on Chip, network on chip; SoC: System on Chip, system on chip; HM: Honeycomb Mesh, cellular mesh network; AHC: additional highway connections, additional highway connection; TSV: Through Silicon Via, through silicon via technology; PE: Processing Element, processing unit.

[0038] The embodiment of the present application provides a network-on-chip construction method, and the specific process of the method is as shown in Figure 2 The method specifically comprises the following steps.

[0039] S201, establishing a bottom layer cellular network based on a cellular topology; each node in the bottom layer cellular network comprises a router and a processing unit, and each node in the bottom layer cellular network interacts with adjacent nodes.

[0040] Specifically, the bottom layer network adopts a cellular topology, that is, a two-dimensional network formed by periodically arranging a plurality of regular hexagonal units. Each hexagonal vertex is a network node, and the nodes are connected by edges to form a honeycomb mesh. Each node comprises a router (responsible for data packet forwarding) and a processing unit (PE) (executing a computing task), and the two are connected through an internal bus. Each node only communicates directly with 3 adjacent nodes, and data interaction is achieved through a physical link.

[0041] S202, selecting at least one node from each region of the bottom layer cellular network as a cross-layer node.

[0042] Specifically, each region is a geometric sector obtained by symmetrically dividing the bottom-layer cellular network with the center node of the bottom-layer cellular network as the origin, and the coverage angle of each sector is the same. For example, when six-sector division is performed, each sector covers 60°. At least one node in each sector (region) is selected as a cross-layer node, and the selection criteria need to be met: the number of cross-layer nodes selected in each sector is equal, and they are distributed along the central symmetry axis. The cross-layer nodes need to cover all the ring layers of the bottom-layer network.

[0043] S203, connecting the cross-layer nodes to the corresponding nodes in the top-layer switching network through additional channels to obtain a hierarchical cellular on-chip network.

[0044] Specifically, each node in the top-layer switching network contains a router, and full interconnection is achieved through additional channels. The additional channels are dedicated physical links independent of the local communication links of the bottom-layer cellular network, and are used to connect the cross-layer nodes and the top-layer switching network. Each cross-layer node is connected to a designated node of the top-layer network through an additional channel in one-to-one or one-to-many. For example, the cross-layer nodes in the same sector can share one top-layer node.

[0045] The on-chip network construction method of the embodiment establishes a bottom network based on a honeycomb topology structure, selects cross-layer nodes by symmetrically dividing geometric sectors, and connects to a top interconnection switch network by using an additional channel to form a hierarchical communication architecture. The bottom honeycomb network adopts a hexagonal honeycomb topology structure, and each node only needs to be connected with 3 adjacent nodes (compared with 4-degree connection of traditional two-dimensional Mesh network), which significantly reduces the number of physical links, greatly reduces the wiring area and power consumption; at the same time, the symmetrically divided geometric sectors ensure that the cross-layer nodes are uniformly distributed in the bottom network, avoiding the wiring congestion problem caused by the high node density in the local area, and improving the overall link utilization. The cross-layer nodes are directly connected to the top interconnection switch network by the additional channel to construct a high-speed transmission channel, so that the cross-sector and long-distance data packets do not need to traverse the bottom honeycomb nodes hop by hop, but directly reach the target area through the top network. The geometric symmetry of the bottom honeycomb topology is closely combined with the sector division rule, supports flexible expansion of the network, and newly added nodes only need to expand outward according to the hexagonal ring layer and follow the symmetric sector layout, without the need to redesign the routing logic or adjust the top architecture; in addition, the division of labor between the bottom layer and the top layer in the hierarchical architecture (the bottom layer is responsible for adjacent communication, and the top layer is responsible for cross-area transmission) is compatible with existing on-chip network protocols, reduces the integration and migration cost of traditional Mesh or NoC systems. The symmetric sector division ensures the uniform distribution of cross-layer nodes in the bottom network, combined with the path diversity of the fully interconnected top network, so that the communication traffic can be dynamically allocated to different sectors and top layer links, avoiding the congestion problem of edge nodes in traditional Mesh network due to single path; at the same time, the fully interconnected characteristics of the top network provide redundant channels for the key path, which can automatically switch the route when the local link fails, improving the system reliability. The application complements the hardware efficiency advantage of the honeycomb topology and the global transmission capacity of the hierarchical architecture, solves the trade-off problem between scalability, delay and resource overhead of traditional on-chip networks.

