Message transmission method, device, electronic device and storage medium for network on chip

Abstract

The embodiment of the present invention provides a message transmission method, device, electronic device and storage medium of an on-chip network. The method comprises: configuring a device router in a mesh network; determining a routing group connection mode based on a target transmission strategy; when the routing group connection mode is an input-output separation group connection, connecting the output end of a processing module to a device router in an X-axis direction of the mesh network, and connecting the input end of the processing module to a device router in a Y-axis direction of the mesh network; when the routing group connection mode is an input-output non-separation group connection, connecting the input end and the output end of the processing module to the same device router, and configuring message transmission restrictions; determining a target transmission path of the message based on the routing group connection mode. The present invention can reduce the arbitration logic of the on-chip network and improve the message transmission efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of network on chip, and in particular to a message transmission method, device, electronic equipment and storage medium of a network on chip. Background Art

[0002] In recent years, as the scale of on-chip networks has become larger and larger, the rise of artificial intelligence (AI) has led to faster and faster iteration of software requirements. The topology architecture of the network on chip (NoC) must support flexible expansion and be easy to configure to adapt to different overall chip structures.

[0003] The advent of the AI ​​era has also led to an increasing demand for on-chip network bandwidth. The new-era NoC architecture needs to reduce the line density in routers to accommodate data packets with larger bandwidth.

[0004] The industry requires NoC to be able to transmit data packets over longer distances at a fixed main frequency to reduce the number of sockets in the routing and reduce costs. Therefore, the combinational logic depth between routers needs to be reduced as much as possible.

[0005] At present, the Ring bus router has only two directions, counterclockwise and clockwise, on the transmission path of data packets. The turning is not flexible enough and the path diversity is low. In addition, the Ring bus has poor expansion capability, which is not conducive to the flexible establishment of a whole-chip interconnection network. The data packet needs to arbitrate with the data packets in four directions and the data packets of other devices connected to the router from the device Mesh bus. The arbitration logic is deep, and the routing in multiple directions is accumulated inside the router, which is easy to cause layout and wiring congestion, which is not conducive to expanding bandwidth. Complex arbitration logic needs to be considered during the transmission of data packets. Therefore, the logic is deep on the main control path, and the transmission distance is shorter at the same main frequency. Summary of the invention

[0006] The embodiments of the present invention provide a method, device, electronic device and storage medium for transmitting messages of a network on chip, which can solve the problems of deep arbitration logic and low message transmission efficiency of the network on chip in the related art.

[0007] On the one hand, an embodiment of the present invention discloses a method for transmitting messages in a network on chip, which is applied to a network on chip, wherein the network on chip includes at least four cross redirectors, a device router, and a processing module, and the method includes:

[0008] Connecting the at least four cross redirectors in pairs to form a mesh network, and configuring the device router in the mesh network;

[0009] Determining a routing group connection mode of the on-chip network based on a target transmission strategy;

[0010] When the routing group connection mode is an input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network;

[0011] In the case where the routing group connection mode is an input-output inseparable group connection, the input end and the output end of the processing module are connected to the same device router, and a message transmission restriction is configured; the message transmission restriction is used to indicate that a message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis;

[0012] A target transmission path of the message is determined based on the routing group connection mode.

[0013] Optionally, the network on chip further includes a path router, and the method further includes:

[0014] Determining a first distance between two processing modules on the target transmission path;

[0015] When the first distance is greater than or equal to a preset threshold, at least one access router is inserted between any two cross redirectors on the target transmission path.

[0016] Optionally, the routing group connection mode includes the input-output inseparable group connection; and determining the target transmission path of the message based on the routing group connection mode includes:

[0017] Based on the XY routing algorithm and the message transmission restriction, determining at least one preset transmission path in the mesh network, and determining a second distance between two device routers in the preset transmission path;

[0018] A target transmission path for the message is determined from the preset transmission path based on the second distance.

[0019] Optionally, when the routing group connection mode is the input-output separation group connection, a device router in the on-chip network is connected to the mesh network along any one of the X-axis and the Y-axis.

[0020] Optionally, the at least four cross redirectors are connected in pairs to form a mesh network, and the device router is configured in the mesh network, including:

[0021] Connecting the at least four cross-diverters in pairs to form a grid network;

[0022] A device router is configured between any two adjacent cross redirectors in the mesh network.

[0023] Optionally, when the routing group connection mode is the input-output separation group connection, configuring a device router between any two adjacent cross redirectors in the mesh network includes:

[0024] For the first cross redirector in the mesh network, a device router is inserted between the second cross redirectors adjacent to the first cross redirector in the X-axis direction, and a device router is inserted between the third cross redirectors adjacent to the first cross redirector in the Y-axis direction; the first cross redirector is any cross redirector in the mesh network.

