KR100901691B1 - Network communication system of mesh-star mixing on-chip and communication method thereof - Google Patents

Abstract

A mesh-star mixing on-chip network communications system and a communications method thereof for implementing the expandability of minimum delay time and optimizing the communication characteristic of SoC design are provided to compose a hierarchical communications structure by connecting mesh-star mixing on-chip. Star switches formulate star networks. The star switches switch packets of star IP cores according to a source routing mode. A plurality of mesh switches forms mesh network(20). Mesh IP cores are one by one connected to each mesh switch. Bridges connect star networks and mesh network. The mesh-star mixing on-chip network communications system optimizes the communication characteristic of system on chip design.

Description

Network communication system of Mesh-star mixing on-chip and communication method

The present invention relates to a communication structure of a system on the chip (SoC) system, in particular, a communication resource contention between heterogeneous communication requests on a network and network scalability and communication path that guarantees maximum communication parallelism and minimum communication delay time. The present invention relates to a mesh-star mixed on-chip network communication system and a communication method thereof, which provide a communication structure of a large SoC system by mixing a two-dimensional mesh network having very good avoidance.

The present invention is derived from the research conducted as part of the IT source technology development of the Ministry of Information and Communication and the Ministry of Information and Communication Research and Development.

SoC design methodology is focused on platform-oriented design that enables short-term development of SoCs for multi-type systems. In this case, the data communication structure together with the processor is the core of the design.

However, as the number of devices integrated in the SoC increases, the amount of data transmitted and received between each component module is rapidly increasing, making the design of the data communication structure more difficult.

On-chip buses, such as the multi-layer Advanced Microcontroller Bus Architecture (AMBA) and Silicon Backplane, which are widely used in embedded systems, are limited by data bandwidth and scalability limitations, excessive power consumption, and various interface requirements. Due to reusability, it is a factor of great performance limitation in applications such as multimedia, which requires a large amount of data communication.

Therefore, the SoC design that can integrate various processor cores, memory, IP, etc. on a single chip requires an on-chip communication structure that is scalable and reusable.

It is on-chip network that applied computer network technology on chip in response to these demands. Examples of such on-chip networks include stars, two-dimensional meshes, hypercubes, trees, and toruss, and an appropriate network structure is selected according to data traffic of an application field. .

However, in the case of the existing on-chip network, it is necessary to design a unified communication structure based on the most communication cores rather than reflecting the local communication characteristics of the application design, so that the performance of the entire system can be guaranteed. In this case, too much communication resources are allocated, resulting in an inefficient communication structure.

In addition, in order to maximize the data bandwidth in the on-chip network, the operating speed of the crossbar switch, which is the basis of the switching, should be maximized, but in this case, power consumption increases rapidly with the increase of the operating clock frequency. There is a difficult problem in reality.

In order to solve this problem, an on-chip network communication structure is required that can lower the operating clock frequency of the crossbar switch and design the communication structure to have an optimal topology to obtain a desired data bandwidth within a predetermined power consumption.

1 is a diagram illustrating a topology of a star type on-chip network according to the related art.

As shown in FIG. 1, the star type on-chip network has a structure in which all IP cores (SIP) are connected to one star switch (SS).

Accordingly, all IP cores (SIP) connected to the star switch (SS) can communicate at the same time, and have the maximum communication parallelism and the minimum delay time.

In addition, it is possible to predict the communication delay time between the IP cores (SIP) to ensure the communication quality of service (QoS), and thus real-time where the delay time of communication between the IP cores (SIP) is very important. Suitable for SoC design.

However, since the number of IP cores that can be connected to one star switch SS is limited, the star type on-chip network structure of FIG. 1 has a big problem in scalability.

2 is a diagram illustrating a topology of a mesh type on-chip network according to the related art.

As shown in FIG. 2, a mesh type on-chip network has a structure in which one IP core MIP is connected to each of a plurality of mesh switches MS connected to each other.

In the case of the mesh type on-chip network, by increasing the number of mesh switches (MS), the connectable IP core (MIP) can also be increased indefinitely, which has the advantage of very good scalability.

However, the basic communication delay time between IP cores (MIP) is relatively long, a large number of mesh switches (MSs) are required, and the communication delay time is difficult to predict.