[0046] In an optional embodiment, each node in the underlying cellular network has a data interaction interface for intra-node and inter-node data interaction. The data interaction interface includes a processing unit interface and a routing interface; the processing unit interface is used for local data interaction between the processing unit and the router in each node of the underlying cellular network; the routing interface is used for data interaction between adjacent nodes in the underlying cellular network. By configuring a data interaction interface for each node, the intra-node (between the processing unit and the router) and inter-node (between adjacent routers) communication paths are clearly distinguished, avoiding resource competition caused by mixed data flow. The physical isolation of the interface (such as independent bus or channel) enables local computing and network transmission to be performed in parallel, improving the utilization rate of the intra-node processing unit (PE) and reducing the waiting delay of cross-node communication. By separating the data interaction interface into a processing unit interface and a routing interface, the processing unit interface directly connects the PE and the router, shortens the local data access path (such as the distance from the register to the router), and reduces the single-hop transmission delay; the routing interface independently occupies physical link resources, ensuring that data forwarding between adjacent nodes is not disturbed by local computing tasks and improving network throughput; interface separation avoids arbitration delay caused by simultaneous access of the PE and the router to the same bus, meeting the parallel communication needs of multi-core chips.

[0047] In an optional embodiment, the cellular topology is a two-dimensional cellular network composed of multiple hexagonal rings; the additional channel is a bidirectional high-speed link. The use of a two-dimensional cellular network (non-three-dimensional stacking) avoids the complexity of TSV technology, is compatible with existing chip manufacturing processes, and reduces implementation costs; the bidirectional high-speed link as an additional channel increases the line width or uses a low-impedance metal layer to improve signal transmission rate, compensates for the physical distance disadvantage of cross-layer links, and ensures that the single-hop delay of the top layer network is lower than the total sum of the bottom layer multi-hop. In addition, the underlying cellular network includes multiple ring layers; the distribution of the ring layers includes: according to the size of the underlying cellular network, expanding from the hexagonal ring at the center of the two-dimensional cellular network outward, and sequentially numbering the hexagonal rings. Through the hexagonal ring layer numbering mechanism (such as ring 1 to ring t), the number of ring layers (t value) can be selected according to the computing demand (such as the number of cores), ensuring that the network size matches the computing load; the ring layer numbering provides a logical basis for cross-layer node selection, making the distribution of cross-layer nodes consistent with the network expansion direction and reducing the fluctuation of the number of long-distance communication hops.

[0048] In an optional embodiment, the size of the underlying cellular network is an integer greater than 1. Limiting the network size t≥2 (such as HM2 to HM6) ensures that the cellular network has practical application value. t=1 (single hexagon) cannot form a hierarchical structure, and t≥2 forces the network to have layered communication capabilities; when t≥2, the symmetric region division and cross-layer node selection rules can be effectively executed, avoiding the path redundancy problem of single-ring networks.

[0049] In an alternative embodiment, the number of cross-layer nodes is determined by the size of the bottom-layer cellular network; and the cross-layer nodes in the same ring layer are equally spaced, and the cross-layer nodes in different ring layers are aligned along the cellular symmetry axis. Through the rules of equal spacing and symmetry axis alignment, the cross-layer nodes in the same ring layer are uniformly distributed, avoiding bandwidth contention caused by dense cross-layer nodes in local areas; and the nodes in different ring layers are aligned along the symmetry axis, making the additional channel (AHC) present as a straight line or a periodic polyline in the chip layout, reducing the wire conflict and improving the utilization of routing resources.

[0050] In an alternative embodiment, the selection of cross-layer nodes includes:

[0051] The bottom-layer cellular network is divided into six regions by setting three coordinate axes at an angle of 120 degrees with the center point of the bottom-layer cellular network as the origin; a series of parallel lines perpendicular to the coordinate axes are made along the positive direction of each coordinate axis, each parallel line passing through the nodes of the bottom-layer cellular network to form a positive direction zigzag chain; starting from the origin, the coordinate values of all nodes on the first zigzag chain perpendicular to the positive direction of the coordinate axis are 1, those on the second zigzag chain are 2, and so on; a series of parallel lines perpendicular to the coordinate axes are made along the negative direction of each coordinate axis, each parallel line passing through the nodes of the bottom-layer cellular network to form a negative direction zigzag chain; starting from the origin, the coordinate values of the nodes on the first negative direction zigzag chain are 0, those on the second negative direction zigzag chain are -1, and so on; the coordinate values of each node are determined according to the zigzag chain sequence number of the coordinate axis direction in which the node is located; for any region in the bottom-layer cellular network, the nodes with coordinate values equal to the ring layer number are selected as cross-layer nodes.

[0052] Through the coordinate axis division and zigzag chain coordinate rules, accurate cross-layer node positioning is achieved. The six regions divided by the three 120-degree coordinate axes cover the entire network, ensuring that there is no blind area in the selection of cross-layer nodes; the zigzag chain coordinate values (such as the ring layer number) provide a unique selection standard for cross-layer nodes, avoiding layout deviation caused by manual intervention or random selection; the coordinate values of cross-layer nodes are consistent with the ring layer number, ensuring that the physical path of the cross-layer link (AHC) is consistent with the network expansion direction, and shortening the average length of the additional channel.

