Data transmission method of network-on-chip bus equipment and system-on-chip
By numerating and reordering data packets in the on-chip network, the problem of limited P2P data transmission bandwidth in the prior art is solved, efficient and strong sequence-saving data transmission is realized, and data transmission between on-chip network devices is improved efficiency.
Patent Information
- Application Number
- CN202411354422.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-27
- Publication Date
- 2024-11-15
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing P2P data transmission scheme in the on-chip network seriously limits the transmission bandwidth while achieving strong order maintenance. The number of unfinished transactions is 1, limiting the data transmission efficiency between devices.
By numbering multiple data packets for the requesting node of the requester device in an on-chip network and recording the numbers in the address field of the data packet, the slave node of the receiver device completes the transaction of the data packet in sequence according to the number, and if necessary Reordering is performed, allowing the requester device to send multiple unfinished transactions simultaneously.
It has achieved a significant increase in P2P data transmission bandwidth under the conditions of strong order protection, and the unfinished transactions can exceed 1, which improves the data transmission efficiency between on-chip network devices, and simplifies the packet routing and processing flow.
Smart Images

Figure CN118972336A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of network-on-chip data transmission methods, and in particular relates to a data transmission method for a network-on-chip bus device and a system-on-chip. Background Art
[0002] A network on chip (NoC) is a communication system circuit. NoC technology applies the theory and methods of computer networks to on-chip communication, improving the scalability of system-on-chip (SoC) and the energy efficiency of complex SoCs. Currently, many NoCs can support mounted bus devices, such as PCIe devices. Figure 5 The figure shows an on-chip network with multiple PCIe devices, main memory, and other requesting devices mounted on it. These devices are interconnected through the on-chip bus. The on-chip bus will create corresponding request nodes and slave nodes for each PCIe device. The PCIe device can send read and write requests to the bus through the corresponding request node as a requester to access the main memory space or the I / O space of another PCIe device. It can also receive and process the access initiated by other requesters through the bus through the corresponding slave node as a receiver. The scenario in which a PCIe device initiates access to the I / O space of another PCIe device is called P2P (Peer-to-Peer) transmission. P2P transmission usually has the requirement of strong order-preserving transmission, that is, each read and write transaction must be executed after the previous read and write transactions are completed.
[0003] For P2P transmission, the existing order-preserving transmission scheme is usually a scheme called Request / Endpoint Ordered. Figure 6 The data transmission process between the requesting node of the requester device and the slave node of the receiver device in the scheme is shown. Taking write transmission as an example, between the requesting node of the requester device and the slave node of the receiver device, a data transmission includes the following three steps: 1) The requesting node sends a write request to the target node (slave node); 2) The target node allocates relevant resources for the request and sends a response to the requesting node, indicating that the write data can be received; 3) The requesting node sends the write data packet to the target node, and the target node ends the table entry of the transaction. After steps 1) to 3) are completed, the requesting node sends the next write request and repeats the above steps. Through this mechanism, although strong order-preserving transmission is achieved, that is, it ensures that the target node will receive and complete the transaction in the order expected by the requesting node, however, this mechanism limits the number of outstanding (unfinished transactions) of P2P transmission to 1, that is, the requesting node cannot send the next request when it does not receive a response, which seriously limits the transmission bandwidth of P2P.
[0004] Therefore, how to implement a P2P data transmission method with strong order preservation and high transmission bandwidth between bus devices in an on-chip network is an urgent problem to be solved. Summary of the invention
[0005] The present invention is made to solve the above problems, and aims to provide a P2P data transmission method in an on-chip network that can achieve strong order-preserving P2P data transmission while greatly improving the transmission bandwidth. The present invention adopts the following technical solutions: The present invention provides a data transmission method for an on-chip network bus device, which is used for data transmission between two bus devices connected to an on-chip bus, wherein the two bus devices are a requester device and a receiver device, respectively, and the two bus devices have a request node and a slave node on the on-chip bus, respectively, and the slave node is configured with an address interval for routing a data packet to the slave node through the on-chip network. The method has the following technical features, which include the following steps: Step S1, the request node of the requester device uses the address interval to number a plurality of data packets according to their transaction processing order, and records the number in the address field of the data packet used for routing; Step S2, the request node of the requester device initiates data transmission and sends the plurality of numbered data packets to the receiver device; Step S3, when the slave node of the receiver device receives the plurality of data packets, it sequentially completes the transaction corresponding to each of the data packets based on the number of each of the data packets.