On the other hand, in case of SoC with application purpose such as MPEG4, H.264, HDTV, etc., it is composed of IP cores with various functions and communication patterns, but the communication patterns between IP cores have predictable consistency at the design stage.

In other words, by analyzing the communication pattern between cores in the initial stage of design by static or dynamic method (profiling), it is possible to design the area, performance and The communication structure can be designed much better in terms of power consumption.

As a result of profiling the multimedia SoC designs, more than 80% of the IP cores communicate with up to four other IP cores.

This means that in most designs, a large amount of communication occurs locally, and by enabling these local communication needs at maximum performance, the overall system's performance efficiency can be maximized.

To this end, a hybrid communication structure combining a highly scalable global communication structure that can easily design a large-scale system based on a variety of regional communication structures by introducing a communication structure that best suits local communication needs. Is required.

In order to satisfy the above requirements, the present invention has the advantages of both a star network and a mesh network, and can be optimized for communication characteristics of an SoC design, and is a dual-layer on-chip network communication having a minimum communication delay time. I would like to propose a structure.

That is, the entire communication structure is not implemented as an on-chip network of one identical topology, but as a whole, it has a hierarchical communication structure implemented with a topology and a communication protocol of a two-dimensional mesh structure and a star topology and a protocol. We propose a mesh-star mixed on-chip network communication system and a communication method thereof.

According to an aspect of the present invention as a means for solving the above problems, a star switch to form a star network and switching the packet of the star IP core according to the source routing scheme; A mesh switch forming a mesh network and switching packets of the mesh IP core according to an X-Y routing scheme; And a connection between the star switch and the mesh switch, inserts routing information according to the XY routing scheme into a packet transmitted from the star switch, and transfers the routing information to the mesh switch. Provided is a mesh-star mixed on-chip network system including a bridge which inserts routing information according to a routing scheme and then transfers the routing information to the star switch.

The packet may include a header field including at least one of source routing information, (X, Y) coordinate information, type of packet (Read / Write), and packet size information; An address field containing a destination address; And a plurality of data fields in which data is stored, wherein each of the header field, the address field, and the plurality of data fields has a size corresponding to a fleet unit.

And the bridge comprises: a routing table for managing source routing information and (X, Y) coordinate information for each of the different bridges, the star IP core, the star switch, the mesh IP core, and the mesh switch; A star / mesh converter which, when the packet is transmitted from the star switch, obtains (X, Y) coordinate information corresponding to the packet from the routing table, adds it to the packet, and then transfers the packet; And a mesh / star converter which, when the packet is transmitted from the mesh switch, obtains source routing information corresponding to the packet from the routing table, adds it to the packet, and then transfers the packet.

In this case, the star switch may be connected to two or more bridges or may be connected to a star switch different from itself.

According to an aspect of the present invention as a means for solving the above problems, the (X, Y) coordinate information is inserted into the packet transmitted from the star network and transmitted to the mesh network, the mesh network is the (X, Y) star / mesh transmission step of delivering the packet to a destination IP core using coordinate information; And inserting source routing information into the packet transmitted from the mesh network and transmitting the source routing information to the star network, wherein the star network transmits the packet to a destination IP core through the source routing information. It provides a mesh-star mixed on-chip network communication method.

The star / mesh transmission step may include: when the packet in which source routing information is inserted is transmitted from a star IP core, the star switch switching the packet according to a source routing scheme; Receiving a packet transmitted from the star switch, acquiring (X, Y) coordinate information corresponding to a destination address of the packet and inserting the packet into the packet; And transmitting the packet by using the (X, Y) coordinate information inserted into the packet.

The mesh / star transmission may include: when a packet including (X, Y) coordinate information is inserted from a mesh IP core, a mesh switch switching the packet using the (X, Y) coordinate information; Receiving a packet transmitted from the mesh switch, acquiring and inserting source routing information corresponding to a destination address of the packet into the packet; And the star switch forwards the packet using the source routing information inserted into the packet.

As such, the present invention proposes a network system and a communication method for synthesizing a multi-layered on-chip communication structure optimized for communication characteristics of SoC design.

Therefore, in the present invention, the entire network system is not implemented as one on-chip network, but each part is connected to a star on chip network and these are connected to a mesh network to have a hierarchical communication structure.