[0053] In an alternative embodiment, the nodes of the top-layer switching network do not have data interaction interfaces; and the number of nodes of the top-layer switching network is equal to the total number of ring layers of the bottom-layer cellular network. The top-layer nodes do not connect data interaction interfaces (such as PEs or memories) and only forward data, avoiding the competition for router resources between computing tasks and communication tasks and improving the throughput of the top-layer network; and the number of top-layer nodes is equal to the number of ring layers of the bottom-layer (such as four top-layer nodes corresponding to HM4), ensuring that each ring layer has dedicated top-layer routing resources and avoiding congestion caused by shared nodes in multiple ring layers.

[0054] In an alternative embodiment, the cross-layer nodes are connected to corresponding nodes in the top-level switching network through additional channels, including:

[0055] The cross-layer nodes in the same hexagonal ring are connected to the same node in the top-level switching network through additional channels; or the cross-layer nodes in adjacent regions are connected to the same node in the top-level switching network.

[0056] The cross-layer nodes in the same hexagonal ring are connected to the same node in the top-level switching network through additional channels, including:

[0057] Based on the total number of ring layers of the bottom-level cellular network, the number of nodes of the top-level switching network is determined; all cross-layer nodes in the same hexagonal ring layer of the bottom-level cellular network are connected to a single corresponding node in the top-level switching network through additional channels. The six regions of the bottom-level cellular network are divided into three groups of cross-regional node clusters; the number of nodes of the top-level switching network is equal to the number of clusters; the cross-layer nodes of each cluster are connected to the same node in the top-level switching network through additional channels.

[0058] Two alternative cross-layer connection strategies are provided to adapt to different design requirements. Type one is same-ring connection, in which the cross-layer nodes in the same ring layer share a top-level node, reducing the size of the top-level network (node number = ring layer number) and the wiring complexity; type two is cross-regional connection, in which the cross-layer nodes in adjacent regions share a top-level node (such as three groups of regional pairs), further reducing the number of top-level nodes to 3, suitable for chip design with limited area. Through one-to-one mapping of ring layers and top-level nodes, the cross-layer communication of each ring layer is handled by an independent top-level node, avoiding cross-ring layer data flow interference; the failure of a single ring layer top-level node only affects the communication of the ring layer, improving the network fault tolerance. Through regional clustering connection, only 3 top-level nodes are needed to cover the whole network, greatly reducing the number of top-level routers and area overhead; the three groups of clusters are symmetrically distributed along the coordinate axes, making the additional channel (AHC) wiring path exhibit 120-degree rotational symmetry, simplifying the physical design rules.

[0059] Further, in the cross-layer node selection process in the above steps, the selection of the cross-layer node further comprises: monitoring the link utilization of the bottom layer cellular network in real time, and when the link utilization of any sector exceeds a preset threshold, switching the connection mode of the cross-layer node and the top layer switching network to the same ring layer aggregation connection mode; when the link utilization of all sectors is lower than the threshold, switching to the cross-area cluster connection mode. The bandwidth of the additional channel is dynamically allocated according to the ring layer number where the cross-layer node is located, and the bandwidth of the additional channel of the cross-layer node with a ring layer number n is n times the base bandwidth, where n≥1; when selecting the routing path of the data packet, the data packet with the largest cross-ring layer number is preferentially transmitted through the additional channel with the highest bandwidth. When a cross-layer link fault is detected, the following steps are performed: according to the coordinate difference value of the fault link, mapping the target path to the mirror sector symmetric to the current sector to generate a redundant communication path; updating the routing priority table of the corresponding node in the top layer switching network, and raising the priority of the redundant path to the highest level.