[0006] The data transmission method of the on-chip network bus device provided by the present invention may also have such a technical feature, wherein the number is an address in the address range, and the data packet includes a write request.
[0007] The data transmission method of the on-chip network bus device provided by the present invention may also have such a technical feature, wherein step S3 includes the following sub-steps: step S3-1, when the slave node of the receiver device receives the first data packet, it establishes a sequential relationship for the multiple data packets to be received, and completes the transaction corresponding to the first data packet; step S3-2, when the slave node of the receiver device receives the nth data packet, it checks and determines whether all data packets before the data packet have been received and the corresponding transactions have been completed according to the address field of the data packet, wherein n≥2; step S3-3, when step S3-2 is judged as no, the slave node of the receiver device blocks the transaction corresponding to the nth data packet; step S3-4, when step S3-2 is judged as yes, the slave node of the receiver device completes the transaction corresponding to the nth data packet.
[0008] The data transmission method of the on-chip network bus device provided by the present invention may also have such technical features, wherein step S3 also includes: step S3-5, after completing the transaction of the nth data packet, the receiver device checks and determines whether there are one or more data packets with consecutive numbers with the nth data packet among the several blocked data packets received before the nth data packet according to the address fields of each data packet; step S3-6, when the judgment in step S3-5 is yes, the receiver device completes the corresponding transactions in sequence according to the numbers of the blocked data packets.
[0009] The data transmission method of the on-chip network bus device provided by the present invention may also have such technical features, wherein, in step S2, the requesting node of the requester device declares the number of data packets to be transmitted and the number of the first data packet by writing the corresponding target node register before initiating the data transmission; in step S3-1, when the slave node of the receiver device receives the first data packet, it obtains the number and number range of the data packets to be received by reading the target node register.
[0010] The data transmission method of the on-chip network bus device provided by the present invention may also have such a technical feature, wherein step S3 includes the following sub-steps: step S3-1, when the slave node of the receiver device receives the first data packet, the slave node completes the transaction corresponding to the first data packet; step S3-2, when the slave node of the receiver device receives the nth data packet, it checks and determines whether all data packets before the data packet have been received and the corresponding transactions have been completed according to the address field of the data packet, wherein n≥2; step S3-3, when step S3-2 is judged as no, the slave node of the receiver device blocks the transaction corresponding to the nth data packet; step S3-4, when step S3-2 is judged as yes, the slave node of the receiver device completes the transaction corresponding to the nth data packet; step S3-7, the slave node of the receiver device determines whether more data packets can be received; step S3-8, when step S3-7 is judged as no, the slave node of the receiver device returns a corresponding response to the request node of the requester device, so that the requester device stops sending data packets.
[0011] The data transmission method of the on-chip network bus device provided by the present invention may also have such technical features, wherein step S3 also includes: step S3-5, after completing the transaction of the nth data packet, the receiver device checks and determines whether there are one or more data packets with consecutive numbers with the nth data packet among the several blocked data packets received before the nth data packet according to the address fields of each data packet; step S3-6, when the judgment in step S3-5 is yes, the receiver device completes the corresponding transactions in sequence according to the numbers of the blocked data packets.
[0012] The data transmission method of the on-chip network bus device provided by the present invention may also have such a technical feature, wherein, in step S3, after completing the transaction corresponding to the data packet, the receiver device returns a corresponding response to the requester device.