[0025] On the other hand, an embodiment of the present invention discloses a message transmission device of a network on chip, which is applied to a network on chip, wherein the network on chip includes at least four cross redirectors, a device router and a processing module, and the device includes:

[0026] A configuration module, used for connecting the at least four cross redirectors in pairs to form a mesh network, and configuring the device router in the mesh network;

[0027] A first determination module, configured to determine a routing group connection mode of the on-chip network based on a target transmission strategy;

[0028] a connection module, used for connecting the output end of the processing module to the device router in the X-axis direction of the mesh network and the input end of the processing module to the device router in the Y-axis direction of the mesh network when the routing group connection mode is the input-output separation type group connection; and for connecting the input end and the output end of the processing module to the same device router when the routing group connection mode is the input-output non-separation type group connection;

[0029] The configuration module is further used to configure message transmission restrictions; the message transmission restrictions are used to indicate that messages cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis;

[0030] The first determination module is further used to determine a target transmission path of the message based on the routing group connection mode.

[0031] Optionally, the device further comprises:

[0032] A second determining module, used to determine a first distance between two processing modules on the target transmission path;

[0033] The inserting module is used to insert at least one access router between any two cross redirectors on the target transmission path when the first distance is greater than or equal to a preset threshold.

[0034] On the other hand, an embodiment of the present invention further discloses an electronic device, comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; the memory is used to store executable instructions, and the executable instructions enable the processor to execute the aforementioned on-chip network message transmission method.

[0035] The embodiment of the present invention further discloses a readable storage medium. When instructions in the readable storage medium are executed by a processor of an electronic device, the electronic device can execute the above-mentioned message transmission method of the on-chip network.

[0036] The embodiments of the present invention include the following advantages:

[0037] The embodiment of the present invention can determine the routing group connection mode of the on-chip network according to the target transmission strategy. Under different target transmission strategies, the routing group connection mode of the on-chip network is different. When the routing group connection mode is an input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network; when the routing group connection mode is an input-output non-separation group connection, the input end and the output end of the processing module are connected to the same device router, and the message transmission restriction is configured; the message transmission restriction is used to indicate that the message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis. In the embodiment of the present invention, by separating the cross redirector and the device router, the arbitration logic of the on-chip network can be reduced, thereby improving the message transmission efficiency. Further, through the target transmission path, the routing deadlock problem can be avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative labor.

[0039] Figure 1 It is a flowchart of a method for transmitting messages in a network on chip according to an embodiment of the present invention;

[0040] Figure 2 It is a structural block diagram of a cross redirector of a network on chip of the present invention;

[0041] Figure 3 It is a structural block diagram of a device router of a network on chip of the present invention;

[0042] Figure 4 It is a structural block diagram of a path router of a network on chip of the present invention;

[0043] Figure 5 It is a schematic diagram of an input-output separation group connection structure of an on-chip network of the present invention;

[0044] Figure 6 It is a schematic diagram of an input-output non-separable group connection structure of a network on chip of the present invention;

[0045] Figure 7 It is a structural block diagram of an embodiment of a message transmission device of a network on chip of the present invention;

[0046] Figure 8 It is a structural block diagram of an electronic device provided by an example of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] The terms "first", "second", etc. in the specification and claims of the present invention are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable when appropriate, so that the embodiments of the present invention can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, the term "and / or" in the specification and claims is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. In the embodiments of the present invention, the term "multiple" refers to two or more, and other quantifiers are similar.

[0049] Method Embodiment

[0050] Reference Figure 1 , shows a flowchart of a method for transmitting a message in a network on chip according to an embodiment of the present invention, the method may specifically include the following steps:

[0051] Step S101, connect at least four cross redirectors in pairs to form a mesh network, and configure the device router in the mesh network;

[0052] Step S102: determining a routing group connection mode of the on-chip network based on the target transmission strategy;

[0053] Step S103: When the routing group connection mode is an input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network;

[0054] Step S104: When the routing group connection mode is an input-output inseparable group connection, the input end and the output end of the processing module are connected to the same device router, and a message transmission restriction is configured; the message transmission restriction is used to indicate that a message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis;

[0055] Step S105: determine the target transmission path of the message based on the routing group connection mode.

[0056] The on-chip network message transmission method provided by the embodiment of the present invention is applied to the on-chip network, and the on-chip network includes at least four cross redirectors, a device router and a processing module.

[0057] In an embodiment of the present invention, a cross redirector can be used to arbitrate and cross-transmit data packets in a bus at a network cross node without interacting with the device for data packets. The components of the cross redirector can include an input buffer, a bus message arbitrator, and a cross switch.

[0058] The input buffer of the cross redirector may include four input buffers, corresponding to the four directions of the bus, namely, the southeast, northwest, and northeast directions. The bus message arbitrator is used to flexibly adapt the routing algorithm and arbitration strategy of the mesh topo structure network on chip. The cross switch of the cross redirector is used to forward the data packets that obtain arbitration priority in the four input channels of the cross redirector router to the corresponding output port according to different arbitration strategies.

[0059] The device router is used to arbitrate and forward data packets in devices connected to the bus and data packets transmitted on the bus. The components of the device router may include input buffers, fast channels, configurable arbiters, and crossbar switches.

[0060] The input buffer of the device router may include four input buffers, and the four input buffers correspond to two opposite directions on the bus and two devices connected to the bus. The fast channel can form a fast channel for the device router to form a data packet between two devices connected to the bus to facilitate private communication between the two devices. The arbitration in the device router can be dynamically configured as priority arbitration with different weights. The cross switch of the device router can forward the data packets that obtain arbitration priority in the four input channels of the device router to the corresponding output port according to different arbitration strategies.