As a result, the communication performance is improved by up to 20% or more compared with the existing on-chip network method, and the area can be reduced by up to 1/3.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, in describing in detail the operating principle of the preferred embodiment of the present invention, if it is determined that the detailed description of the related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

In addition, the same reference numerals are used for parts having similar functions and functions throughout the drawings.

3A to 3C illustrate a topology of a mesh-star mixed on chip network according to embodiments of the present invention.

First, the topology of FIG. 3A is a structure in which a star network is connected to a mesh network, and not only an IP core (MIP) but also a star switch (SS) constituting a star network may be connected to a mesh switch (MS) constituting a mesh network. have.

This is because the IP core is so large that only one IP core can fit in a tile assigned to one mesh switch in the mesh network, or there is an IP core that is difficult to connect to one star network due to high communication with all IP cores. It is a suitable topology.

3B is a structure in which a communication link with an additional mesh switch is allocated to a star network, and one star switch SS may be commonly connected to two or more mesh switches MS.

This is a suitable topology when there is a large amount of communication from one star network to the outside, or when there is a star network lacking one external link due to multiple communication occurring simultaneously.

In addition, the topology of FIG. 3c is a structure in which a communication link is directly connected between star switches, in which one star switch SS is directly connected not only to the mesh switch MS but also to the star switch SS forming a different star network from the star switch SS. Can be connected.

This is a suitable topology when there is a large amount of communication between adjacent star networks, and reduces communication delay time by performing communication through a communication link directly connected between star networks without a mesh network.

4 is a diagram illustrating the structure of a packet transmitted and received through the mesh-star mixed on-chip network of the present invention.

The mesh-star mixed on-chip network of the present invention supports a packet switching network, and uses the same packet format for efficient mutual communication between the star network and the mesh network.

All packets are transmitted and received in flits, which are the sizes that can be sent in one clock over the communication link, and the size of the flits is usually determined by the trade-off of the area and performance occupied by the communication hardware. In the present invention, it is assumed that the design is 32 bits for ease of explanation.

The packet consists of a header field, an address field, and a data field. The header field contains source routing information, (X, Y) coordinate information, packet type (Read / Write), and packet size information.

In this case, the source routing information is routing information defined according to the source routing method used in the star network, and the (X, Y) coordinate information is routing information defined according to the X-Y routing method used in the mesh network.

That is, in the present invention, the star network may deliver the packet using routing information defined according to the source routing scheme, and the mesh network may deliver the packet using routing information defined according to the XY routing scheme. It supports storing all kinds of routing information.

At this time, the source routing method determines the path from the network interface where the packet is generated to the destination of the packet and sends information of input / output ports to pass through each switch on the path in the header part of the packet. Minimize the required hardware. The star switch (SS) extracts the output port number, which is routing information of the corresponding switch, from the header portion of each packet flowing into the input port and sends the packet to the corresponding output port. In this case, up to 4 levels of source routing information may be stored in the header, and packets may be transmitted through up to 4 levels of star networks.

In the XY routing method, the packet is first sent to the switch having the same X coordinate value as the X coordinate of the destination switch of the packet in the X axis direction, and then the path is set to the switch of the Y coordinate value of the destination switch in the Y axis direction. It has low hardware cost and very good performance. Accordingly, the mesh switch MS calculates an optimal path by applying an X-Y routing algorithm according to the (X, Y) coordinates stored in the packet header, and switches the packet to the destination mesh switch or the destination IP core according to the calculated path.

The address field stores the address of the packet's destination in 32-bit size, and each of the plurality of data fields stores N data in 32-bit size. The size of the data field is specified in the packet size information of the header field.

5 is a diagram illustrating a mesh-star mixed on-chip network system according to an exemplary embodiment of the present invention, which forms a topology in which a star network is connected to a mesh network as in FIG. 3a.

Referring to FIG. 5, a communication system of a mesh-star mixed on-chip network includes star switches 11-1 and 11-2 and star switches 11-1 and 11 that form star networks 10-1 and 10-2. 2) a plurality of star IP cores 12-1 to 12-23 connected to each other, a plurality of mesh switches 21-1 to 21-4 forming a mesh network 20, and each mesh switch 21-4; Mesh IP cores 22-3 and 22-4 connected one by one, and at least one bridge 40-1 and 40- connecting between the star networks 10-1 and 10-2 and the mesh network 20. 2) is provided.