[0060] In summary, the on-chip network construction method provided by the embodiments of the present application constructs a set of efficient, low-consumption and scalable on-chip interconnection system through the cooperative design of the cellular topology structure, the symmetric sector division, the cross-layer node dynamic selection rule and the full interconnection top layer switching network, and the depth correlation and complementary effect among the technical features significantly optimize the network performance and resource utilization. The bottom layer network based on the cellular topology adopts a hexagonal ring layer expansion structure, and each node only needs to be connected with 3 adjacent nodes, compared with the 4-degree connection of the traditional Mesh network, the number of hardware links is significantly reduced, and the geometric symmetry of the hexagon provides a physical basis for subsequent cross-layer node selection and sector division. The nodes in the bottom layer network realize isolated communication of the local processing unit (PE) and the router through the modular data interaction interface, avoid data flow conflicts, and the standardized design of the interface further seamlessly compatible with the additional channel of the top layer switching network, ensuring the protocol consistency during cross-layer data transmission, and reducing the logical overhead of heterogeneous network integration. The bottom layer network is symmetrically divided into six 60-degree sectors through a three-axis coordinate system, and the cross-layer nodes are accurately positioned based on the zigzag chain coordinate value, so that the cross-layer nodes are uniformly distributed along the cellular symmetry axis, combined with the ring layer numbering rule to expand outward layer by layer, forming a logical clear node selection framework. This design not only ensures the regularity of the cross-layer nodes in the physical layout (for example, the same ring layer node spacing is equal, and different ring layers are aligned along the symmetry axis), but also deeply couples with the full interconnection architecture of the top layer switching network: when the cross-layer nodes are connected to the top layer through the additional channel, two connection modes can be selected according to the network size, mode one aggregates the cross-layer nodes of the same hexagonal ring layer into a single top layer node, and the network diameter is compressed to the minimum through the direct path of the top layer; mode two connects the cross-layer node clusters of adjacent sectors to three top layer nodes, simplifies the wiring complexity by reducing the number of top layer nodes, while retaining the natural support of cellular symmetry for load balancing. The two modes can dynamically adapt to different scene requirements, for example, mode one is preferred in high-performance computing scenarios to minimize delay, and mode two is selected for area-sensitive chips to reduce wiring density. The full interconnection characteristics of the top layer switching network and the low connectivity of the bottom layer cellular network form a complement: the top layer node only contains a router and is not connected to a processing unit, and focuses on high-speed relay of cross-area data, while the bottom layer node efficiently handles adjacent communication through the local interface, forming a division system of "local refinement of the bottom layer and global high-speed of the top layer". This hierarchical architecture transfers long-distance communication traffic to the top layer through the additional channel, significantly reducing the bottom layer hop count, and the symmetric expansion of the cellular topology supports flexible expansion from HM2 to HM6, and the newly added nodes only need to extend outward according to the hexagonal ring layer rule, without reconstructing the routing logic or adjusting the top layer mapping relationship, greatly improving the scalability of the network.In addition, the symmetrical distribution of cross-layer nodes in combination with the path diversity of the fully interconnected top layer enables dynamic allocation of traffic to different ring layers and sectors, avoiding the congestion problem of edge nodes in traditional mesh networks; and the redundant paths of the top layer network further enhance the fault tolerance, enabling fast switching of routes through mirror-symmetrical sectors in the event of local link failure, thereby improving system reliability. Through the close linkage of geometric rules, hierarchical division of labor, and dynamic adaptation mechanisms, the various technical features achieve a breakthrough in multiple dimensions. At the hardware level, the cellular low-connectivity and modular interface design reduce the wiring area, and the bidirectional high-speed characteristics of the additional channels improve the throughput of the top layer; at the performance level, the hierarchical architecture significantly compresses the maximum number of hops in an end-to-end connection, reducing the average delay and achieving load balancing through the symmetrical sector and ring layer rules; at the scalability level, the hexagonal ring layer expansion and coordinate system support seamless scale expansion, and the two top layer connection modes provide flexible selection for different application scenarios. This method fundamentally solves the contradiction between scalability, delay, and resource overhead in traditional on-chip networks.

[0061] Based on the on-chip network construction method provided in the above embodiments, a specific example will be used to explain in detail.

[0062] The present embodiment also provides a hierarchical cellular on-chip network topology based on the on-chip network construction method of the above embodiment. The bottom layer of the proposed hierarchical cellular interconnection on-chip network is a two-dimensional cellular topology network, and the top layer is a switch network using additional high-speed channel interconnection (AHC). This structure enables long-distance, multi-hop data packets in the on-chip network to be transmitted directly through the top layer fast channel, effectively reducing the number of hops and network delay. The bottom layer network and the top layer network of the hierarchical cellular network proposed in this example have two different types of connection methods, type one has a smaller network diameter, and type two has a more easily implemented hardware layout. The hierarchical cellular interconnection topology network proposed in this example is superior to hierarchical mesh networks (CMesh and CHMesh) in terms of small network diameter, small network delay, good network scalability, low structural complexity, small link overhead, and strong load capacity.

[0063] The bottom layer network of the hierarchical cellular on-chip network topology of the present embodiment is constructed based on a cellular topology structure. The cellular topology structure is composed of many hexagons, and the definition of the cellular network is as follows: for a cellular grid topology, the network is composed of a hexagon with a size of 1, denoted as HM1. Adding a hexagonal ring (i.e., 6 hexagons) to the boundary (6 edges) of HM1 results in a cellular network with a size of 2, denoted as HM2. Similarly, adding a hexagonal ring to the boundary of HM(t-1) results in a cellular network with a size of t, denoted as HMt. A typical cellular topology network can be referred to Figure 3 , Figure 3 For a two-dimensional HM4 cellular topology network, Figure 3The HM3 honeycomb network includes common routing nodes and bidirectional links in the underlying honeycomb network.