[0061] The processing module in the embodiment of the present invention may be a processor, a processing unit or an electronic device with a data processing function.

[0062] In an embodiment of the present invention, at least four cross redirectors are first connected in pairs to form a grid network.

[0063] For example, in the embodiment of the present invention, the Mesh network is a mesh network structure formed by interconnecting multiple nodes (usually routers). These nodes can be wirelessly interconnected to form a dynamic and scalable network architecture.

[0064] In an embodiment of the present invention, the target transmission strategy may be a message transmission strategy, which is used to indicate that the routing group connection mode corresponding to the message is determined according to the transmission distance or the transmission power consumption. Exemplarily, the target transmission strategy may include determining the routing group connection mode corresponding to the message according to the transmission power consumption. In this case, an input-output non-separated group connection mode may be selected, that is, not introducing too many device routers, which can realize the transmission of the message and avoid the routing deadlock problem.

[0065] Alternatively, the target transmission strategy may include determining the routing grouping method corresponding to the message based on the transmission distance. In this case, in order to avoid the routing deadlock problem, the input-output separation grouping method can be selected, and then the message is transmitted according to the strategy corresponding to the input-output separation grouping.

[0066] It should be noted that the routing grouping mode of the on-chip network can include an input-output separated grouping mode and an input-output non-separated grouping mode.

[0067] Among them, the input-output separation group connection method means that the input and output ends of the processing module are respectively connected to different device routers. For example, the output port of the processing module can be connected to the X-axis of the Mesh network, and the input port can be connected to the Y-axis of the Mesh network, so as to realize deadlock-free data interaction of the entire chip network under the XY routing algorithm.

[0068] In the embodiment of the present invention, the device router in the X-axis direction in the mesh network may include the device router in the horizontal axis direction, and the device router in the X-axis direction may include multiple ones.

[0069] The device router in the Y-axis direction in the mesh network may include the device router in the longitudinal direction, and the device router in the Y-axis direction may include multiple ones.

[0070] The input-output inseparable group connection method refers to connecting the input and output ends of the processing module to the same device processor.

[0071] The message transmission restriction may include that messages cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis.

[0072] The Y-axis direction of the grid network may include the longitudinal axis direction in the grid network. The positive direction of the X-axis may be based on any processing module in the grid network as the origin, and the direction to the right of the processing module may be defined as the positive direction of the X-axis.

[0073] In the embodiment of the present invention, the processing module can be connected to the mesh network through the device router according to the routing group connection mode of the on-chip network. Specifically, when the routing group connection mode is the input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network, that is, the processing module is connected to two different device routers; when the routing group connection mode is the input-output non-separation group connection, the input end and the output end of the processing module are connected to the same device router.

[0074] After the processing module is connected to the grid network, the target transmission path is determined according to the positions of the source processing module sending the message and the target processing module receiving the message in the grid network, and the message is transmitted from the source processing module to the target processing module according to the target transmission path.

[0075] For example, when the routing group connection mode is an input-output separation group connection, the target transmission path can be determined according to the positions of the device router connected to the output end of the source processing module and the device router connected to the input end of the target processing module in the mesh network, and the routing algorithm. For example, the path with the shortest transmission distance between the device router connected to the output end of the source processing module and the device router connected to the input end of the target processing module is determined as the target transmission path.

[0076] In a mesh network, data can only be transmitted along the edges of the mesh. In other words, two nodes on the same diagonal line in the mesh cannot communicate directly.

[0077] When the routing group connection mode is an input-output non-separable group connection, the target transmission path can be determined according to the position of the device router connected to the output end of the source processing module in the mesh network, the position of the device router connected to the input end of the target processing module in the mesh network, the message transmission restriction, and the routing algorithm. For example, the path with the shortest transmission power consumption between the two processing modules is determined as the target transmission path.

[0078] In an embodiment of the present invention, the on-chip network can determine the routing group connection mode of the on-chip network according to the target transmission strategy. Under different target transmission strategies, the routing group connection mode of the on-chip network is different. When the routing group connection mode is an input-output separated group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network; when the routing group connection mode is an input-output inseparable group connection, the input end and the output end of the processing module are connected to the same device router, and a message transmission restriction is configured; the message transmission restriction is used to indicate that the message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis; it can improve the problem of the arbitration logic depth of the on-chip network and improve the message transmission efficiency.

[0079] Furthermore, in the embodiment of the present invention, a path router may be inserted into the mesh network, and the path router may be used to increase the main frequency of the entire chip. Optionally, the message transmission method of the on-chip network provided by the embodiment of the present invention may also include the following steps:

[0080] Step S106: Determine a first distance between two processing modules on the target transmission path.

[0081] In the embodiment of the present invention, during the process of transmitting messages on the network on chip, the network on chip may also determine a first distance between two processing modules on a target transmission path.

[0082] In the embodiment of the present invention, the first distance between two processing modules may include the distance between any two processing modules, which may be two adjacent processing modules or two non-adjacent processing modules.