[0013] The method for transmitting sequence-preserving data of an on-chip network bus device provided by the present invention may also have such a technical feature, wherein the bus device is a PCIe device.
[0014] The present invention provides an on-chip system, characterized in that it includes: an on-chip bus; and at least two bus devices, which are respectively connected to the on-chip bus, and have corresponding request nodes and slave nodes on the on-chip bus, wherein any one of the at least two bus devices serves as a requester device, and the other serves as a receiver device, and the slave node of the receiver device is configured with an address interval for routing a data packet to the slave node through an on-chip network, and the request node of the requester device uses the address interval to number a plurality of data packets with order preservation requirements according to their transaction processing order, and records the number in the address field of the data packet used for routing; then initiates data transmission, and sends the plurality of numbered data packets to the receiver device, and when the slave node of the receiver device receives the plurality of data packets, it reorders the plurality of data packets based on the number of each of the data packets, and completes the transactions corresponding to each of the data packets in sequence according to the reordering results.
[0015] Function and Effect of the Invention According to the on-chip network bus device's order-preserving data transmission method and on-chip system of the present invention, the method includes the steps of a request node of a requester device numbering a plurality of data packets, a step of the request node initiating data transmission, and a step of a slave node of a receiver device reordering the plurality of numbered data packets after receiving them and completing corresponding transactions. Since the receiver device can handle the disorder of data packets with such a mechanism, the requester device does not need to send the next request after receiving the response, but can directly send a plurality of data packets, and the number of uncompleted transactions can be greater than 1, thereby improving the bandwidth of P2P data transmission between on-chip network bus devices; and since the request node uses the address interval reserved by the on-chip system for the target node to number the data packets and record them in the address field of the data packets, that is, the number can be used to route the data packets to the target node and allow the target node to identify the order of the plurality of data packets, and there is no need to add a new field for the number. Therefore, through such a clever design, the implementation of the method in the on-chip system is very convenient. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 It is a flow chart of a method for transmitting sequence-preserving data by an on-chip network bus device in Embodiment 1 of the present invention.
[0017] Figure 2 This is a flowchart of step S3 in the first embodiment of the present invention.
[0018] Figure 3 It is a schematic diagram of the interaction process between the requesting node and the target node in the first embodiment of the present invention.
[0019] Figure 4 This is a flow chart of step S3 in the second embodiment of the present invention.
[0020] Figure 5 It is a schematic diagram of the structure of the on-chip network in the prior art.
[0021] Figure 6 It is a schematic diagram of the interaction process between a requesting node and a target node in the prior art.
[0022] Reference numerals: On-chip bus 10; first PCIe device 21; requesting node 211; target node 212; second PCIe device 22; requesting node 221; target node 222; main memory 30; target node 31; other requester devices 40; requesting node 41. DETAILED DESCRIPTION
[0023] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the following specifically describes the method for transmitting sequence-preserving data in a network-on-chip bus device of the present invention in combination with embodiments and drawings.
[0024] <Example 1> The present embodiment provides an order-preserving data transmission method for an on-chip network bus device, which is used to implement strong order-preserving data transmission between multiple bus devices mounted on the on-chip network.
[0025] Figure 5 Schematic diagram of a network on chip in the prior art. The same network on chip structure is also used in this embodiment for specific description.
[0026] like Figure 5 As shown, the system on chip includes an on-chip bus 10, multiple PCIe devices (two of which are shown in the figure, a first PCIe device 21 and a second PCIe device 22), a main memory 30, and one or more other requester devices 40 (for example, processor cores). The PCIe devices, the main memory 30, and the other requester devices 40 are all connected to the on-chip bus 10 to form an interconnected on-chip network. Each mounted PCIe device has a corresponding request node and a target node on the on-chip bus 10. In addition, the main memory 30 also has a corresponding target node 31 on the on-chip bus 10, and the other requester devices 40 also have a corresponding request node 41 on the on-chip bus 10.
