CN112486871B - Routing method and system for on-chip bus - Google Patents

Abstract

An embodiment of the present application provides a routing method and system for an on-chip bus. The routing method includes: acquiring the actual transmission bandwidth occupancy of each destination port in at least one destination port on the bus; Predicted transmission bandwidth occupancy of the destination port in the future estimated time period; according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy, a transmission strategy is determined for the data to be transmitted from the source port. Some embodiments of the present application monitor the actual bandwidth occupancy and predicted bandwidth occupancy of each destination port in the bus transmission process, and then provide a suitable target exit for the data to be transmitted. When the data transmission direction is uplink, the data to be transmitted Send it to the host (or on-chip storage unit) connected to the target port through the target outlet.

Description

A routing method and system for on-chip bus

technical field

The present application relates to the field of computer IO, and specifically, the embodiments of the present application relate to a routing method and system for an on-chip bus.

Background technique

Computer I/O technology is always a very important key technology in the development of high performance computing technology. Its technical characteristics determine the processing capability of computer I/O, and then determine the overall performance and application environment of the computer. Fundamentally speaking, no matter now or in the future, I/O technology will restrict the application and development of computer technology, especially in the field of high-end computing.

Computer I/O peripherals mainly include devices connected to PCIE, USB, SATA, and Ethernet. In server applications, access to memory by multiple I/O devices and P2P (Peer to Peer) access between multiple I/O devices are involved. The access of these I/O devices needs to meet the bandwidth requirements, and also requires unified and flexible routing management. In a system-on-chip or system-on-a-chip SoC chip, the routing management is implemented by a dedicated IO routing module, and the IO routing module is collectively referred to as IOHUB. IOHUB connects various I/O peripherals and CPU memory, and also routes local to remote I/O devices and memory access. As the server's demand for I/O devices increases, the server SoC chip needs to support more I/O devices to meet greater bandwidth requirements.

Therefore, how to meet the simultaneous access of more peripherals to CPU memory and improve the routing capability of IOHUB is a technical problem that needs to be solved urgently.

Contents of the invention

The purpose of the embodiments of the present application is to provide a routing method and system for an on-chip bus. Through the technical solutions of the embodiments of the present application, at least the bandwidth and transmission efficiency of the IOHUB can be effectively improved. For example, in some embodiments of the present application, by increasing the number of target ports, more source ports can be supported, and the transmission efficiency can be improved by improving the bandwidth monitoring mechanism and flexible routing.

In a first aspect, some embodiments of the present application provide a routing method for an on-chip bus. The routing method includes: obtaining the actual transmission bandwidth occupancy of each destination port in at least one destination port on the bus; obtaining the at least one The predicted transmission bandwidth occupancy of each of the destination ports in the expected future time period; according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy, a transmission strategy is determined for the data to be transmitted from the source port.

Some embodiments of the present application monitor the actual bandwidth occupancy and predicted bandwidth occupancy of each destination port in the bus transmission process, and then provide a suitable target exit for the data to be transmitted. When the data transmission direction is uplink, the data to be transmitted Send it to the host (or on-chip storage unit) connected to the target port through the target outlet.

In some embodiments, the destination port is a port connected to a host or an on-chip memory, and there are multiple destination ports; according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy, it is Determining a transmission strategy for the data to be transmitted from the source port includes: determining at least one target exit from a plurality of destination ports for the data to be transmitted according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy ; Wherein, the target exit is used to provide the data to be transmitted to the host or the on-chip memory.

Some embodiments of the present application can provide more destination ports for data from more I/O devices by increasing the number of bus interfaces connected to the host or internal memory. By setting multiple destination ports, it can meet the increase of I/O devices, or the increase of I/O interface speed and the increase of bandwidth requirements caused by system applications, and improve the IO routing by using a connection with the host or on-chip memory. The bandwidth bottleneck problem caused by the destination port.

In some embodiments, the determining at least one target exit from multiple destination ports for the data to be transmitted according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy includes: according to the determining at least one target egress from the plurality of destination ports according to the transmission bandwidth threshold and the predicted bandwidth threshold respectively set for each destination port in the plurality of destination ports.

Some embodiments of the present application can determine whether these destination ports can transmit data from Data to be transmitted on the source port.

In some embodiments, the determining at least one target egress from the plurality of destination ports according to the transmission bandwidth threshold and the predicted bandwidth threshold respectively set for each of the plurality of destination ports includes: confirming the first The first actual transmission bandwidth occupancy corresponding to a destination port is smaller than the first transmission bandwidth threshold corresponding to the first destination port, and confirming that the first predicted transmission bandwidth occupancy corresponding to the first destination port is smaller than the first transmission bandwidth threshold corresponding to the first destination port A first predicted bandwidth threshold corresponding to a destination port; selecting the first destination port as the target egress.

In some embodiments of the present application, a destination port with sufficient bandwidth margin at the current moment and a larger bandwidth margin within a predictable period of time from the current moment is selected as the target egress of the current data to be transmitted, so that the The transmission data is output to the host or memory connected to the destination port through enough destination ports, which ensures the smooth transmission of the data to be transferred and improves the resource utilization of the destination port.

In some embodiments, the determining at least one target egress from multiple destination ports according to the transmission bandwidth threshold and predicted bandwidth threshold respectively set for each of the multiple destination ports includes: according to the Describe the attribute characteristics of the data to be transmitted, and confirm that at least one of the transmission bandwidth threshold and the predicted bandwidth threshold is used to determine the target exit, wherein the attribute characteristics are used to characterize the amount of data to be transmitted .

Some embodiments of the present application also select a more reasonable target exit according to the amount of data corresponding to the data to be transmitted, and make full use of the bandwidth of each destination port on the basis of ensuring the smooth transmission of the data to be transmitted.

In some embodiments, according to the attribute characteristics of the data to be transmitted, confirming that at least one of the transmission bandwidth threshold and the predicted bandwidth threshold is used to determine the target egress includes: confirming the data volume of the data to be transmitted greater than the first set threshold; confirming that the second predicted transmission bandwidth occupancy corresponding to the second destination port is less than the second predicted bandwidth threshold corresponding to the second destination port; selecting the second destination port as the target egress; Or, confirming that the amount of data to be transmitted is less than a second set threshold; confirming that the second actual transmission bandwidth occupation corresponding to the second destination port is less than the second transmission bandwidth threshold corresponding to the second destination port; Selecting the second destination port as the target egress; wherein the first set threshold is greater than the second set threshold.

In some embodiments of the present application, when the amount of data to be transmitted is large, the destination port with sufficient predicted bandwidth margin (that is, whether to select a certain destination port as the target exit according to the predicted actual transmission bandwidth occupancy) is preferred. When the amount is small, the current remaining bandwidth can be preferentially selected (that is, to determine whether to select a certain destination port as the target exit according to the actual transmission bandwidth occupancy). These embodiments fully combine the characteristics of the data to be transmitted and the remaining bandwidth of each destination port , which can more accurately determine the target outlet connected to the host or memory for the data to be transmitted.