[0083] Step S107: When the first distance is greater than or equal to a preset threshold, insert at least one access router between any two cross redirectors on the target transmission path.

[0084] In an embodiment of the present invention, after determining the first distance between two processing modules on the target transmission path, the on-chip network can also insert at least one path router between any two cross redirectors on the target transmission path when the first distance is greater than or equal to a preset threshold.

[0085] In the embodiment of the present invention, the preset threshold may be any distance threshold greater than or equal to 0.

[0086] In the embodiment of the present invention, the access router is used to increase the main frequency of the entire chip by setting up the access router when the distance between two routers is long and the message cannot pass within one cycle; there is no need for arbitration in the access router, and the data packet input from the bus will be directly output from the opposite direction.

[0087] In the embodiment of the present invention, when the first distance is greater than or equal to a preset threshold, at least one access router is inserted between any two cross redirectors on the target transmission path.

[0088] In an embodiment of the present invention, a path router can be connected to a mesh network according to the routing group connection mode of the on-chip network. Specifically, when the distance between two processing modules on the target path is greater than a preset threshold, at least one path router can be inserted between any two cross redirectors on the target transmission path. For example, multiple cross redirectors need to be passed between two processing modules on the target path, that is, the distance for the message to be transmitted on the target path is longer. Assuming that it passes through a cross redirector, the distance is increased by 1. If there are three cross redirectors on the target path and the preset threshold is 2, at least one path router is inserted between any two cross redirectors on the target transmission path.

[0089] Furthermore, in the embodiment of the present invention, the routing group connection mode includes input-output non-separable group connection; in the process of determining the target transmission path of the message based on the routing group connection mode, the method may further include the following steps:

[0090] Step S1051: Based on the XY routing algorithm and the message transmission restriction, at least one preset transmission path is determined in the mesh network, and a second distance between two device routers in the preset transmission path is determined.

[0091] In the process of determining the target transmission path of the message based on the routing group connection method, the on-chip network can also determine at least one preset transmission path in the mesh network based on the XY routing algorithm and the message transmission restriction, and determine the second distance between two device routers in the preset transmission path.

[0092] Among them, the XY routing algorithm is an algorithm used to determine the path of a data packet from a starting point to an end point in a network topology.

[0093] For example, the XY routing algorithm is a deterministic routing algorithm that represents nodes or routers in the network with two-dimensional coordinates (X, Y), and the address information in the data packet is also represented by two-dimensional coordinates. The core idea of ​​the algorithm is that during the transmission process, the data packet first moves along the X dimension (horizontal direction) until it reaches the same X coordinate position as the destination node, and then moves along the Y dimension (vertical direction) until it reaches the destination node. This routing method can ensure that the data packet is transmitted along a predetermined path in the network, avoiding the randomness and uncertainty of the path.

[0094] In a network, each node or router is assigned a unique two-dimensional coordinate (X, Y). Similarly, the destination address in a data packet is also represented by a two-dimensional coordinate.

[0095] In an embodiment of the present invention, the process of determining the target path may include that when a message or data packet needs to be transmitted from a source processing module to a target processing module, an XY routing algorithm determines a unique path according to the coordinate information of the source processing module and the destination node. The path first moves along the X dimension until it is the same as the X coordinate of the target processing module, and then moves along the Y dimension until it reaches the target processing module.

[0096] For example, in order to avoid loop deadlock, the XY routing algorithm restricts the direction of data packets. Specifically, only data packets moving eastward (or westward) are allowed to turn northward (or southward), while data packets moving northward (or southward) are not allowed to turn eastward (or westward). This direction restriction ensures that data packets will not fall into an infinite loop during transmission.

[0097] Advantages of the XY routing algorithm may include being easy to understand, avoiding deadlock, and load balancing, etc. In the embodiment of the present invention, loop deadlock may be effectively avoided by limiting the transmission direction of data packets or messages.

[0098] In the embodiment of the present invention, the preset transmission path may include one determined in the mesh network according to the XY routing algorithm and the message transmission restriction, and the preset transmission path may include at least one.

[0099] In the embodiment of the present invention, the second distance between two device routers in the preset transmission path may be a distance value of any integer greater than or equal to 0.

[0100] Step S1052: determine a target transmission path for the message from preset transmission paths based on the second distance.

[0101] In an embodiment of the present invention, after determining at least one preset transmission path in the mesh network based on the XY routing algorithm and the message transmission restriction, and determining the second distance between two device routers in the preset transmission path, the on-chip network can also determine the target transmission path of the message from the preset transmission path based on the second distance.

[0102] In the embodiment of the present invention, the preset transmission path may include at least one transmission path.

[0103] In an embodiment of the present invention, when the routing group connection mode is an input-output inseparable group connection, the target transmission path can determine at least one preset transmission path based on the positions of the device router connected to the output end of the source processing module and the device router connected to the input end of the target processing module in the mesh network, the message transmission restriction, and the routing algorithm, and then the target path can be determined from the preset transmission paths based on the distance between the two device routers in the preset transmission paths.