[0027] For the convenience of description, hereinafter, one of two PCIe devices performing P2P data transmission that initiates a transmission request is referred to as a requester device, and the other is referred to as a receiver device.
[0028] Figure 1 4 is a flow chart of the method for transmitting data in an order-preserving manner using the on-chip network bus device in this embodiment.
[0029] like Figure 1 As shown, the method comprises the following steps: Step S1, the requesting node of the requester device uses the address interval reserved by the system for the target node of the receiver device to number multiple data packets with order preservation requirements in the order of transmission (that is, the transaction processing order), and records the number in the address field of the data packet.
[0030] Step S2, the requesting node of the requester device initiates data transmission and sends a plurality of numbered data packets to the target node of the receiver device in sequence.
[0031] Step S3, when the target node of the receiver device receives the multiple data packets, it completes the transactions corresponding to the respective data packets in sequence based on the numbers of the respective data packets.
[0032] The above steps will be described in detail below.
[0033] Step S1, the requesting node of the requester device uses the address interval reserved by the on-chip system for the target node of the receiver device to number multiple data packets with order preservation requirements in the order of transmission, and records the numbers in the address field of the data packets.
[0034] Among them, the on-chip system will allocate (reserve) an address range for the target node of each bus device. If the address recorded in the address field in a data packet belongs to an address range, the on-chip network will route the data packet to the target node corresponding to the address range.
[0035] In this embodiment, the above-mentioned address range is also used as the sequential numbering of multiple data packets, that is, the encoding range of these data packets, so that the numbered data packets can be routed to the target node through their address fields, and the target node can identify the sequential relationship between these data packets, and there is no need to add new fields for the numbering.
[0036] In addition, in step S1, the requesting node combines the request packet and the data packet involved in the prior art method into one data packet, and sends it to the target node at one time.
[0037] by Figure 2 In the schematic structure shown, the first PCIe device 21 is used as a requester device, and the second PCIe device 22 is used as a receiver device. The system on chip allocates a reserved first address interval for the target node 212 of the first PCIe device 21, and allocates a reserved second address interval for the target node 222 of the PCIe device 22, and the first address interval has no overlapping part with the second address interval.
[0038] In this embodiment, the first address interval allocated to the target node 212 of the first PCIe device 21 is 0x0000-0x00ff, and the second address interval allocated to the target node 222 of the PCIe device 22 is 0x1000-0x10ff.
[0039] The first PCIe device 21 wants to initiate a P2P transmission to the second PCIe device 22. The transmission involves four data packets and requires order preservation. The requesting node 211 of the first PCIe device 21 uses the four addresses ADDR1 to ADDR4 in the second address interval as the numbers of the four data packets according to the order of the four data packets and the second address interval of the target node 222 of the second PCIe device 22, and records the four numbers in the address field of each data packet.
[0040] In this embodiment, the address field of the first data packet is set to 0x1000 (ADDR1), the address field of the second data packet is set to 0x1001 (ADDR2), the address field of the third data packet is set to 0x1002 (ADDR3), and the address field of the fourth data packet is set to 0x1003 (ADDR4).
[0041] Step S2, the requesting node of the requester device initiates data transmission and sends a plurality of numbered data packets to the target node of the receiver device in sequence.
[0042] In this embodiment, the first PCIe device 21 sends the above four data packets in sequence to the second PCIe device 22. Since the four addresses 0x1000-0x1003 in the address field of the above data packets all belong to the second address interval reserved by the on-chip system for the target node 222 of the second PCIe device 22, the on-chip network will route all the four data packets to the target node 222 of the second PCIe device 22.
[0043] That is, the requesting node of the requester device does not need to be like Figure 5 The interactive process shown is the same as that of first sending a write request, sending a write data packet after receiving a response, and then sending the next write request after completion. Instead, multiple write data packets containing write requests can be directly sent in sequence.