Hereinafter, the function of each component will be described.

When each star switch 11-1 transmits a packet from a star IP core 12-11, 12-12 or a bridge 40-1 connected thereto, the star switch 11-1 analyzes the packet transmission path according to the source routing method. After checking, it forwards the packet entered to the path.

Each star IP core 12-11 packetizes control signals and data through a network interface (not shown) such as a master network interface (MNI) and a slave network interface (SNI), and communicates in a fleet unit. The routing information of the packet is set according to the source routing method.

At this time, the MNI is a network interface connected to a master core such as a processor and a direct memory access controller (DMAC) and requests a communication transaction to the network, and the SNI is an interface connected to a slave core such as a memory and the like. Provide service.

Each mesh switch 21-1, when a packet is transmitted from a different mesh switch 21-1, a mesh IP core 22-3 or a bridge 40-1 connected thereto, analyzes the packet according to the XY routing method. After checking the transmission path of the packet, deliver the packet entered to the path.

Each mesh IP core 22-3 packetizes control signals and data through a network interface such as MNI and SNI like the star IP core 12-11, and communicates in a fleet unit. Set the routing information according to the XY routing method.

Each bridge 40-1 is an interface module for connecting a star network and a mesh network. As illustrated in FIG. 6, a star / mesh converter 41, a mesh / star converter 42, and a routing table 43 are shown. Each of the converters 41 and 42 adds routing information of the network to which the packet is to be transmitted.

More specifically, when a new packet is transmitted from the star network 10-1, the star / mesh converter 41 routes (X, Y) coordinate information on the mesh network corresponding to the destination address of the packet to the routing table. In step (43), it is retrieved, inserted into the packet header, and forwarded to the mesh network (20).

When the packet is transmitted from the mesh network 20, the mesh / star converter 42 receives source routing information from the destination address of the packet transmitted from the mesh network 20 to the destination IP core on the star network connected to the corresponding bridge. Search and obtain from routing table 43. After the source routing information in the packet header is additionally stored or updated through the obtained information, the packet is transmitted to the star network 10-1.

If the packet is transmitted through the mesh network 20 after the packet is generated by the mesh IP core 22-3, only the (X, Y) coordinate information is stored in the packet header, so that the mesh / star converter 42 ) Additionally stores source routing information in the packet header.

On the other hand, if the packet is transmitted through both the star network 10-1 and the mesh network 20 after the packet is generated by the star IP core (for example, 12-11), the (X, Y) coordinate is included in the packet header. Since not only the information but also the source routing information is stored, the mesh / star converter 42 deletes the previously stored source routing information and stores the source routing information obtained by the mesh / star converter 42. That is, the existing source routing information is updated with source routing information to be used for the network to which the packet is to be transmitted.

The routing table 43 acquires and stores source routing information and (X, Y) coordinate information of all communication resources (bridges, star IP cores and switches, and mesh IP cores and switches) located in the mesh-star mixed on-chip network. do.

In the above description, the star switch and the mesh switch are separated from each other and the packets are switched through different routing schemes. The star switch and the mesh switch are configured as one device as shown in FIG. You can also perform the function and the function as a mesh switch at the same time.

7 is a diagram illustrating a configuration of an integrated switch according to an embodiment of the present invention.

Referring to FIG. 7, the integrated switch 50 includes n input units 51-1 to 51-n for controlling each of n input ports, and m arbiters 52- for controlling each of the m output ports. 1-52-m) and n x m crossbar fabrics 53.

Each input unit 51-1 analyzes the packet input through the input port managed by the source port through the source routing algorithm and the XY routing algorithm to determine the output port for outputting the packet, and then identifies the packet with the identified output port. Request to print.

That is, if only the source routing information is stored in the packet header, the output port number to output the input packet is determined through the source routing algorithm. If the packet is transmitted from the bridge and the source routing information and (X, Y) coordinates are stored in the packet header at the same time, the source routing algorithm determines the output port number to output the input packet. . Then, it sends an input request signal including its identifier and packet priority information to the arbiters 52-1 to 52-m controlling the identified output ports.