[0104] Optionally, in the embodiment of the present invention, the message transmission method of the network on chip provided by the embodiment of the present invention may further include the following steps:

[0105] Step S1031: When the routing group connection mode is input-output separation group connection, a device router in the on-chip network is connected to the mesh network along any one of the X axis and the Y axis.

[0106] In an embodiment of the present invention, during the process of transmitting messages in the on-chip network, the on-chip network can also connect a device router in the on-chip network to the mesh network along any one of the X-axis and the Y-axis when the routing group connection mode is an input-output separation group connection.

[0107] In the embodiment of the present invention, device routers from different directions in the mesh network may be mounted on one processing module.

[0108] In the embodiment of the present invention, different device routers may be mounted on different processing modules.

[0109] In the embodiment of the present invention, the processing module can be connected to the mesh network through the device router according to the routing group connection mode of the on-chip network. Specifically, when the routing group connection mode is an input-output separation group connection, a device router in the on-chip network is connected to the mesh network along any one of the X axis and the Y axis.

[0110] Optionally, after at least four cross redirectors are connected in pairs to form a mesh network, and the device router is configured in the mesh network, the message transmission method of the network on chip provided by the embodiment of the present invention may further include the following steps:

[0111] Step S1011, connect at least four cross redirectors in pairs to form a grid network.

[0112] In the embodiment of the present invention, during the process of transmitting messages in the network on chip, the network on chip may also connect at least four cross redirectors in pairs to form a mesh network.

[0113] In an embodiment of the present invention, the cross-steering devices connected in pairs may include cross-steering devices connected in the horizontal axis or the vertical axis direction.

[0114] Step S1012: configure a device router between any two adjacent cross redirectors in the mesh network.

[0115] In the embodiment of the present invention, after at least four cross redirectors are connected in pairs to form a mesh network, the on-chip network can also configure a device router between any two adjacent cross redirectors in the mesh network.

[0116] In the embodiment of the present invention, any two adjacent cross-directors may include two adjacent cross-directors in the transverse axis direction or two adjacent cross-directors in the longitudinal axis direction.

[0117] In the embodiment of the present invention, the device router can be configured between any two adjacent cross redirectors in the mesh network according to the routing group connection mode of the on-chip network.

[0118] Further, in the embodiment of the present invention, when the routing group connection mode is an input-output separation group connection, in the process of configuring a device router between any two adjacent cross redirectors in the mesh network, the message transmission method of the on-chip network provided by the embodiment of the present invention may also include the following steps:

[0119] Step S1021: For the first cross redirector in the mesh network, insert a device router between the second cross redirectors adjacent to the first cross redirector in the X-axis direction, and insert a device router between the third cross redirectors adjacent to the first cross redirector in the Y-axis direction; the first cross redirector is any cross redirector in the mesh network.

[0120] In an embodiment of the present invention, when the routing group connection mode is an input-output separated group connection, in the process of configuring a device router between any two adjacent cross redirectors in the mesh network, the on-chip network can also, for the first cross redirector in the mesh network, insert a device router between the second cross redirector adjacent to the cross redirector in the X-axis direction, and insert a device router between the third cross redirector adjacent to the cross redirector in the Y-axis direction; the first cross redirector is any cross redirector in the mesh network.

[0121] In the embodiment of the present invention, the first cross redirector C1 may be any cross redirector in the mesh network.

[0122] The X-axis direction may be a horizontal axis direction, that is, a horizontal axis direction in a grid network. The Y-axis direction may be a vertical axis direction, that is, a vertical axis direction in a grid network.

[0123] The first cross-director and the second cross-director may be adjacent to each other in the transverse axis direction. The first cross-director and the third cross-director may be adjacent to each other in the longitudinal axis direction.

[0124] In the embodiment of the present invention, the device router can be connected to the mesh network according to the routing group connection mode of the on-chip network. Specifically, when the routing group connection mode is the input-output separation group connection, for the first cross redirector C1 in the mesh network, a device router DX is inserted between the second cross redirector C1-X adjacent to the cross redirector in the X-axis direction, and a device router DY is inserted between the third cross redirector C1-Y adjacent to the first cross redirector in the Y-axis direction; the first cross redirector is any cross redirector in the mesh network.

[0125] In the prior art, the Ring bus topo may include a data packet being sent from a device onto the bus, and only arbitrating with messages transmitted on the bus and messages from other devices mounted on the router. There is less arbitration logic, and the congestion density generated after layout and routing is small, and the bus can support messages with a larger bandwidth. During transmission on the bus, the data packet can only move in one direction, and its path control logic is simple, so the data packet can be transmitted over a longer distance at the same main frequency.

[0126] However, the Ring bus router only has two directions for data packet transmission: counterclockwise and clockwise. The direction is not flexible enough and the path diversity is low. In addition, the Ring bus has poor expansion capability, which is not conducive to the flexible establishment of a whole-chip interconnection network.

[0127] The Mesh bus topo solution may include a Mesh bus router with four data packet transmission directions, which has flexible steering, can use more complex routing algorithms, and has strong expansion capabilities, and is adaptable to large-scale on-chip network interconnection.