[0044] Step S3, when the target node of the receiver device receives the multiple data packets, it completes the transactions corresponding to the data packets in sequence based on the numbers of the data packets.
[0045] The order in which the four data packets arrive at the target node 222 of the second PCIe device 22 may be out of order. The target node 222 of the second PCIe device 22 may reorder the data packets according to their address fields, and process the corresponding transactions of the data packets in sequence according to the reordering structure. After completing the transaction, the target node of the receiver device returns a corresponding response to the request node of the requester device.
[0046] Figure 3 is a flow chart of step S3 in this embodiment.
[0047] like Figure 3 As shown, step S3 specifically includes the following sub-steps: Step S3-1, when the target node of the receiver device receives the first data packet, it establishes a sequence relationship for multiple data packets to be received and completes the transaction corresponding to the data packet.
[0048] In this embodiment, in step S2, the requesting node of the requester device declares the number of data packets to be transmitted and the starting address number of the first data packet by writing the corresponding target node register before initiating P2P data transmission. When the target node of the receiver device receives the first data packet, it can learn the number of data packets to be received and the address number range by reading the above-mentioned target node register. Regarding the operation of establishing a sequential relationship for multiple data packets, the receiver device can also adopt other feasible methods in the prior art, which is not limited in this application.
[0049] Step S3-2, when the target node of the receiver device receives the nth (n≥2) data packet, it checks and determines according to the address field of the data packet whether the 1st to n-1th data packets before the data packet have all been received and the corresponding transactions have all been processed.
[0050] Step S3-3, when the judgment in step S3-2 is no, the target node of the receiver device blocks the transaction corresponding to the nth data packet and returns to step S3-2, that is, waits for the next data packet.
[0051] Step S3-4, when the judgment in step S3-2 is yes, the target node of the receiver device completes the transaction corresponding to the nth data packet.
[0052] Step S3-5, after completing the transaction of the nth data packet, the target node of the receiver device checks and determines whether there are one or more data packets with consecutive numbers with the nth data packet among the data packets with blocked transactions received before the nth data packet according to the address fields of each data packet. Step S3-6, when the judgment in step S3-5 is yes, the receiver device completes the transactions corresponding to the blocked data packets in sequence according to the numbers of the blocked data packets, and then returns to step S3-2, that is, waits for the next data packet.
[0053] That is, determine whether the n+1th data packet has arrived before the nth data packet, or whether consecutive n+1th data packets, n+2th data packets, etc. have arrived before. After completing the transactions of n data packets, the transactions of these data packets can be completed in sequence.
[0054] In step S3, the received multiple data packets that may be out of order are reordered according to the above process and the corresponding transactions are completed in sequence until all the data packets are received and the transaction processing is completed.
[0055] In the on-chip system, the functions of dependency checking and back pressure can be implemented through circuits. That is, if a transaction with an earlier order has not been completed and a transaction with a later order is received, the circuit can recognize that the dependency between the two transactions has not been eliminated, and thus the transaction with the later order will not be executed until the dependency is eliminated.
[0056] Figure 4 This is an example diagram of a receiver device reordering multiple received data packets in this embodiment.
[0057] like Figure 4 As shown, for example, the target node 222 of the second PCIe device 22 (receiver device) receives four data packets in succession, and these four data packets are out of order. The target node 222 receives these four data packets in the order of: data packet 1, data packet 3, data packet 2, and data packet 4, that is, the third data packet 3 is received before the second data packet 2. Figure 4 As shown in Table t1 in , when data packet 1 is received, according to the above step S3-1, the receiver device 22 establishes a sequential relationship for data packet 1 and the next three data packets to be received, completes the transaction corresponding to data packet 1, records the numbers of the four data packets and their sequential relationship, and records the status that data packet 1 has been received and the corresponding transaction has been completed.