In some embodiments, the actual transmission bandwidth occupancy is determined according to the transmission bandwidth of the current transmission within the transmission window.

Some embodiments of the present application determine the actual transmission bandwidth occupancy of each destination port by counting the transmission bandwidth within the transmission window, which improves the accuracy and objectivity of estimating the bandwidth occupancy of each destination port.

In some embodiments, the calculation formula of the actual transmission bandwidth occupancy is as follows:

Among them, CLK COUNT is the number of system clock cycles counted within the set transmission window, and TRANS COUNT is used to represent and count the number of data bytes corresponding to all valid transmissions at the current moment.

Some embodiments of the present application can determine the actual transmission bandwidth occupancy of each destination port through the above formula, which improves the objectivity of bandwidth occupancy estimation.

In some embodiments, the value of the parameter TRANS COUNT is counted by monitoring the transmission effective signal in the channel.

Some embodiments of the present application determine and record the actual transmission bandwidth occupancy by detecting valid signals of data transmission of each channel on the bus, thereby improving the accuracy of estimating the actual transmission bandwidth occupancy.

In some embodiments, the predicted transmission bandwidth occupancy is determined according to the transmission bandwidth of the expected transmission within the expected transmission window, wherein the expected transmission window is determined according to the expected transmission corresponding to The total number of system clock cycles determined by the transfer length and transfer bit width.

Some embodiments of the present application determine the predicted transmission bandwidth occupancy of each destination port by counting the expected transmission bandwidth of each destination port within the expected transmission window, which improves the accuracy and objectivity of estimating the bandwidth occupancy of each destination port.

In some embodiments, the calculation formula of the predicted transmission bandwidth occupancy is as follows:

Wherein, CLK COUNT1 is a system clock count value within the predictable transmission window, and EST is a count value obtained by counting the transmission length carried by the current transmission request to represent the data.

Some embodiments of the present application can determine the predicted transmission bandwidth occupancy of each destination port through the above formula, which improves the objectivity of bandwidth occupancy estimation.

In some embodiments, the value of the EST is obtained by extracting the information carried in the transmission request signal in the channel.

In some embodiments of the present application, the value of the predicted transmission bandwidth occupancy is determined by detecting the information carried in the transmission request signal of each channel on the bus, which improves the accuracy of estimation of the predicted transmission bandwidth occupancy.

In a second aspect, some embodiments of the present application provide a routing device for an on-chip bus. The routing device includes: a bandwidth occupancy calculation module configured to: obtain the bandwidth occupancy calculation module of each destination port in the current at least one destination port on the bus The actual transmission bandwidth occupancy; obtaining the predicted transmission bandwidth occupancy of each destination port in the at least one destination port in a preset time period in the future; the arbitration module is configured to transmit according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy Bandwidth occupancy, which determines the transmission strategy for the data to be transmitted from the source port.

In a third aspect, some embodiments of the present application provide a routing system for a system on chip, the routing system includes: a system clock counting module configured to add 1 for each transmission clock cycle when monitoring starts, and when the When the transmission window value is set, it is cleared and restarts the clock counting in the next monitoring transmission window; the actual transmission volume counting module is configured to count the actual transmission volume corresponding to the effective transmission on the channel according to the effective transmission signal; predict the transmission volume The counting module is configured to count the predicted transmission volume on the channel in the expected time period in the future according to the transmission request signal; the bandwidth calculation module is configured to: according to the set transmission window size and a plurality of actual The transmission amount determines the actual transmission bandwidth occupancy of the destination port; determines the predicted transmission bandwidth occupancy according to the predictable transmission window size and a plurality of the predicted transmission amounts corresponding to the destination port; the arbiter is configured to: compare Obtaining a first comparison result between the actual transmission bandwidth occupancy of the destination port and the set transmission bandwidth threshold; comparing the predicted transmission bandwidth occupancy of the destination port with the set prediction transmission bandwidth threshold to obtain a second comparison result; The first comparison result and the second comparison result generate a target exit selection signal, wherein the target exit is an exit selected from a plurality of destination ports for transmitting data to be transmitted; the router is configured as receiving the target exit selection signal to send the data to be transmitted to the target exit.

Some embodiments of the present application count the actual transmission volume, the predicted transmission volume, and the number of system clocks through multiple technical devices, and then determine the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy based on these data, which improves the accuracy of bandwidth occupancy estimation sex.

In some embodiments, when it is confirmed that the transmission valid signal is at a first level, the actual transmission amount counting module is started to perform one count, wherein the first level is a high level or a low level.

In some embodiments of the present application, by transmitting a valid signal to trigger the statistics of the actual transmission volume, the accuracy of the statistical result of the actual transmission volume can be improved.

In some embodiments, the counting step of one count is determined according to the transmission bit width of the channel.

In some embodiments of the present application, by setting a uniform step size value, the comparability of bandwidth comparison results of multiple destination ports is realized, thereby improving the rationality of the obtained transmission strategy.

In some embodiments, when it is confirmed that the transmission request signal is at a second level, the predicted transmission amount counting module is started to perform one count, wherein the second level is a high level or a low level.

In some embodiments of the present application, the transmission request signal is used to trigger the statistics of the predicted transmission volume, which can improve the accuracy of the statistics result of the predicted transmission volume.

In some embodiments, the counting step of one count is determined according to the transmission length of the data to be transmitted this time carried by the transmission request signal.

In some embodiments of the present application, by setting a uniform step size value, the comparability of bandwidth comparison results of multiple destination ports is realized, thereby improving the rationality of the obtained transmission strategy.

In a fourth aspect, some embodiments of the present application provide a system-on-chip data transmission method, the data transmission method including: before an effective transmission starts: setting the value of the transmission window corresponding to each destination port among the multiple destination ports, And set the bandwidth comparison threshold corresponding to each destination port in the plurality of destination ports, wherein the bandwidth comparison threshold includes a transmission bandwidth comparison threshold and a predicted transmission bandwidth threshold; when effective transmission starts, continuously update the plurality of destination ports The actual transmission bandwidth occupancy of each destination port and the value of the predicted transmission bandwidth occupancy, wherein the actual transmission bandwidth occupancy is calculated according to the value of the transmission window and the statistical actual transmission volume, and the predicted transmission bandwidth The bandwidth occupancy is calculated according to the value of the predicted transmission window and the predicted transmission volume obtained by statistics; during arbitration, the output routing result is dynamically adjusted according to the bandwidth comparison threshold of each destination port in the plurality of destination ports, so as to Data to be transmitted from the source port is routed to the adjusted destination egress.