[0128] However, in the Mesh bus topo solution, the data packets on the device need to arbitrate with the data packets in four directions and the data packets of other devices connected to the router. The arbitration logic is deep, and the routing in multiple directions is accumulated inside the router, which is easy to cause layout and wiring congestion and is not conducive to expanding the bandwidth. The data packet needs to consider complex arbitration logic during the transmission process, so the logic is deep on the main control path, and the transmission distance is shorter at the same main frequency.

[0129] At present, the industry's demands may include the following: In recent years, as the scale of on-chip networks has become larger and larger, the rise of AI has led to faster iteration of software-side requirements. The topology of NoC must support flexible expansion and be easy to configure to adapt to different overall chip structures; the advent of the AI ​​era has also led to an increasing demand for on-chip network bandwidth. The new era of NoC architecture needs to reduce the line density in routers to accommodate data packets with larger bandwidth; the industry needs NoC to be able to transmit data packets over longer distances at a fixed main frequency, so as to reduce the number of strips in the routing and reduce costs, so it is necessary to reduce the combinational logic depth between routers as much as possible.

[0130] The present invention proposes a high-performance, configurable, Mesh-like on-chip network topology architecture, which can be flexibly expanded to be applicable to large-bandwidth, high-frequency, large-scale on-chip network whole-chip interconnection.

[0131] Exemplarily, in an embodiment of the present invention, three types of routing components are provided for flexibly building whole-chip interconnections.

[0132] Reference Figure 2 , shows a cross redirector structure block diagram of a network on chip of the present invention, which may specifically include the following:

[0133] The crossbar redirector is used to arbitrate and cross-transmit data packets in the bus at the network crossbar node, and does not interact with the device for data packets. The components of the crossbar redirector may include an input buffer, a bus message arbitrator, and a crossbar switch.

[0134] The input buffer may include four input buffers corresponding to the four directions of the bus, namely, the southeast, northwest, and northeast directions. The bus message arbitrator can flexibly adapt the routing algorithm and arbitration strategy of the mesh topo structure network-on-chip. The crossbar switch forwards the data packets that obtain arbitration priority in the four input channels of the crossbar router to the corresponding output port according to different arbitration strategies.

[0135] Reference Figure 3 , shows a structural block diagram of a device router of a network on chip of the present invention, which may specifically include the following:

[0136] The device router is used to arbitrate and forward data packets in the devices connected to the bus and data packets transmitted on the bus. The device router components may include input buffers, fast channels, configurable arbiters, and cross switches.

[0137] The input buffer may include four input buffers corresponding to two opposite directions on the bus and two devices connected to the bus. The fast channel can form a fast channel for the device router to form a data packet between device 0 and device 1 to facilitate private communication between the two devices. The arbitration in the device router can be dynamically configured as priority arbitration with different weights. The crossbar switch forwards the data packets that obtain arbitration priority in the four input channels of the device router to the corresponding output port according to different arbitration strategies.

[0138] Reference Figure 4 , shows a structural block diagram of a path router of a network on chip of the present invention, which may specifically include the following:

[0139] The access router is used to increase the main frequency of the entire chip when the distance between two routers is long and the message cannot pass within one cycle. There is no need for arbitration in the access router, and the data packet input from the bus will be directly output from the opposite direction.

[0140] The present invention also provides two routing group connection solutions to avoid the routing deadlock problem under this type of Mesh topo.

[0141] Reference Figure 5 , shows a schematic diagram of an input-output separation group connection structure of an on-chip network of the present invention, which may specifically include the following:

[0142] The input-output separation group connection structure may include at least four cross redirectors, a device router, a channel router and a processing module.

[0143] In an embodiment of the present invention, the device output port is connected to the X-axis of the Mesh network, and the input port is connected to the Y-axis of the Mesh network, so as to achieve deadlock-free data interaction of the entire chip network under the XY routing algorithm.

[0144] Reference Figure 6 , showing a schematic diagram of an input-output non-separable group connection structure of an on-chip network of the present invention, which may specifically include the following:

[0145] The input-output inseparable group connection structure may include at least four cross redirectors, a device router, a channel router and a processing module.

[0146] In the embodiment of the present invention, if the input and output terminals of the device are connected to a single axis of the Mesh network, the data packet may make two turns during the routing process, so the routing is restricted to turn from the Y-axis direction to the negative direction of the X-axis, and the deadlock loop path is broken by restricting the turn from the Y-axis to the positive direction of the X-axis. All data packets from the bus on the device first reach the nearest cross-turn node on the right side of the device through the XY routing algorithm, and then are transmitted to the negative direction of the X-axis through the cross-turn node.

[0147] In actual multi-core on-chip network planning, the three components described in the present invention can be used to perform flexible global interconnection division, distributing the logic and a large number of connections concentrated in traditional routers to cross-steering routers and device routers to increase the main frequency of the entire chip, and effectively interrupting the long-distance transmission of data packets on the path in large-core application scenarios through access routers. In addition, the storage device interface and the entire chip cache device can be hung on the same device router, and efficient private communication can be carried out through fast channels.