[0058] like Figure 4 As shown in Table t2 in FIG, upon receiving data packet 3, according to the above step S3-2, the receiver device 22 checks whether data packets 1 and 2 before data packet 3 have been received and the corresponding transactions have been completed. Since it is checked that data packet 2 has not been received and the corresponding transaction has not been completed, according to step S3-3, the receiver device 22 blocks the transaction corresponding to data packet 3 and then waits for the next data packet.
[0059] like Figure 4 As shown in Table t3 in , when receiving data packet 2, according to the above step S3-2, the receiver device 22 checks whether data packet 1 before data packet 2 has been received and the corresponding transaction has been completed. Since data packet 1 has been received and the corresponding transaction has been completed, according to step S3-4, the receiver device 22 completes the transaction corresponding to data packet 2; then, according to step S3-5, the receiver device 22 checks whether there are one or more data packets with consecutive numbers with data packet 2 among the blocked data packets received before data packet 2, and checks to obtain the blocked data packet 3; then, according to step S3-6, the receiver device 22 completes the transaction corresponding to the blocked data packet 3; and then waits for the next data packet.
[0060] like Figure 4As shown in Table t4 in FIG, upon receiving data packet 4, according to the above step S3-2, the receiver device 22 checks whether data packets 1-3 before data packet 4 have all been received and the corresponding transactions have all been completed. Since data packets 1-3 have all been received and the corresponding transactions have all been completed, according to step S3-4, the receiver device 22 completes the transaction corresponding to data packet 4.
[0061] It can be seen that between packets 1 and 4, outstanding = 2. It can be understood that when more packets are transmitted, outstanding may also be other values greater than 2.
[0062] That is, through the above steps, strong order-preserving P2P data transmission between two bus devices can be achieved, and the transmission bandwidth is larger.
[0063] Functions and Effects of Embodiment 1 According to the on-chip network bus device's order-preserving data transmission method and on-chip system provided in this embodiment, the method includes the steps of a requesting node of a requester device numbering a plurality of data packets, a step of the requesting node initiating data transmission, and a step of a slave node of a receiver device reordering the plurality of numbered data packets after receiving them and completing corresponding transactions. Since the receiver device can handle the disorder of data packets with such a mechanism, the requester device does not need to send the next request after receiving the response, but can directly send a plurality of data packets, and the number of uncompleted transactions can be greater than 1, thereby improving the bandwidth of P2P data transmission between on-chip network bus devices; and since the requesting node uses the address interval reserved by the on-chip system for the target node to number the data packets and record them in the address field of the data packets, that is, the number can be used to route the data packets to the target node and allow the target node to identify the order of the plurality of data packets, and there is no need to add a new field for the number. Therefore, through such a clever design, the implementation of the method in the on-chip system is very convenient.
[0064] In an embodiment, when a slave node of a receiver device receives a data packet, it can check and determine, based on the number of the data packet, whether all previous data packets have been received and whether the corresponding transactions have been completed, and block the transaction corresponding to the data packet if the determination is no, thereby achieving reordering of multiple data packets at the receiving end so that the number of unfinished transactions can be greater than 1.
[0065] Furthermore, the data packet sent by the requesting node of the requester device includes the original write request, so the method not only improves the transmission bandwidth but also simplifies the interaction process between the requesting node and the target node.
[0066] <Example 2> The present embodiment provides an order-preserving data transmission method for an on-chip network bus device. In the present embodiment, the same symbols are assigned to the same components as those in the first embodiment and the corresponding descriptions are omitted.
[0067] Compared with the first embodiment, the difference is that in the method of this embodiment, the specific process of step S3 is different.
[0068] Figure 4 is a flow chart of step S3 in this embodiment.
[0069] like Figure 4 As shown, step S3' of this embodiment includes the following sub-steps: Step S3-1', when the target node of the receiver device receives the first data packet, it completes the transaction corresponding to the data packet.
[0070] Step S3-2', when the target node of the receiver device receives the nth data packet, it checks and determines according to the address field of the data packet whether the 1st to n-1th data packets before the data packet have been received and the corresponding transactions have been processed.