[0064] Based on the on-chip network construction method of the above embodiment, the hierarchical honeycomb on-chip network topology construction of the present embodiment needs to divide the region using the coordinate system of the honeycomb network. The present embodiment uses the coordinate system as shown in Figure 4 Figure 4 which is the HM3 honeycomb network coordinate system, Figure 4 The HM3 honeycomb network includes common routing nodes and bidirectional links in the underlying honeycomb network, and the coordinate axes intersect at the center point of the honeycomb network, and the positive half axes of the three coordinate axes evenly divide the honeycomb network into three parts. On a certain coordinate axis (x, y, or z), or , the nodes with the same coordinates form a zigzag chain perpendicular to the axis. If the node coordinates need to be determined, first determine the node coordinates, find a straight line passing through the center point of the honeycomb network and perpendicular to the axis, denoted as (as shown by the dashed line in Figure 4 ), and along the positive direction of the axis, the coordinates of all nodes on the first zigzag chain perpendicular to the axis and parallel to the axis are 1, the coordinates of all nodes on the second zigzag chain perpendicular to the axis and parallel to the axis are 2, and so on to obtain the coordinate values of the positive half axis of the axis; along the negative direction of the axis, the coordinates of all nodes on the first zigzag chain perpendicular to the axis and parallel to the axis are 0, the coordinates of all nodes on the second zigzag chain perpendicular to the axis and parallel to the axis are -1, and so on to obtain the coordinate values of the negative half axis of the axis. Similarly, the and coordinate values of the nodes are obtained.

[0065] Based on the honeycomb topology of the above Figure 3 and the honeycomb network coordinate system of Figure 4 , the specific construction of the hierarchical honeycomb on-chip network topology of the present embodiment is as follows.

[0066] ​​​​This example is based on the size of the cellular network is 2~6, namely HM2, HM3, HM4, HM5, HM6. The hierarchical cellular interconnection network proposed by the application, the bottom network is a two-dimensional cellular network, mainly responsible for the communication between adjacent points, each node is composed of a router and a PE unit, and a port is connected on each node. The top network uses a full interconnection method to connect, which is an additional highway connection (AHC) interconnection switch network, the nodes in the network only contain routers, mainly responsible for communication, and each node is not connected to a port. The nodes in the bottom network connected to the top network through additional highway connections are defined as cross-layer nodes, and the additional highway connections between the two-layer networks are cross-layer links.

[0067] The specific design process of the hierarchical cellular network is as follows.

[0068] Firstly, the size of the bottom cellular network needs to be determined.

[0069] The size of the bottom network of the cellular hierarchical network is determined, which requires that the size of the cellular network HMt is greater than 1, that is HM3 is taken as an example for illustration, as shown in Figure 5 According to the straight line , , The entire cellular network is divided into six regions, which are sequentially recorded as region I, region II, region III, region IV, region V, and region VI in a clockwise direction. From the center of the cellular network, the hexagonal rings are sequentially recorded as ring 1, ring 2, ring 3, and so on.

[0070] Next, the cross-layer nodes in each region need to be determined.

[0071] The cross-layer nodes in the HM3 network are shown in Figure 6 The nodes in the bottom network connected to the top network through additional highway connections are defined as cross-layer nodes, and the cross-layer nodes in each region are as follows:

[0072] In region I, the nodes on ring are cross-layer nodes;

[0073] In region II, the nodes on ring are cross-layer nodes;

[0074] In region III, the nodes on ring are cross-layer nodes;

[0075] In region IV, the nodes on ring are cross-layer nodes;

[0076] In region V, the nodes on the ring are cross-layer nodes; are cross-layer nodes;

[0077] In region VI, the nodes on the ring are cross-layer nodes; are cross-layer nodes.

[0078] Figure 6 The cross-layer nodes in the cellular network are marked in Fig. 5, which includes the common routing nodes in the underlying cellular network, the bidirectional links, the cross-layer nodes in ring 1 and the cross-layer nodes in ring 2.

[0079] Finally, the top-layer network needs to be constructed.

[0080] The construction of the top-layer network includes two types.

[0081] Type I is to connect the cross-layer nodes in the same hexagonal ring in the underlying cellular network to the same node in the top-layer network respectively, and the number of routing nodes in the top-layer network is determined by the size of the underlying HMt network, i.e. for the underlying HMt network, the number of nodes in its top-layer network is , and the connection mode of the nodes is full interconnection.

[0082] The construction of type I in the hierarchical HM2 network topology is shown in Fig. 6, which includes the common routing nodes, the cross-layer nodes, the nodes in the top-layer network, the bidirectional links and the additional high-speed channels, wherein the additional high-speed channels are also bidirectional transmission. Figure 7 Figure 7 The construction of type I in the hierarchical HM3 network topology is shown in Fig. 7, which includes the common routing nodes, the cross-layer nodes in ring 1, the nodes in the top-layer network connected with the cross-layer nodes in ring 1, the cross-layer nodes in ring 2, the nodes in the top-layer network connected with the cross-layer nodes in ring 2, the bidirectional links and the additional high-speed channels.