[0148] By separating the device attachment node and the cross-steering node on the bus, the following problems are solved: the layout and routing congestion caused by the traditional router attaching all devices to the same router, and the difficulty in widening the channel width. Therefore, the present invention can be applied to large-bandwidth AI-oriented whole-chip on-chip network interconnection. Traditional routers attach all devices to the same router, resulting in deep router arbitration logic and difficulty in increasing the main frequency of the whole chip. Traditional router devices can only be attached to cross nodes, which is not conducive to flexible layout of whole-chip devices. By attaching the device to the device router described in the present invention, the device position and the upper bus interface can be arbitrarily placed globally, and the architecture is more flexible, which is conducive to the rapid layout of heterogeneous on-chip networks of devices.

[0149] Setting a fast channel in a device router is beneficial for high-speed private communication between devices connected to the same device router, while the routing layer communication between devices in the prior art will be affected by bus data packets.

[0150] In the embodiment of the present invention, a bus cross-router and a device-router are separated from each other, and a Mesh-like high-performance configurable topo architecture is formed under the components.

[0151] The device router is configured with a fast channel, which is a communication solution that separates private communication between devices from bus data packets.

[0152] In the embodiment of the present invention, a deadlock solution in a Mesh topology-like environment can be implemented through a message transmission method of an on-chip network.

[0153] In summary, the embodiment of the present invention can determine the routing group connection mode of the on-chip network according to the target transmission strategy. Under different target transmission strategies, the routing group connection mode of the on-chip network is different. When the routing group connection mode is an input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network; when the routing group connection mode is an input-output non-separation group connection, the input end and the output end of the processing module are connected to the same device router, and the message transmission restriction is configured; the message transmission restriction is used to indicate that the message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis; in the embodiment of the present invention, by separating the cross redirector and the device router, the arbitration logic of the on-chip network can be reduced, thereby improving the message transmission efficiency. Further, through the target transmission path, the routing deadlock problem can be avoided.

[0154] Device Embodiment

[0155] Reference Figure 7 , shows a structural block diagram of a message transmission device of a network on chip of the present invention, the device is applied to the network on chip, the network on chip includes at least four cross redirectors, a device router and a processing module, the device may specifically include:

[0156] A configuration module 701 is used to connect the at least four cross redirectors in pairs to form a mesh network, and configure the device router in the mesh network;

[0157] A first determining module 702, configured to determine a routing group connection mode of the on-chip network based on a target transmission strategy;

[0158] The connection module 703 is used to connect the output end of the processing module to the device router in the X-axis direction of the mesh network, and connect the input end of the processing module to the device router in the Y-axis direction of the mesh network when the routing group connection mode is an input-output separation group connection; and connect the input end and the output end of the processing module to the same device router when the routing group connection mode is an input-output non-separation group connection;

[0159] The configuration module is further used to configure message transmission restrictions; the message transmission restrictions are used to indicate that messages cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis;

[0160] The first determination module is further used to determine a target transmission path of the message based on the routing group connection mode.

[0161] Optionally, the device further comprises:

[0162] A second determining module, used to determine a first distance between two processing modules on the target transmission path;

[0163] The inserting module is used to insert at least one access router between any two cross redirectors on the target transmission path when the first distance is greater than or equal to a preset threshold.

[0164] Optionally, the determining module further includes:

[0165] The first determination submodule is used to determine at least one preset transmission path in the mesh network based on the XY routing algorithm and the message transmission restriction, and determine a second distance between two device routers in the preset transmission path; and determine a target transmission path of the message from the preset transmission path based on the second distance.

[0166] Optionally, the device further comprises:

[0167] The mounting module is used to connect a device router in the on-chip network to the mesh network along any one of the X-axis and the Y-axis when the routing group connection mode is the input-output separation group connection.

[0168] Optionally, the configuration module further includes:

[0169] A connection module, used for connecting the at least four cross diverters in pairs to form a grid network;

[0170] The configuration submodule is used to configure a device router between any two adjacent cross redirectors in the mesh network.

[0171] Optionally, the configuration submodule further includes:

[0172] An insertion submodule is used to insert a device router between a second cross redirector adjacent to the first cross redirector in the X-axis direction, and to insert a device router between a third cross redirector adjacent to the first cross redirector in the Y-axis direction for the first cross redirector in the mesh network; the first cross redirector is any cross redirector in the mesh network.

[0173] In summary, the embodiment of the present invention can determine the routing group connection mode of the on-chip network according to the target transmission strategy. Under different target transmission strategies, the routing group connection mode of the on-chip network is different. When the routing group connection mode is an input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network; when the routing group connection mode is an input-output non-separation group connection, the input end and the output end of the processing module are connected to the same device router, and the message transmission restriction is configured; the message transmission restriction is used to indicate that the message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis; in the embodiment of the present invention, by separating the cross redirector and the device router, the arbitration logic of the on-chip network can be reduced, thereby improving the message transmission efficiency. Further, through the target transmission path, the routing deadlock problem can be avoided.