[0071] Step S3-3', when the judgment in step S3-2 is no, the target node of the receiver device blocks the transaction corresponding to the nth data packet.
[0072] Step S3-4', when the judgment in step S3-2 is yes, the target node of the receiver device completes the transaction corresponding to the nth data packet.
[0073] Step S3-5', after completing the transaction of the nth data packet, the slave node of the receiver device detects and determines whether there are one or more data packets with consecutive numbers with the nth data packet among several data packets with blocked transactions received before the nth data packet.
[0074] Step S3-6', when the judgment in step S3-5' is yes, the target node of the receiver device completes the transactions corresponding to the blocked data packets in sequence according to the numbers of the blocked data packets.
[0075] Step S3-7', the target node of the receiver device determines whether more data packets can be received, and if the determination is yes, returns to step S3-2', ie, receives the next data packet.
[0076] Step S3-8', when the judgment in step S3-7' is no, that is, when no more data packets can be received, the target node of the receiver device returns a corresponding response to the request node of the requester device, so that the request node of the requester device temporarily stops sending data packets.
[0077] That is, compared with the first embodiment, the present embodiment does not utilize the target node register for declaration, but the requester device directly initiates transmission to the receiver device, and the receiver device receives the data packet when it can, and returns an appropriate response to the requester device when it cannot receive any further data packets, so that the requester device stops sending.
[0078] In this embodiment, other methods are the same as those in the first embodiment, and thus will not be described again.
[0079] Functions and Effects of Embodiment 2 According to the order-preserving data transmission method of the on-chip network bus device provided by this embodiment, since the same order-preserving transmission mechanism as that of the first embodiment is adopted, P2P data transmission with strong order preservation and higher transmission bandwidth can also be achieved. In addition, since the transmission of data packets is controlled by the receiving device returning a response when it cannot receive more data packets, there is no need to declare the number of transmitted data packets by writing to the corresponding registers, etc., so less resources are occupied, and adaptive data packet transmission adjustment can be achieved according to the performance and processing conditions of the receiving device.
[0080] The above embodiments are only used to illustrate the specific implementation of the present invention, and the present invention is not limited to the description scope of the above embodiments. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only to illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, and these changes and improvements fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.
[0081] For example, in the above embodiment, a PCIe device is taken as an example for specific description. In fact, the method can also be applied to bus devices of other bus protocols in a similar manner.
Claims
1. A data transmission method for an on-chip network bus device, for performing data transmission between two bus devices connected to an on-chip bus, wherein the two bus devices are a requester device and a receiver device, respectively, and the two bus devices have a request node and a slave node on the on-chip bus, respectively, and the slave node is configured with an address interval for routing a data packet to the slave node through the on-chip network, characterized in that: The following steps are involved: Step S1, the request node of the requester device uses the address interval to number multiple data packets according to their transaction processing order, and records the number in the address field of the data packet for routing; Step S2, the requesting node of the requester device initiates data transmission and sends the plurality of numbered data packets to the receiver device; Step S3: When receiving the plurality of data packets, the slave node of the receiver device sequentially completes the transactions corresponding to the respective data packets based on the serial numbers of the respective data packets.
2. The data transmission method of the on-chip network bus device according to claim 1, characterized in that: in, The number is an address in the address range, The data packet contains a write request.
3. The data transmission method of the on-chip network bus device according to claim 1 or 2, Features: Wherein, step S3 includes the following sub-steps: Step S3-1, when receiving the first data packet, the slave node of the receiver device establishes a sequence relationship for multiple data packets to be received and completes the transaction corresponding to the first data packet; Step S3-2, when the slave node of the receiver device receives the nth data packet, it checks and determines whether all data packets before the data packet have been received and the corresponding transactions have been completed according to the address field of the data packet, where n≥2; Step S3-3, when the judgment in step S3-2 is no, the slave node of the receiver device blocks the transaction corresponding to the nth data packet; Step S3-4, when the judgment in step S3-2 is yes, the slave node of the receiver device completes the transaction corresponding to the nth data packet.