Each arbiter 52-1 selects a signal having the highest priority among several input request signals that intend to use the output port managed by the arbiter 52-1, and connects the communication path from the input unit where the input port is generated to the output port. Generates a control signal to be transmitted to the crossbar fabric (53). In addition, by transmitting a grant signal to the corresponding input unit, the input unit receiving the grant signal can transmit a packet through a communication path set in the crossbar fabric 53.

The crossbar fabric 53 establishes a communication path between an input unit (or an input port) and an output port in response to a control signal transmitted from the arbiters 52-1 to 52-m.

8 is a diagram for explaining an operation of transmitting a packet of a star IP core by a mesh-star mixed on-chip network system according to the present invention.

First, the star IP core 12-11 located in the star network 10-1 includes source routing information of a destination IP core in a packet header, and then generates a packet having a structure as shown in FIG. 11-1) (S101).

The star switch 11-1 analyzes the packet transmitted from the star IP core 12-11 according to the source routing method (S102).

As a result of analysis, if the destination IP core is a star IP core (eg, 12-12) located in its network, the packet is output to an output port to which the corresponding star IP core 12-12 is connected (S103). On the other hand, if the destination IP core is the star IP core 12-21 located in the different star network 10-2 or the mesh IP core 22-3 located in the mesh network 20, the star switch 11-1 ) Outputs the packet to an output port to which the bridge 40-1 is connected (S104).

After the bridge 40-1 grasps the destination address from the address field of the packet transmitted from the star switch 11-1, additionally stores (X, Y) coordinate information corresponding thereto in the header of the packet (S105). ) It is transmitted to the mesh switch 21-1 connected to it (S106).

The mesh switch 21-1 analyzes the packet transmitted from the bridge 40-1 according to the XY routing method to identify the next mesh switch 21-2 or 21-3 (S107). It transfers this to the switch 21-2 or 21-3 (S108, S110).

Then, the next mesh switch 21-2 or 21-3 analyzes the received packet according to the XY rerouting method and delivers the received packet to the mesh IP core 22-3 connected thereto (S109), or transmits the corresponding packet to itself. It is transmitted to the bridge 40-2 connected to (S111).

If the bridge 40-2 receives the packet, the bridge 40-2 grasps the source routing information from itself to the destination IP core 12-21 from the destination address stored in the address field of the packet to determine the packet header. After updating the source routing information of (S112), the packet is forwarded to the star switch 11-2 connected thereto (S113).

The star switch 11-2 forwards the packet to the destination IP core according to the source routing information updated by the bridge 40-2.

As described above, the mesh-star mixed on-chip network system of the present invention connects a star network to a mesh network so that IP cores located in different networks can communicate with each other.

9 is a diagram for explaining an operation of transmitting a packet of a mesh IP core in a mesh-star mixed on-chip network system according to the present invention.

First, the mesh IP core 22-3 includes (X, Y) coordinate information of the destination IP core in a packet header, and then generates a packet having a structure as shown in FIG. 4 and transmits the packet to the mesh switch 21-3. (S201).

The mesh switch 21-3 analyzes the packet transmitted from the mesh IP core 22-3 according to the X and Y routing room (that is, by analyzing the (X, Y) coordinate information) and then the next unit mesh switch ( 21-3) is identified (S202), and transferred to the mesh switch 21-3 (S203).

Then, the mesh switch 21-1 of the next stage analyzes the received packet according to the X-Y rerouting method and transfers the received packet to the bridge 40-1 connected to the packet (S204).

The bridge 40-1 grasps the source routing information from itself to the destination IP core 12-21 from the destination address stored in the address field of the packet transmitted from the mesh switch 21-1, and thus source routing information of the packet header. After the additional storage (S112), and transmits the packet to the star switch (11-1) connected to it (S113).

The star switch 11-1 transfers the packet to the star IP core 12-11, which is a destination IP core, according to the source routing information additionally stored by the bridge 40-1.

As described above, the mesh-star mixed on-chip network system of the present invention connects a star network to a mesh network so that IP cores located in different networks can communicate with each other.

In particular, since the mesh-star mixed on-chip network system according to an embodiment of the present invention allocates one mesh IP core to one mesh switch, the size of the mesh-star mixed on-chip network system is too large or the amount of communication with all the IP cores is high. It is suitable when there is an IP core that is difficult to connect to the network.