In the fifth aspect, some embodiments of the present application provide a system on chip, the system on chip includes a processor, a memory, at least one I/O device, and the routing system described in any one of the above third aspects; wherein, the At least one I/O device is interconnected with the processor through the routing system; or, the at least one I/O device is interconnected with the memory through the routing system.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the accompanying drawings that need to be used in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present application, so It should not be regarded as a limitation on the scope, and those skilled in the art can also obtain other related drawings according to these drawings without creative work.

Fig. 1 is a relational diagram of IO routing provided by related technologies and multiple I/O devices and host connections;

FIG. 2 is an internal structure diagram of IO routing provided by related technologies;

FIG. 3 is a schematic diagram provided by an embodiment of the present application to explain the blockage caused by the small number of destination ports corresponding to uplink data transmission in the related art;

FIG. 4 is a flowchart of a routing method for an on-chip bus provided in an embodiment of the present application;

FIG. 5 is a schematic diagram of transmission window movement provided by an embodiment of the present application;

Fig. 6 is the relationship diagram of the IO routing of the multi-bus interface provided by the embodiment of the present application and the connection between multiple I/O devices and the host;

7 is a schematic diagram of the connection relationship between the source port and the destination port provided by the embodiment of the present application;

FIG. 8 is a schematic diagram of routes corresponding to two destination ports for uplink data transmission provided by the embodiment of the present application;

FIG. 9 is a block diagram of a routing device for an on-chip bus provided in an embodiment of the present application;

FIG. 10 is a block diagram of a routing system for an on-chip bus provided in an embodiment of the present application;

FIG. 11 is an internal structural diagram of the IO routing provided by the embodiment of the present application;

FIG. 12 is another flowchart of the routing method provided by the embodiment of the present application.

Detailed ways

The technical solutions in the embodiments of the present application will be described below with reference to the drawings in the embodiments of the present application.

It should be noted that like numerals and letters denote similar items in the following figures, therefore, once an item is defined in one figure, it does not require further definition and explanation in subsequent figures. Meanwhile, in the description of the present application, the terms "first", "second" and the like are only used to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

The meanings of the abbreviations involved in this application are as follows.

SoC: System-on-a-Chip, SoC is called a system-on-a-chip, which is an integrated circuit with a dedicated target, which contains a complete system and has all the content of embedded software.

P2P: Peer-to-Peer, PCIE device end-to-end transmission.

Host: I/O device transmission and interconnection objects, such as Data Fabric.

Monitor: A functional module for monitoring transmission bandwidth.

Source port: The port that initiates the transfer request in the client port of the IO route.

Target port: The port that receives the transfer request and outputs it in the client port of the IO route.

IOHUBs of related technologies and the technical problems and defects existing in these IOHUBs are exemplarily described below with reference to FIG. 1 and FIG. 2 .

The IOHUB design of the related art interconnects the I/O device with the host Host through the bus interface Bus Interface. Taking five I/O devices interconnected with the host Host through IOHUB as an example, the interconnection structure is shown in FIG. 1 . The upstream data stream in Fig. 1 is that the I/O device 300 sends data to the host 100. The specific process is that five I/O devices are connected to the IOHUB (i.e. the IO router 200 in Fig. 1 ) through the bus interface BUSIF (Bus Interface). The router 200 connects the request and data from the I/O device to the host 100 through the bus interface BUSIF connected to the host through routing; the downstream data flow downstream in Figure 1 sends data from the host 100 to the I/O device 300, the specific process Access to five I/O devices is realized after the request and data from the host 100 are routed and distributed to the bus interface BUSIF connected to the I/O device. In FIG. 1 , five I/O devices share one host 100 , and P2P access can also be implemented between I/O devices through routing. For IOHUB, each port is a client port, and each client port can be either a source port or a destination port target port, which depends on the direction of data transmission.

The design scheme of the IOHUB in the related art is shown in FIG. 2 , which is different from FIG. 1 in that FIG. 2 replaces the host 100 in FIG. 1 with a memory 110 . The memory 110 in Figure 2 can receive and store data from the I/O device through the bus, and can also send data to the I/O device connected to the bus, so Figure 2 illustrates two bus interface units connected to the memory 110, And these two bus interface units respectively correspond to the upstream interface (corresponding to the destination port) of the upstream data flow and the downstream interface (corresponding to the source port) of the downstream data flow respectively. In some embodiments of the application, the two A bus interface in one direction is represented as a bus interface as shown in FIG. 1 . It should be noted that, in some embodiments of the present application, by increasing the number of bus interfaces connected to the host or on-chip memory, it may refer to increasing the number of bus interfaces serving as uplink interfaces.

Fig. 2 shows the main functional modules that the IOHUB of related art comprises, and these functional modules include: bus interface BUSIF, decoder 201 (or be called decoder), bus buffer 202 (or be called BUF), multiplexer 203 (or called MUX) and arbitrator 204 (or called arbitrition), wherein the core functions of each module are introduced as follows.

The bus interface BUSIF belongs to the bus interface of the I/O device and the memory on the bus, and is used for the I/O device and the memory to access the IOHUB. The bus interface realizes the bus protocol, and processes the command channel, data channel and response channel.

Decoder 201, this module mainly realizes the decoding of the access request address. For the upstream direction of the upstream data flow, the decoder 201 decodes according to the range of the transmission request address request address to determine whether the target device of this transmission is a memory or other I/O devices, or an invalid address space. If the access space is a storage memory or a remote I/O device, the destination port target port is the port connected to the host or storage; if the access space is an I/O device on the same routing Hub, the destination port target port is the same as The port to which other I/O devices are connected; if the access space is invalid, an invalid response will be returned. For the downstream direction of the downstream data flow, the Host port of the host computer is used as the source port sourceport, and its decoder 201 decodes according to the address space of each I/O device configured by the software. target port is the port of the corresponding I/O device.

The bus buffer 202 (for example, FIFO including different channels of the bus) is used for temporarily storing data or temporarily storing access requests. For example, before each access request and data are arbitrated and output, the FIFO buffers these access requests and data to ensure that the access is not interrupted, wherein the depth and bandwidth of the FIFO are related to the number of client ports.

The arbiter 204 is mainly responsible for arbitrating when different source ports access the same target port, and only one source port is routed to the target port at the same time. Arbitration adopts round-robin (Round Robin) scheduling method, that is, each time the request from the client port client port is allocated to the destination port target port in turn.

The multiplexer 203 is responsible for selecting the source port source port to output to the destination port target port according to the results of the decoder 201 and the arbiter 204 .

It should be noted that the upstream data flow direction in Figure 2 has only one destination port (corresponding to two data flow directions), so when the number of I/O devices increases, the bandwidth of this destination port may become a constraint between the I/O device and Bottleneck for memory 110 communication.

The technical defects existing in the related technologies of FIG. 1 and FIG. 2 are exemplarily described below with reference to FIG. 3 .