4. The data transmission method of the on-chip network bus device according to claim 3, characterized in that: in, Step S3 also includes: Step S3-5, after completing the transaction of the nth data packet, the receiver device checks and determines, according to the address fields of each data packet, whether there are one or more data packets with consecutive numbers with the nth data packet among the blocked data packets received before the nth data packet; Step S3-6, when the judgment in step S3-5 is yes, the receiver device completes the corresponding transactions in sequence according to the numbers of the blocked data packets.
5. The data transmission method of the on-chip network bus device according to claim 3, characterized in that: in, In step S2, before initiating the data transmission, the requesting node of the requester device declares the number of data packets to be transmitted and the number of the first data packet by writing the corresponding target node register. In step S3-1, when the slave node of the receiver device receives the first data packet, it obtains the number and number range of the data packets to be received by reading the register of the target node.
6. The data transmission method of the on-chip network bus device according to claim 1 or 2, Features: Wherein, step S3 includes the following sub-steps: Step S3-1, when the slave node of the receiver device receives the first data packet, the slave node completes the transaction corresponding to the first data packet; Step S3-2, when the slave node of the receiver device receives the nth data packet, it checks and determines whether all data packets before the data packet have been received and the corresponding transactions have been completed according to the address field of the data packet, where n≥2; Step S3-3, when the judgment in step S3-2 is no, the slave node of the receiver device blocks the transaction corresponding to the nth data packet; Step S3-4, when the judgment in step S3-2 is yes, the slave node of the receiver device completes the transaction corresponding to the nth data packet; Step S3-7, the slave node of the receiver device determines whether more data packets can be received; Step S3-8, when the judgment in step S3-7 is no, the slave node of the receiver device returns a corresponding response to the request node of the requester device, so that the requester device stops sending data packets.
7. The data transmission method of the on-chip network bus device according to claim 6, characterized in that: in, Step S3 also includes: Step S3-5, after completing the transaction of the nth data packet, the receiver device checks and determines, according to the address fields of each data packet, whether there are one or more data packets with consecutive numbers with the nth data packet among the blocked data packets received before the nth data packet; Step S3-6, when the judgment in step S3-5 is yes, the receiver device completes the corresponding transactions in sequence according to the numbers of the blocked data packets.
8. The data transmission method of the on-chip network bus device according to claim 1 or 2, characterized in that: in, In step S3, after completing the transaction corresponding to the data packet, the receiver device returns a corresponding response to the requester device.
9. The data transmission method of the on-chip network bus device according to claim 1 or 2, characterized in that: in, The bus device is a PCIe device.
10. A system on chip, characterized in that: include: On-chip bus; as well as at least two bus devices, respectively connected to the on-chip bus, with corresponding request nodes and slave nodes on the on-chip bus, Wherein, any one of the at least two bus devices serves as a requester device, and the other serves as a receiver device, The slave node of the receiver's device is configured with an address interval for routing the data packet to the slave node through the on-chip network, The request node of the requester device uses the address interval to number multiple data packets with order preservation requirements according to their transaction processing order, and records the number in the address field of the data packet for routing; then initiates data transmission and sends the multiple numbered data packets to the receiver device, When receiving the multiple data packets, the slave node of the receiver device sequentially completes the transactions corresponding to the data packets based on the numbers of the data packets.
Citation Information
Patent Citations
Directory cache method for node control chip in cache coherent non-uniform memory access (CC-NUMA) system
CN102708190A
Bus master, bus system, and bus control method
CN105190583A
Method for processing data by using network card, network equipment and computer system
CN114489465A
Data transmission method of multi-core processor on-chip network system
CN116303221A
Method and system for reordering out-of-order response based on AXI protocol
CN116414767A
Cited By
On-chip transaction processing system and method, electronic equipment and storage medium
CN120045514A