[0083] The construction of type I in the hierarchical HM4 network topology is shown in Fig. 8, which includes the common routing nodes, the cross-layer nodes in ring 1, the nodes in the top-layer network connected with the cross-layer nodes in ring 1, the cross-layer nodes in ring 2, the nodes in the top-layer network connected with the cross-layer nodes in ring 2, the cross-layer nodes in ring 3, the nodes in the top-layer network connected with the cross-layer nodes in ring 3, the bidirectional links and the additional high-speed channels. Figure 8 Figure 8

[0084] Figure 9 Figure 9

[0085] ​​​​​​Type two is to connect the cross-layer nodes in the bottom network area I and area II with the same node in the top layer network through an additional high-speed channel; to connect the cross-layer nodes in the bottom network area III and area IV with the same node in the top layer network through an additional high-speed channel; and to connect the cross-layer nodes in the bottom network area V and area VI with the same node in the top layer network through an additional high-speed channel. Therefore, the top layer network has three nodes, and the three nodes are all interconnected.

[0086] Figure 10 is the construction of type two in the hierarchical HM2 network topology, Figure 11 is the construction of type two in the hierarchical HM3 network topology, Figure 12 is the construction of type two in the hierarchical HM4 network topology. Figure 10 、 Figure 11 and Figure 12 all contain the ordinary routing nodes, the cross-layer nodes in area I and area II, the nodes in the top layer network connected with the cross-layer nodes in area I and area II, the cross-layer nodes in area III and area IV, the nodes in the top layer network connected with the cross-layer nodes in area III and area IV, the cross-layer nodes in area V and area VI, the nodes in the top layer network connected with the cross-layer nodes in area V and area VI, bidirectional links and additional high-speed channels.

[0087] Through the above three steps, the hierarchical cellular on-chip network topology is constructed. Compared with other networks, the comparison results are shown in Table 1. From the comparison of the Mesh network and the hierarchical Mesh network and the comparison of the HM network and the hierarchical HM network in Table 1, it can be seen that the hierarchical structure can effectively reduce the network diameter, thereby reducing the number of hops and network delay. The network diameter refers to the maximum value of the shortest distance between all node pairs in the network, and the network diameter greatly affects the network delay.

[0088] As can be seen from Table 1, under the condition that the number of network nodes is the same, the connectivity of the top layer nodes of the CMesh network and the hierarchical cellular network is not much different, but the number of top layer nodes and the diameter of the hierarchical cellular network proposed by the present application are smaller than those of the CMesh network, which shows that the resource consumption and delay of the hierarchical cellular network are smaller.

[0089] As can be seen from Table 1, under the condition that the number of network nodes is the same, the number of top layer nodes and the connectivity of the top layer nodes of the CHMesh network and the hierarchical cellular network are not much different, but the diameter of the hierarchical cellular network proposed by the present application is smaller than that of the CHMesh network, which shows that the hierarchical cellular network has smaller delay.

[0090] The complexity of the interconnection network topology is related to the connectivity and the number of links, and the Mesh network is widely used in the network-on-chip due to its simple structure, and it can be seen from Table 1 that when the number of nodes is not much different, the number of links of the hierarchical cellular interconnection network is between that of the CMesh network and that of the CHMesh network, and the connectivity of the network nodes is not much different, so the hierarchical cellular interconnection network proposed in the application is easy to be implemented in hardware and has a broad application prospect.

[0091] In summary, the performance of the hierarchical cellular network-on-chip topology proposed in the present example is better than that of the four network topologies in Table 1, and has the advantages of small network delay, good network scalability, simple structure, easy hardware implementation, etc. The bottom layer of the hierarchical cellular interconnection network-on-chip proposed in the application is a two-dimensional cellular topology network, and the top layer is a switch network using additional high-speed channel interconnection (AHC). This structure can make the long-distance and multi-hop data packets in the network-on-chip directly transmitted through the top layer fast channel, which can effectively reduce the number of hops and network delay. The hierarchical cellular network proposed in the application has two different types of connection between the bottom layer network and the top layer network, type one has a smaller network diameter, and type two has a more easily implemented hardware layout. In summary, the hierarchical cellular interconnection topology network proposed in the application is better than the hierarchical Mesh network (CMesh and CHMesh), has the advantages of small network diameter, small network delay, good network scalability, low structure complexity, small link overhead, strong load capacity, etc., and has a broader application prospect.

[0092] Table 1

[0093]

[0094] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment.