In related technical solutions (for example, FIG. 1 and FIG. 2 ), multiple I/O devices can only access the host or memory through one target port, and the arbitration module on the IO router adopts a polling method for arbitration. Therefore, with the increase of I/O devices and the increase of bandwidth requirements, there are the following problems: First, the bandwidth is not enough: with the increase of I/O devices, the client port corresponding to the I/O devices mounted on IOHUB will be increase; with the improvement of the interface rate corresponding to the I/O device and the increase of the bandwidth demand of the system application, a target port becomes a bandwidth bottleneck. As shown in Fig. 3, the bandwidth of path ① and path ② are both 32GB/s. Since the bus interface BUSIF connected to the host 100 has only one outlet, the bandwidth of 32GB/s of the bus interface BUSIF will become the bottleneck of the transmission bandwidth. When performing upstream data flow transmission, the calculation of the egress bandwidth should be as follows:

Wherein, the left side of the equation represents the bandwidth of the destination port connected to the host (there is only one destination port), and the parameter on the right side of the equation represents the sum of the bandwidths of the source ports connected to the I/O device. It can be seen that when there are more source ports for simultaneous transmission, the bandwidth pressure of the destination port target port is greater.

Second, the arbitration mechanism of the arbitration module in the related art does not consider the bandwidth. As mentioned above, the arbitrator 204 of the related art adopts a polling mechanism without considering the bandwidth factor. If the transmission length of the previous client client is very large and occupies the destination port target port all the time, then the latter client client will be in front of the previous client. The client cannot obtain the transmission of the target port before the transmission is completed, which affects the transmission of key data. For example, a kind of transmission situation is, as shown in Figure 3, if channel ① is transmitting, and channel ② has the data of real-time transmission need to obtain the transmission right of the BUSIF connected with host 100 as soon as possible, but because channel ① has not been transmitted, and If the transmission cannot be interrupted, the channel ② cannot obtain the transmission right, which will cause the real-time transmission of the channel ② to be interrupted. Another transmission situation is that the transmission type sent by path ① each time is batch transmission, that is, the size of data transmitted each time is large, while the size of data transmitted by path ② is small, but the transmission frequency is high. When the channel ② releases the destination port target port soon each time, the channel ① occupies the destination port target port for a long time each time, and the average bandwidth of the channel ② cannot be guaranteed on the whole.

Just because of the above-mentioned problems existing in the related technology, some embodiments of the present application increase the bandwidth monitoring mechanism on the basis of the IOHUB of the related technology, improve the arbitration mechanism according to the bandwidth, and increase the quantity of the destination port target port at the same time, by increasing the destination port target port To increase the export bandwidth flowing to the host or on-chip memory, further in some embodiments of the present application, it is possible to flexibly adjust the bandwidth of the data to be transmitted in multiple destinations according to the bandwidth required by the source port source port and the bandwidth of the monitored destination port Bandwidth on port target port.

For example, some embodiments of the present application increase the number of bus interfaces interconnected with the host or memory (each bus interface corresponds to a port), and then increase the number of corresponding destination ports when the I/O device acts as a data sender; Some embodiments of the application also add a bandwidth monitoring mechanism for ports, improve the arbitration and routing involved in the existing transmission strategy based on the bandwidth monitoring results, and increase the flexibility of routing.

The routing method for the on-chip bus executed by the IOHUB of the embodiment of the present application will be exemplarily described below with reference to FIG. 4 .

Please refer to Fig. 4, Fig. 4 is the routing method for the on-chip bus provided by the embodiment of the present application, including: S101, obtaining the actual transmission bandwidth occupancy of each destination port in at least one destination port on the bus; S102, obtaining the at least one destination port The predicted transmission bandwidth occupancy of each destination port in a destination port in the expected future time period; S103, according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy, determine the transmission for the data to be transmitted from the source port Strategy.

It should be noted that, for uplink data transmission, the destination port refers to a port connected to a host or an on-chip memory; for downlink data transmission, the destination port is a port connected to an on-chip bus and an I/O device. When the destination port is a port connected to an I/O device, the number of destination ports is one, and the transmission strategy determined according to the bandwidth occupancy (including the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy) in the embodiment of the present application is used It is used to solve the problem of how to make decisions when multiple hosts transmit data to the destination port at the same time. When the destination port is a port connected to a host or an on-chip storage unit, the number of destination ports corresponding to some embodiments of the present application is increased to multiple, and the bandwidth occupancy of this application (including actual transmission bandwidth occupancy and predicted transmission) is adopted. Bandwidth Occupancy) The transmission strategy determined is used to select one or more exits from multiple destination ports for the data to be transmitted, through which the data to be transmitted can be sent to the host or on-chip memory.

Taking uplink data transmission as an example below, various steps involved in the scheme in FIG. 4 are exemplarily described.

In some embodiments of the present application, the actual transmission bandwidth occupation involved in S101 is determined according to the transmission bandwidth of the current transmission within the transmission window of any destination port. For example, the formula for calculating the actual transmission bandwidth usage of any destination port is as follows:

Among them, BW_TRANS is used to represent the transmission bandwidth of the current transmission within the transmission window, CLK COUNT is used to count the number of system clock cycles within the set transmission window, and TRANS COUNT is used to represent the number of bytes corresponding to all valid transmissions of any port .

CLK COUNT is the number of clock cycles (that is, the number of clock cycles) within the statistically obtained transmission window range. The size of the transmission window (statistical transmission cycle range) can be obtained based on empirical values, and the value of the transmission window is preset through the register.

The calculation window for calculating the actual transmission bandwidth usage is shown in Figure 10. After the software sets the size of the transmission window, the size of the transmission window does not change during the moving process, but the bandwidth will change as the window moves. As shown in FIG. 5 , the actual transmission bandwidth occupancy in the first transmission window Cal_Window is greater than the actual transmission bandwidth occupancy in the second transmission window Cal_Window.

In some embodiments, the predicted transmission bandwidth occupancy involved in S102 is determined according to the transmission bandwidth of the expected transmission of any destination port within the expected transmission window, wherein the expected transmission window is determined according to the The total number of system clock cycles determined by the calculation of the transfer length and transfer bit width corresponding to the expected transfer. For example, the predictable transmission window is the total number of system clock cycles obtained by transmitting a piece of data with a data bit width and length at the start of transmission at each system clock cycle according to the transmission length corresponding to the predictable transmission. For example, the calculation formula of the predicted transmission bandwidth occupancy is as follows:

BW_TRANS=EST/CLK COUNT1 where CLK COUNT1 is the count value of the system clock within the expected transmission window, and EST is the count value obtained by counting the transmission length carried by the current transmission request to represent the data.

It should be noted that the expected transmission refers to the amount of transmission obtained according to the length of the transmission. For example, if the transmission length in the transmission request is 256Byte, the expected transmission is 256Byte; the expected transmission window refers to the calculation based on the transmission length and transmission bit width The obtained transmission cycle number, if the transmission length is 256Byte, and the transmission bus bit width is 32Byte, then the expected transmission window is 8 Cycles. For the specific values of the above parameters, reference may be made to the description below, and details will not be repeated here.