[0174] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0175] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0176] Regarding the processor in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0177] Reference Figure 8 , is a structural block diagram of an electronic device for message transmission of an on-chip network provided by an embodiment of the present invention. Figure 8 As shown, the electronic device includes: a processor, a memory, a communication interface and a communication bus, and the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store executable instructions, and the executable instructions enable the processor to execute the on-chip network message transmission method of the aforementioned embodiment.

[0178] The processor may be a central processing unit (CPU), a general processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other editable devices, transistor logic devices, hardware components or any combination thereof. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0179] The communication bus may include a path to transmit information between the memory and the communication interface. The communication bus may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 8 The fact that only one line is used in the diagram does not mean that there is only one bus or only one type of bus.

[0180] The memory may be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or may be an electrically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, etc.

[0181] The embodiment of the present invention also provides a non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by a processor of an electronic device (server or terminal), the processor can execute Figure 1 The message transmission method of the on-chip network is shown.

[0182] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0183] It should be understood by those skilled in the art that the embodiments of the present invention may be provided as methods, devices or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0184] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems) and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0185] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing terminal device to operate in a predictable manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0186] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0187] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0188] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0189] The above is a detailed introduction to the message transmission method, device, electronic device and storage medium of a network on chip provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A message transmission method for a network on chip, characterized in that: Applied to a network on chip, the network on chip includes at least four cross redirectors, a device router and a processing module, the method includes: Connecting the at least four cross redirectors in pairs to form a mesh network, and configuring the device router in the mesh network; Determining a routing group connection mode of the on-chip network based on a target transmission strategy; When the routing group connection mode is an input-output separation group connection, the output end of the processing module is connected to the device router in the X-axis direction of the mesh network, and the input end of the processing module is connected to the device router in the Y-axis direction of the mesh network; In the case where the routing group connection mode is an input-output inseparable group connection, the input end and the output end of the processing module are connected to the same device router, and a message transmission restriction is configured; the message transmission restriction is used to indicate that a message cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis; A target transmission path of the message is determined based on the routing group connection mode.

2. The method according to claim 1, characterized in that The network on chip further includes a path router, and the method further includes: Determining a first distance between two processing modules on the target transmission path; When the first distance is greater than or equal to a preset threshold, at least one access router is inserted between any two cross redirectors on the target transmission path.

3. The method according to claim 1, characterized in that The routing group connection mode includes the input-output non-separable group connection; the determining of the target transmission path of the message based on the routing group connection mode includes: Based on the XY routing algorithm and the message transmission restriction, determining at least one preset transmission path in the mesh network, and determining a second distance between two device routers in the preset transmission path; A target transmission path for the message is determined from the preset transmission path based on the second distance.

4. The method according to claim 1, characterized in that: In the case where the routing group connection mode is the input-output separation group connection, a device router in the on-chip network is connected to the mesh network along any one of the X-axis and the Y-axis.

5. The method according to claim 1, characterized in that: The step of connecting the at least four cross redirectors in pairs to form a mesh network, and configuring the device router in the mesh network includes: Connecting the at least four cross-diverters in pairs to form a grid network; A device router is configured between any two adjacent cross redirectors in the mesh network.

6. The method according to claim 5, characterized in that When the routing group connection mode is the input-output separation group connection, a device router is configured between any two adjacent cross redirectors in the mesh network, including: For the first cross redirector in the mesh network, a device router is inserted between the second cross redirectors adjacent to the first cross redirector in the X-axis direction, and a device router is inserted between the third cross redirectors adjacent to the first cross redirector in the Y-axis direction; the first cross redirector is any cross redirector in the mesh network.

7. A message transmission device for a network on chip, characterized in that: Applied to a network on chip, the network on chip includes at least four cross redirectors, a device router and a processing module, and the device includes: A configuration module, used for connecting the at least four cross redirectors in pairs to form a mesh network, and configuring the device router in the mesh network; A first determination module, configured to determine a routing group connection mode of the on-chip network based on a target transmission strategy; A connection module, used for connecting the output end of the processing module to the device router in the X-axis direction of the mesh network, and connecting the input end of the processing module to the device router in the Y-axis direction of the mesh network when the routing group connection mode is an input-output separation group connection; and for connecting the input end and the output end of the processing module to the same device router when the routing group connection mode is an input-output non-separation group connection; The configuration module is further used to configure message transmission restrictions; the message transmission restrictions are used to indicate that messages cannot be transmitted from the Y-axis direction of the mesh network to the positive direction of the X-axis; The first determination module is further used to determine a target transmission path of the message based on the routing group connection mode.

8. The device according to claim 7, characterized in that The device also includes: A second determining module, used to determine a first distance between two processing modules on the target transmission path; The inserting module is used to insert at least one access router between any two cross redirectors on the target transmission path when the first distance is greater than or equal to a preset threshold.

9. An electronic device, characterized in that: The electronic device includes a processor, a memory, a communication interface and a communication bus, and the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store executable instructions, and the executable instructions enable the processor to execute the on-chip network message transmission method as described in any one of claims 1 to 6.

10. A readable storage medium, characterized in that: When the instructions in the readable storage medium are executed by a processor of an electronic device, the processor is enabled to execute the message transmission method of the on-chip network as described in any one of claims 1 to 6.

Images