[0095] The embodiment of the application also provides a network-on-chip construction device, and the structure of the device is as shown in Figure 13 The device comprises:

[0096] The bottom layer network construction module 1301 is configured to establish a bottom layer cellular network based on a cellular topology structure; each node in the bottom layer cellular network comprises a router and a processing unit, and each node in the bottom layer cellular network interacts with adjacent nodes for data interaction;

[0097] The cross-layer node selection module 1302 is configured to select at least one node from each region of the bottom layer cellular network as a cross-layer node; each region is a geometric sector obtained by symmetrically dividing the bottom layer cellular network with the center node of the bottom layer cellular network as the origin, and the coverage angle of each sector is the same.

[0098] connecting the cross-layer nodes to corresponding nodes in the top-layer switching network through additional channels to obtain a hierarchical cellular on-chip network; each node in the top-layer switching network comprises a router and is fully interconnected through the additional channels.

[0099] In an optional implementation, the cross-layer node selection module 1302 is specifically configured to:

[0100] taking the center point of the bottom-layer cellular network as an origin, setting three coordinate axes that are mutually at an angle of 120 degrees, and dividing the bottom-layer cellular network into six regions;

[0101] drawing a series of parallel lines perpendicular to the coordinate axes in the positive direction of each coordinate axis, each parallel line passing through the nodes of the bottom-layer cellular network to form a positive-direction zigzag chain, and starting from the origin, the coordinate values of all the nodes on the first zigzag chain perpendicular to the positive direction of the coordinate axis are 1, the coordinate values of all the nodes on the second zigzag chain are 2, and the coordinate values are sequentially increased;

[0102] drawing a series of parallel lines perpendicular to the coordinate axes in the negative direction of each coordinate axis, each parallel line passing through the nodes of the bottom-layer cellular network to form a negative-direction zigzag chain, and starting from the origin, the coordinate values of all the nodes on the first negative-direction zigzag chain are 0, the coordinate values of all the nodes on the second negative-direction zigzag chain are -1, and the coordinate values are sequentially decreased;

[0103] determining the coordinate values of the nodes according to the serial numbers of the zigzag chains in the direction of the coordinate axes where the nodes are located in the bottom-layer cellular network;

[0104] for any region in the bottom-layer cellular network, taking the nodes with a coordinate value equal to the ring layer number as the cross-layer nodes.

[0105] In an optional implementation, the connecting module 1303 is specifically configured to:

[0106] connecting the cross-layer nodes located in the same hexagonal ring to the same node in the top-layer switching network through the additional channels; or connecting the cross-layer nodes in adjacent regions to the same node in the top-layer switching network.

[0107] In an optional implementation, the connecting module 1303 is further configured to:

[0108] confirming the number of nodes in the top-layer switching network based on the total number of ring layers of the bottom-layer cellular network;

[0109] connecting all the cross-layer nodes located in the same hexagonal ring layer in the bottom-layer cellular network to a single corresponding node in the top-layer switching network through the additional channels.

[0110] In an optional implementation, the connecting module 1303 is further configured to:

[0111] The six areas of the bottom-layer cellular network are divided into three groups of cross-area node clusters; the number of nodes of the top-layer switching network is equal to the number of clusters;

[0112] The cross-layer nodes of each cluster are connected to the same node in the top-layer switching network through additional channels.

[0113] The description of the features in the embodiments of the on-chip network construction device can refer to the related description of the embodiments of the on-chip network construction method, which will not be repeated here.

[0114] The embodiments of the present application also provide an electronic device, as shown in the accompanying drawings, comprising a memory 14 and a processor 20, the memory 10 stores a computer program, and the processor 20 is configured to run the computer program to perform the steps in any of the above-mentioned embodiments of the on-chip network construction method. Figure 14

[0115] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, wherein the computer program is configured to perform the steps in any of the above-mentioned embodiments of the on-chip network construction method when running.

[0116] In an example embodiment, the above-mentioned computer readable storage medium can include but is not limited to: a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store computer programs.

[0117] The embodiments of the present application also provide a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the steps in any of the above-mentioned embodiments of the on-chip network construction method.

[0118] The embodiments of the present application also provide another computer program product, which comprises a non-volatile computer readable storage medium, and the non-volatile computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps in any of the above-mentioned embodiments of the on-chip network construction method.

[0119] ​Those skilled in the art will further realize that the mere concepts, teachings, and embodiments described herein are merely meant to provide an enabling description of the claimed application. Accordingly, modifications and / or additions, other than those explicitly described herein, can be obvious to those skilled in the art in the light of this disclosure. The claimed application is intended to embrace all such modifications and / or additions.

[0120] The above provides a method, system, device, equipment and storage medium for constructing an on-chip network. The principles and implementation modes of the present application are described by applying specific examples. The above description of the embodiments is only used to help understand the method and its core idea of the present application. It should be pointed out that, for those skilled in the art, without departing from the principles of the present application, some improvements and modifications can be made to the present application, and these improvements and modifications also fall within the protection scope of the claims of the present application.