The following takes the bus interface connected to the host or the on-chip memory as the destination port (that is, transmits the upstream data stream) as an example to illustrate S103.

In order to improve the problem that there is only one port connected to the host or the on-chip memory in the related art, the data to be transmitted from the I/O device cannot be sent to the host or the memory immediately. In some embodiments of the present application, the host or the on-chip memory It is connected to the IO router through multiple bus interfaces, and multiple bus interfaces correspond to multiple ports when performing data transmission.

As shown in FIG. 6 , the difference from FIG. 1 and FIG. 2 is that the host 100 in FIG. 6 is connected to the IO router through four bus interfaces. It can be understood that the data from the I/O device can be transmitted to the host 100 through four bus interfaces, that is, one or more destination ports in the four destination ports, which significantly increases the total number of simultaneous data transfers to the host or memory. bandwidth.

As shown in Figure 6, when the I/O device is used as the source port, the destination port target port is increased to four. At this time, the export bandwidth is increased by four times that of the original solution, and more I/O devices can be provided at the same time. transmission. As shown in FIG. 7 , the source port and the target port can implement arbitrary routing, that is, any source port can be routed to any target port. In actual use, when the source port is a bus interface connected to a host or an on-chip memory, the corresponding target port is a port connected to an I/O device. It should be noted that, in some embodiments of the present application, when the data flow direction is downstream, the route from the source port source port to the destination port target port is determined by the address space it visits; when the source port sourceport is the same as For the interface connected to the I/O device, when the destination port target port is a bus interface connected to the host or memory, that is, when the data flow direction is upstream, the route from the source port source port to the destination port target port is determined by some embodiments of the present application The monitoring bandwidth is determined.

As an example, when the total bandwidth of the source port source port is lower than the destination port target port, the source port source port can be routed to any destination port target port to ensure that the bandwidth requirements of all source port source ports are met and are not used The destination port target port is in the idle IDLE state; when the total bandwidth of the source port source port is equal to or higher than the destination port target port, the route from multiple source ports source port to the destination port target port needs to be based on the bandwidth of the destination port target port Dynamic adjustment. For example, the first source port Source_0 has a transfer to the destination port target port. When the transfer request is sent, the first destination port Target_0port and the second destination port Target_1port are both in the transmission state, and the third destination port Target_2port is in idle IDLE state, the data to be transmitted on the first source port Source_0 is routed to the third destination port Target_2; if the four destination ports Target ports are all occupied, then according to the bandwidth calculation result (that is, the estimated transmission bandwidth occupancy is calculated), the second The data to be transmitted on the source port Source_0 is routed to the target port Target port with the lowest expected bandwidth (ie, expected transmission bandwidth occupancy).

An exemplary strategy for routing according to bandwidth will be described below with reference to examples.

When the I/O device transmits data to the host computer or memory through multiple destination ports, there is a technical problem of reasonable routing. The purpose of the amount of data transmitted by the memory, correspondingly, S103 includes: according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy, at least one target exit is determined from multiple destination ports for the data to be transmitted; wherein, the target exit uses for providing the data to be transmitted to the host or the on-chip memory. For example, S103 may include determining at least one target egress from the multiple destination ports according to the transmission bandwidth threshold and the predicted bandwidth threshold respectively set for each of the multiple destination ports.

When both the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy of a certain destination port are smaller than the corresponding set threshold, S103 determines that this destination port can be used as a target egress of the data to be transmitted. In some embodiments of the present application, the S103 includes: confirming that the first actual transmission bandwidth occupancy corresponding to the first destination port (which is any one of the multiple destination ports) is smaller than the first destination port corresponding to the first destination port a transmission bandwidth threshold, and confirm that the first predicted transmission bandwidth occupancy corresponding to the first destination port is less than the first predicted bandwidth threshold corresponding to the first destination port; select the first destination port as the target egress .

That is to say, when the actual transmission bandwidth of a destination port is less than the transmission bandwidth threshold and the predicted transmission bandwidth is less than the predicted bandwidth threshold, this destination port can be used as a target egress of the data to be transmitted. It should be noted that if the amount of data to be transmitted from a certain source port is greater than the bandwidth that the destination port can provide, the data to be transmitted from the source port can be distributed to multiple target outlets. An export is just one of these target exports.

As an example, see Figure 8. In Figure 8, there are three I/O devices accessing the DDR in the host host. Before exemplifying the transmission strategy, it is assumed that the device or link in Figure 8 also has the following initial parameters:

First, in Figure 8, there are 3 devices (that is, 3 I/O devices) that initiate write transfer requests, and the information carried by these three write transfer requests indicates that the average bandwidth of the data to be transferred is 32GB/s.

Second, the three devices in Figure 8 are connected to the host through two bus interfaces. Based on the first assumption, it can be known that the number of destination ports for this transmission is two, and the full-load bandwidth of these two destination ports is 48GB/ s.

Third, assume that the initial default transmission route in Figure 8 is S1→T1; S2→T2; S3→T2

Fourth, the bus bit width of the Device is 32Byte, that is, the effective transmission of a system transmission cycle cycle is 32Byte; the bus bit width of the host Host is 48Byte, that is, the effective transmission of a system clock cycle cycle is 48Byte.

Fifth, assume that the operating frequency is 1GHz, that is, the ideal bandwidth of the bus is 32GB/s.

Sixth, assume that the length of the write transfer initiated by each Device is 256Byte.

The workflow in Figure 8 is illustrated below in combination with the above six assumptions:

S201, setting the transmission parameters related to the device Device and the host Host, the device device is for write transmission, and the transmission length is 256 Byte.

S202. Set the transmission bandwidth threshold of T1 to 38GB/s (80% of the ideal bandwidth), and the predicted bandwidth threshold of T1 to 48GB/s (ideal bandwidth).

S203. Set the transmission bandwidth threshold of T2 to 38GB/s (80% of the ideal bandwidth), and the predicted bandwidth threshold of T1 to 48GB/s (ideal bandwidth).

S204, setting the size of the transmission window of the monitor of T1 to 1024 (that is, the size of the set transmission window is 1024 Cycles), and setting the size of the window of the monitor of T2 to 1024.

S205, start the transmission, and the monitors of each destination port start to work.

S206, when the first transmission window ends, perform the following steps:

S206-1, the actual transmission volume TRANS COUNT of T1 is 1000*32Byte, and the CLK COUNT is 1000. According to the calculation formula of BW_TRANS, the bandwidth conversion (that is, the actual transmission bandwidth occupancy) is 32GB/s; the predicted transmission volume ESTCOUNT of T1 is 256Byte , the CLK COUNT required to complete 256Byte transmission is 8 system clock cycles Cycle, so according to the bandwidth calculation formula of EST_TRANS, the bandwidth conversion (that is, the predicted transmission bandwidth occupancy) is 32GB/s. Both bandwidth values are smaller than the preset threshold, that is, the bandwidth on T1 is not fully occupied, and transmissions from other source ports can be received.

S206-2, the actual transmission volume TRANS COUNT of T2 is 1000*48Byte, and the CLK COUNT is 1000. According to the calculation formula of BW_TRANS, the bandwidth conversion (that is, the actual transmission bandwidth occupancy) is 48GB/s.

S206-3, T2's predicted transmission amount EST COUNT from S2 is 256Byte, and the CLK COUNT required to complete 256Byte transmission is 6 Cycles, so T2's BW_EST conversion from S2 (that is, the actual transmission bandwidth occupancy) is 43GB/s; The predicted transmission amount EST_COUNT from S3 of T2 is 256Byte, and the CLK COUNT required to complete the 256Byte transmission is 6 Cycles, so the BW_EST conversion (that is, the actual transmission bandwidth usage) of T2 from S2 is 43GB/s. So the total estimated bandwidth requirement on the T2 port is 86GB/s.

S206-4, so the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy on T2 are both higher than the corresponding threshold.

S206-5, the software changes the route of S2 to T1.

S207, at the end of the second transmission window, the bandwidth calculation results of T1 and T2 are opposite to #6, and the software changes the route of S2 to T2.

On an overall average, the bandwidth of S2 in FIG. 8 is evenly allocated 16 GB/s to the destination port T1 and the destination port T2.

It should be noted that, in some embodiments of the present application, the transmission strategy is determined by judging characteristics of the data to be transmitted from any source port. For example, if the source port still has a large amount of data to be transmitted, it is necessary to refer to the predicted transmission threshold to determine whether a destination port can be used as the target egress; if the amount of data to be transmitted on the source port is small, the transmission bandwidth threshold can be used to Determines whether a destination port can be used as a target egress.

As an example, when one of the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy of a certain destination port in some embodiments of the present application is less than the corresponding set threshold, S103 may include: according to the attribute of the data to be transmitted The characteristic is to confirm that at least one of the transmission bandwidth threshold and the predicted bandwidth threshold is used to determine the target egress, wherein the attribute characteristic is used to characterize the amount of data to be transmitted. For example, it includes: confirming that the amount of data to be transmitted is greater than a first set threshold; confirming that the second predicted transmission bandwidth occupancy corresponding to the second destination port is smaller than the second predicted bandwidth threshold corresponding to the second destination port; Selecting the second destination port as the target egress; or, confirming that the amount of data to be transmitted is less than a second set threshold; confirming that the second actual transmission bandwidth occupancy corresponding to the second destination port is less than the set threshold A second transmission bandwidth threshold corresponding to the second destination port; selecting the second destination port as the target egress; wherein the first set threshold is greater than the second set threshold. The above-mentioned first destination port is any one of multiple destination ports.

Please refer to FIG. 9. FIG. 9 shows the routing device used in the bus through the embodiment of the present application. It should be understood that the device corresponds to the method embodiment in FIG. 4 above, and can perform various steps involved in the method embodiment above. For specific functions of the device, reference may be made to the above description, and to avoid repetition, detailed descriptions are appropriately omitted here. The device includes at least one software function module that can be stored in the memory in the form of software or firmware or solidified in the operating system of the device. The routing device on the bus includes: a bandwidth occupancy calculation module 801 configured to: Acquire the actual transmission bandwidth occupancy of each destination port in at least one destination port on the bus; obtain the predicted transmission bandwidth occupancy of each destination port in the at least one destination port in the future preset time period; the arbitration module 802 is configured to A transmission policy is determined for the data to be transmitted from the source port according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy.

Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the method in FIG. 4 , which will not be repeated here.

The routing system for the on-chip bus according to the embodiment of the present application is exemplarily described below with reference to FIG. 10 .

Some embodiments of the present application provide a routing system for an on-chip bus, the routing system includes: a system clock counting module 210 configured to add 1 to each transmission clock cycle when monitoring starts, and when the set transmission clock cycle is reached When the window value is cleared, and restart the clock counting in the next monitoring transmission window; the actual transmission amount counting module 220 is configured as the actual transmission amount corresponding to the effective transmission according to the transmission effective signal statistics (i.e. the TRANSCOUNT recorded above) ); The predicted transmission volume counting module 230 is configured to predict the transmission volume (i.e. the EST recorded above) on the channel according to the transmission request signal in the expected time period in the future; the bandwidth calculation module 240 is configured to: according to the set Transmission window size (i.e. REG_WINDOW of Fig. 10) and a plurality of said actual transmission quantities corresponding to a purpose port determine the actual transmission bandwidth occupation of said destination port (i.e. BW_TRANS of Fig. 10); according to the predictable transmission window size and A plurality of the predicted transmission amounts corresponding to the destination port determine the predicted transmission bandwidth occupancy (ie, BW_TRANS_EST in FIG. 10 ); the arbiter 250 is configured to: compare the actual transmission bandwidth occupancy of the destination port with the set The transmission bandwidth threshold (i.e. REG_THRESHOLD of Figure 10), obtains the first comparison result; compares the predicted transmission bandwidth occupancy of the destination port with the set predicted transmission bandwidth threshold, obtains the second comparison result; according to the first comparison result and As a result of the second comparison, a target exit selection signal (ie TRANS_BLOCK in FIG. 10 ) is generated, wherein the target exit is an exit selected from a plurality of destination ports for transmitting data to be transmitted; router 260 is selected by It is configured to receive the target outlet selection signal, so as to send the data to be transmitted to the target outlet.

In some embodiments of the present application, when it is confirmed that the transmission valid signal is at a first level, the actual transmission amount counting module is started to perform one count, wherein the first level is a high level or a low level. For example, the counting step of the one count is determined according to the transmission bit width of the channel.

As an example, a valid signal, Valid_Transfer, is transmitted as shown in FIG. 10 . Valid_Transfer refers to a valid bus transfer. Valid_Transfer is calculated separately for each channel on the bus. The valid condition is the valid transmission signal corresponding to the channel, and the counter of the corresponding actual transmission amount counting module 220 is incremented once for each valid transmission. On the data channel, each counter increases the number of Bytes corresponding to the transmission, that is, if the data channel is 256-bit wide, and the transmitted byte enable is valid, the counter corresponding to each valid transmission should be incremented by 32.

In some embodiments, when it is confirmed that the transmission request signal is at a second level, the predicted transmission amount counting module is started to perform one count, wherein the second level is a high level or a low level. For example, the counting step of the one count is determined according to the transmission length of the data to be transmitted this time carried by the transmission request signal.