Claims

1. A method of constructing a network-on-chip, characterized by, The method comprises: establishing a bottom layer cellular network based on a cellular topology; each node in the bottom layer cellular network comprises a router and a processing unit, and each node in the bottom layer cellular network interacts with adjacent nodes for data exchange; selecting at least one node in each region of the bottom layer cellular network as a cross-layer node; the regions are geometric sectors obtained by symmetrically dividing the bottom layer cellular network with a center node of the bottom layer cellular network as the origin, and the coverage angle of each sector is the same; connecting the cross-layer nodes to corresponding nodes in a top layer switching network through additional channels to obtain a hierarchical cellular on-chip network; each node in the top layer switching network comprises a router, and full interconnection is realized through the additional channels; wherein the selection of the cross-layer nodes comprises: dividing the bottom layer cellular network into six regions with the center node of the bottom layer cellular network as the origin and setting three coordinate axes at an angle of 120 degrees to each other; along the positive direction of each coordinate axis, a series of parallel lines perpendicular to the coordinate axis are drawn, and each parallel line passes through the nodes of the bottom layer cellular network to form a positive direction zigzag chain; starting from the origin, the coordinate values of all nodes on the first zigzag chain perpendicular to the positive direction of the coordinate axis are 1, those on the second zigzag chain are 2, and the values increase sequentially; along the negative direction of each coordinate axis, a series of parallel lines perpendicular to the coordinate axis are drawn, and each parallel line passes through the nodes of the bottom layer cellular network to form a negative direction zigzag chain; starting from the origin, the coordinate values of the nodes on the first negative direction zigzag chain are 0, those on the second negative direction zigzag chain are -1, and the values decrease sequentially; determining the coordinate values of each node according to the zigzag chain sequence number of the coordinate axis direction in which each node is located in the bottom layer cellular network; for any region in the bottom layer cellular network, the node with a node coordinate value equal to the ring layer number is selected as the cross-layer node.

2. The method of claim 1, wherein, Each node in the bottom layer cellular network has a data exchange interface for intra-node and inter-node data exchange.

3. The method of claim 2, wherein, The data exchange interface comprises a processing unit interface and a routing interface; the processing unit interface is used for local data exchange between the processing unit and the router of each node in the bottom layer cellular network; the routing interface is used for data exchange between adjacent nodes in the bottom layer cellular network.

4. The method of claim 3, wherein, The cellular topology is a two-dimensional cellular network composed of a plurality of hexagonal rings; the additional channels are bidirectional high-speed links.

5. The method of claim 4, wherein, The bottom layer cellular network comprises a plurality of ring layers; the distribution of the ring layers comprises: according to the size of the bottom layer cellular network, the hexagonal rings are expanded outward from the center of the two-dimensional cellular network, and the hexagonal rings are sequentially numbered.

6. The method of claim 5, wherein, The size of the bottom layer cellular network is an integer greater than 1.

7. The method of claim 6, wherein, The number of cross-layer nodes is determined by the size of the bottom layer cellular network; and the cross-layer nodes in the same ring layer are equally spaced, and the cross-layer nodes in different ring layers are aligned along the cellular symmetry axis.

8. The method of claim 1, wherein, The nodes of the top layer switching network do not have a data exchange interface; the number of nodes of the top layer switching network is equal to the total number of ring layers of the bottom layer cellular network.

9. The method of claim 8, wherein, The connecting of the cross-layer nodes to the corresponding nodes in the top layer switching network through the additional channels comprises: The cross-layer nodes in the same hexagon ring are connected to the same node in the top-level switching network through additional channels; or the cross-layer nodes in adjacent areas are connected to the same node in the top-level switching network.

10. The method of claim 9, wherein, The connecting of the cross-layer nodes in the same hexagon ring to the same node in the top-level switching network through additional channels comprises: Confirming the number of nodes in the top-level switching network based on the total number of ring layers of the bottom-level cellular network; Connecting all the cross-layer nodes in the same hexagon ring layer of the bottom-level cellular network to a single corresponding node in the top-level switching network through additional channels.

11. The method of claim 9, wherein, The connecting of the cross-layer nodes in adjacent areas to the same node in the top-level switching network comprises: Dividing the six areas of the bottom-level cellular network into three groups of cross-area node clusters; the number of nodes in the top-level switching network is equal to the number of clusters; Connecting the cross-layer nodes of each cluster to the same node in the top-level switching network through additional channels.

12. An electronic device, comprising: Comprise: A memory for storing a computer program; A processor for executing the computer program to implement the steps of the network-on-chip construction method according to any one of claims 1 to 11.

13. A computer-readable storage medium, characterized in that, The computer program stored in the computer readable storage medium, wherein the computer program is executed by the processor to implement the steps of the network-on-chip construction method according to any one of claims 1 to 11.

14. A computer program product comprising a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the network-on-chip construction method according to any one of claims 1 to 11.

Images