As shown in FIG. 10 , the transmission request signal is Request: used to predict the bandwidth occupation of the upcoming transmission, and refers to the transmission command channel on the bus. In the transmission command channel, when the transmission request signal Request is valid, the number of bytes that the counter increases each time is the transmission length. If the transmission length is 256 Byte, the counter increases by 256. The prediction method is: in the next 256 transmission cycles, the corresponding The data channel will transmit 256 Bytes of data. The predicted transmission amount counting module 230 is used to predict the bandwidth of the data channel, but does not consider the situation of transmission congestion, so the actual bandwidth should be less than the predicted bandwidth.

Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the device described above can refer to the corresponding process in the method in FIG. 4 , which will not be repeated here.

As shown in FIG. 11 , this figure is a schematic structural diagram of the interconnection between the IOHUB, the host (or storage) and the IO device provided by the embodiment of the present application. Different from Fig. 2, each host or memory in Fig. 11 has four bus interfaces (one uplink signal and one downlink signal in Fig. 11 correspond to one bus interface), and each bus interface module in Fig. 11 can also be set Monitor monitor, the output signal of the monitor monitor (for example, the output signal includes the actual transmission bandwidth occupancy BW_TRANS and the predicted transmission bandwidth occupancy BW_TRANS_EST as shown in Figure 10) after being buffered by the buffer BUF, the input value arbitration is arbirtition to perform routing arbitration , and then select a target outlet with an appropriate bandwidth for the data sent from the source port. For the same units in FIG. 11 as in FIG. 2 , no process statement is made here.

The bandwidth monitoring device monitor in FIG. 11 includes the system clock counting module 210 , the actual transmission volume counting module 220 , the predicted transmission volume counting module 230 and the bandwidth calculation module 240 in FIG. 9 . For the functions of these modules, reference can be made to the above description, and in order to avoid repetition, details are not repeated here.

Some embodiments of the present application provide a system-on-chip data transmission method. The data transmission method includes: before an effective transmission starts: setting the value of the transmission window corresponding to each of the multiple destination ports, and setting the A bandwidth comparison threshold corresponding to each destination port in the plurality of destination ports, wherein the bandwidth comparison threshold includes a transmission bandwidth comparison threshold and a predicted transmission bandwidth threshold; when effective transmission starts, continuously update each destination port in the plurality of destination ports The actual transmission bandwidth occupancy and the value of the predicted transmission bandwidth occupancy, wherein the actual transmission bandwidth occupancy is calculated according to the value of the transmission window and the statistical actual transmission amount, and the predicted transmission bandwidth occupancy is Calculated according to the value of the predicted transmission window and the predicted transmission volume obtained by statistics; during arbitration, dynamically adjust the output routing result according to the bandwidth comparison threshold of each destination port in the plurality of destination ports, so as to transfer the routing results from the source port The data to be transmitted is routed to the adjusted target egress.

That is to say, the data transmission method of the system on chip in the embodiment of the present application is shown in FIG. 12 . Before starting the transmission, the software sets the transmission window size and bandwidth comparison threshold for bandwidth calculation through registers. When the transmission starts, it calculates the update bandwidth (that is, calculates the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy). Arbitration is based on each purpose The bandwidth threshold of the port target port dynamically adjusts the output routing result, and the routing result is updated according to the bandwidth to route the source port source port to the adjusted destination port target port. When a transmission ends, the source port sourceport will end if there is no subsequent transmission. If the transmission continues, the previous bandwidth calculation and routing adjustment will be repeated. When doing arbitration routing, the software needs to set REG_THRESHOLD of the target port corresponding to the destination port, that is, the bandwidth threshold of the port port (that is, set the transmission bandwidth threshold and set the predicted bandwidth threshold), when the actual transmission bandwidth of the corresponding destination port target port is greater than the set bandwidth When the threshold is set, other source port source ports will not be routed to this destination target port, otherwise, other source port source port routes to this destination port target port can be accepted.

Some embodiments of the present application provide a system on a chip, the system on a chip includes a processor, a memory, at least one I/O device, and the above-mentioned routing system and device; wherein, the at least one I/O device and the The processors are interconnected through the routing system or routing device; or, the at least one I/O device and the memory are interconnected through the routing system or routing device.

In the several embodiments provided in this application, it should be understood that the disclosed devices and methods may also be implemented in other ways. The device embodiments described above are only illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functions and possible implementations of devices, methods and computer program products according to multiple embodiments of the present application. operate. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of code that includes one or more Executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions.

In addition, each functional module in each embodiment of the present application may be integrated to form an independent part, each module may exist independently, or two or more modules may be integrated to form an independent part.

If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including Several instructions are used to make a computer device (which may be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk and other media that can store program codes. .

The above descriptions are only examples of the present application, and are not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and changes may be made to the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this application shall be included within the protection scope of this application. It should be noted that like numerals and letters denote similar items in the following figures, therefore, once an item is defined in one figure, it does not require further definition and explanation in subsequent figures.

The above is only a specific implementation of the application, but the scope of protection of the application is not limited thereto. Anyone familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the application. Should be covered within the protection scope of this application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a relationship between these entities or operations. There is no such actual relationship or order between them. Furthermore, the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus comprising a set of elements includes not only those elements, but also includes elements not expressly listed. other elements of or also include elements inherent in such a process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

Claims (19)

1. A routing method for on-chip bus, applied to the IO routing module, is characterized in that, the routing method includes: Obtaining the actual transmission bandwidth occupancy of each destination port among the plurality of destination ports on the bus, wherein the destination port is a port connected to an on-chip memory or a host; Acquiring the predicted transmission bandwidth occupancy of each destination port among the plurality of destination ports in a future estimated time period; Comparing at least one of the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy with a corresponding set threshold to obtain a comparison result; Selecting at least one target exit from the plurality of destination ports based at least on the comparison result, wherein the target exit is used to provide the host or the on-chip memory with data to be transmitted from a source port, the source port It is a port connected with various I/O devices. 2. The routing method according to claim 1, wherein said determining at least one target exit from a plurality of said destination ports based at least on said comparison result comprises: If the comparison result is that the first actual transmission bandwidth occupancy corresponding to the first destination port is less than the first transmission bandwidth threshold corresponding to the first destination port and the first predicted transmission bandwidth occupancy corresponding to the first destination port is smaller than the first predicted bandwidth threshold corresponding to the first destination port, then select the first destination port as the target egress. 3. The routing method according to claim 1, wherein the selecting at least one target outlet from the plurality of destination ports based at least on the comparison result comprises: Selecting at least one target exit from the plurality of destination ports according to the attribute feature of the data to be transmitted and the comparison result, wherein the attribute feature is used to characterize the amount of the data to be transmitted. 4. The routing method according to claim 3, wherein the selecting at least one target exit from the plurality of destination ports according to the attribute characteristics of the data to be transmitted and the comparison result comprises: confirming that the data volume of the data to be transmitted is greater than a first set threshold; If the comparison result is that the second predicted transmission bandwidth occupancy corresponding to the second destination port is smaller than the second predicted bandwidth threshold corresponding to the second destination port, then selecting the second destination port as the target egress. 5. The routing method according to claim 3, wherein the selecting at least one target exit from the plurality of destination ports according to the attribute characteristics of the data to be transmitted and the comparison result comprises: confirming that the data volume of the data to be transmitted is less than a second set threshold; If the comparison result is that the second actual transmission bandwidth occupation corresponding to the second destination port is smaller than the second transmission bandwidth threshold corresponding to the second destination port, then selecting the second destination port as the target egress. 6. The routing method according to any one of claims 1-5, wherein the actual transmission bandwidth occupancy is determined according to the transmission bandwidth of the current transmission within the transmission window. 7. routing method as claimed in claim 6 is characterized in that, the calculation formula of described actual transmission bandwidth occupancy is as follows:

Among them, CLK COUNT is the number of system clock cycles counted within the set transmission window, COUNT is used to represent and count the number of bytes corresponding to all valid transmissions at the current moment. 8. The routing method according to claim 7, characterized in that, the value of the parameter TRANSCOUNT is counted by monitoring the transmission effective signal in the channel. 9. The routing method according to any one of claims 1-5, wherein the predicted transmission bandwidth occupancy is determined according to the transmission bandwidth of the expected transmission within the expected transmission window, wherein the The predictable transmission window is the total number of system clock cycles determined according to the transmission length and transmission bit width corresponding to the predictable transmission. 10. routing method as claimed in claim 9 is characterized in that, the computing formula of described predicted transmission bandwidth occupancy is as follows:

Wherein, CLK COUNT1 is a system clock count value within the predictable transmission window, and EST is a count value obtained by counting the transmission length carried by the current transmission request to represent the data. 11. The routing method according to claim 10, wherein the value of the EST is obtained by extracting the information carried in the transmission request signal in the channel. 12. A routing device for on-chip bus, characterized in that, the routing device comprises: The bandwidth occupancy calculation module is configured to: Obtaining the actual transmission bandwidth occupancy of each destination port among the plurality of destination ports on the bus, wherein the destination port is a port connected to an on-chip memory or a host; Acquiring the predicted transmission bandwidth occupancy of each destination port among the plurality of destination ports in a future estimated time period; an arbitration module configured to: Comparing at least one of the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy with a corresponding set threshold to obtain a comparison result; Selecting at least one target exit from the plurality of destination ports based at least on the comparison result, wherein the target exit is used to provide the host or the on-chip memory with data to be transmitted from a source port, the source port It is a port connected with various I/O devices. 13. A routing system for a system on chip, characterized in that the routing system comprises: The system clock counting module is configured to increase by 1 for each transmission clock cycle when the monitoring starts, and clear to zero when the set transmission window value is reached, and restart the clock counting in the next monitoring transmission window; The actual transmission amount counting module is configured to count the actual transmission amount corresponding to the effective transmission according to the effective transmission signal; The forecasted transmission volume counting module is configured to count the forecasted transmission volume in the expected time period in the future according to the transmission request signal; The bandwidth calculation module is configured to: Obtain the actual transmission bandwidth occupancy of each of the multiple destination ports on the bus, wherein the actual transmission of the destination port is determined according to the set transmission window size and the multiple actual transmission volumes corresponding to one destination port bandwidth usage; Obtaining the predicted transmission bandwidth occupancy of each destination port in the plurality of destination ports in a future expected time period, wherein the determined according to the predictable transmission window size and the plurality of predicted transmission amounts corresponding to the destination ports Predict transmission bandwidth occupancy; Arbitrator, configured as: Comparing the actual transmission bandwidth occupancy of the destination port with the set transmission bandwidth threshold to obtain a first comparison result; Comparing the predicted transmission bandwidth occupancy with the set predicted transmission bandwidth threshold to obtain a second comparison result; generating a selection signal of a target outlet according to the first comparison result and the second comparison result, and selecting at least one target outlet from a plurality of destination ports based on the selection signal, wherein the target outlet is An exit selected from multiple destination ports for transmitting data to be transmitted, the target exit is used to provide the host or on-chip memory with data to be transmitted from the source port, and the source port is connected to various I/O A port connected to the device, the destination port is a port connected to the on-chip memory or the host; A router configured to receive the target exit selection signal, so as to send the data to be transmitted to the target exit. 14. The routing system for a system on chip according to claim 13, characterized in that, when it is confirmed that the transmission valid signal is at the first level, the actual transmission amount counting module is started to count once, wherein the first level Level is high level or low level. 15. The routing system for a system on chip according to claim 14, characterized in that, the counting step of one count is determined according to the transmission bit width of the channel. 16. The routing system for a system-on-chip as claimed in claim 13, wherein when the transmission request signal is confirmed to be at a second level, the predicted transmission amount counting module is started to perform a count, wherein the second level Level is high level or low level. 17. The routing system for a system on chip according to claim 16, wherein the counting step of the one count is determined according to the transmission length of the data to be transmitted carried in the transmission request signal. 18. A data transmission method of a system on chip, characterized in that the data transmission method comprises: Before a valid transfer begins: Setting the value of the transmission window corresponding to each destination port in the plurality of destination ports, and setting the bandwidth comparison threshold corresponding to each destination port in the plurality of destination ports, wherein the bandwidth comparison threshold includes a transmission bandwidth comparison threshold and a predicted transmission bandwidth Threshold, the destination port is a port connected to the on-chip memory or the host; Acquiring the actual transmission bandwidth occupancy of each of the multiple destination ports on the bus; Acquiring the predicted transmission bandwidth occupancy of each destination port among the plurality of destination ports in a future estimated time period; When effective transmission starts, continuously update the values of the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy of each of the plurality of destination ports, wherein the actual transmission bandwidth occupancy is based on the The value of the transmission window and the calculated actual transmission volume are calculated, and the predicted transmission bandwidth occupancy is calculated according to the value of the predicted transmission window and the statistically obtained predicted transmission volume; During arbitration, at least one of the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy is compared with a corresponding set threshold to obtain a comparison result; at least based on the comparison result, the output routing result is dynamically adjusted, so that the output from The data to be transmitted at the source port is routed to the adjusted target exit, wherein at least one of the target exits is selected from the plurality of destination ports according to the actual transmission bandwidth occupancy and the predicted transmission bandwidth occupancy, the The target outlet is used to provide the host or the on-chip memory with data to be transmitted from a source port, and the source port is a port connected to various I/O devices. 19. A system on a chip, characterized in that the system on a chip comprises a processor, a memory, at least one I/O device, and the routing system according to any one of claims 13-17; Wherein, the at least one I/O device is interconnected with the processor through the routing system; or, The at least one I/O device is interconnected with the memory through the routing system.

Images