If necessary, as shown in FIG. 10, two or more bridges may be connected to one star switch to form a topology as shown in FIG. 3B, and thus more mesh switches may be allocated to one star network. That is, it is possible to increase the communication capacity of a particular star network.

In addition, as shown in FIG. 11, a predetermined star switch is directly connected to form a topology as shown in FIG. 3C. Accordingly, a large amount of communication with each other enables star networks to communicate directly without passing through a mesh network. In other words, communication delay between star networks can be reduced.

The present invention described above is not limited to the above-described embodiments and the accompanying drawings, and it is common in the art that various substitutions, modifications, and changes can be made without departing from the technical spirit of the present invention. It will be apparent to those skilled in the art.

1 is a diagram illustrating a topology of a star type on-chip network according to the prior art;

2 is a diagram illustrating a topology of a mesh type on-chip network according to the prior art;

3A to 3C illustrate a topology of a mesh-star mixed on chip network according to embodiments of the present invention;

4 is a diagram showing the structure of a packet transmitted and received via the mesh-star mixed on-chip network of the present invention;

5 illustrates a mesh-star mixed on-chip network system according to an embodiment of the present invention;

6 is a view showing the configuration of a bridge according to an embodiment of the present invention;

7 is a view showing the configuration of an integrated switch according to an embodiment of the present invention;

8 is a diagram for explaining an operation of transmitting a packet of a star IP core by a mesh-star mixed on-chip network system according to the present invention;

9 is a view for explaining an operation of transmitting a packet of a mesh IP core in a mesh-star mixed on-chip network system of the present invention;

10 illustrates a mesh-star mixed on-chip network system according to another preferred embodiment of the present invention; and

11 illustrates a mesh-star mixed on-chip network system according to another preferred embodiment of the present invention.

Claims (15)

A star switch forming a star network and switching packets of the star IP core according to a source routing scheme; A mesh switch forming a mesh network and switching packets of the mesh IP core according to an X-Y routing scheme; And Connected between the star switch and the mesh switch, the routing information according to the XY routing method is inserted into the packet transmitted from the star switch and then transferred to the mesh switch, and the source routing is transmitted to the packet transmitted from the mesh switch. Mesh-star mixed on-chip network system including a bridge for inserting the routing information according to the scheme and forwarded to the star switch. The method of claim 1, wherein the packet is A header field that contains source routing information, (X, Y) coordinate information, type of packet (Read / Write), and packet size information; An address field containing a destination address; And A mesh-star mixed on-chip network system comprising a plurality of data fields containing data. The method of claim 2, wherein the packet is A mesh-star mixed on-chip network system comprising a header field, an address field, and a data field, and is transmitted in a fleet unit. The method of claim 1, wherein the bridge A routing table for managing source routing information and (X, Y) coordinate information for each of a different bridge, the star IP core, the star switch, the mesh IP core, and the mesh switch; A star / mesh converter that acquires (X, Y) coordinate information corresponding to the packet from the routing table when the packet is transmitted from the star switch, adds it to the packet, and then transfers the packet; And When the packet is transmitted from the mesh switch, the mesh-star mixed on-chip network system comprising a mesh / star converter to obtain the source routing information corresponding to the packet from the routing table to add the packet to the packet and then transfer the packet to the packet; . The method of claim 1, wherein the star switch A plurality of input units for analyzing the packets input through the plurality of input ports through a source routing method to identify an output port for outputting the packets, and then requesting to output the packets to the identified output ports; A plurality of arbiters for generating a control signal for connecting a communication path between an input port and an output port in response to a request of each of the plurality of input units; And And a crossbar fabric configured to establish a communication path between an input port and an output port in response to the control signal. The method of claim 1, wherein the mesh switch A plurality of input units for analyzing the packets input through the plurality of input ports through an X-Y routing method to determine an output port for outputting the packets, and then requesting to output the packets to the identified output ports; A plurality of arbiters for generating a control signal for connecting a communication path between an input port and an output port in response to a request of each of the plurality of input units; And And a crossbar fabric configured to establish a communication path between an input port and an output port in response to the control signal. The method of claim 1, wherein the star switch and the mesh switch A plurality of input units for analyzing the packets input through the plurality of input ports through a source routing method or an X-Y routing method to identify an output port for outputting the packets, and then requesting to output the packets to the identified output ports; A plurality of arbiters for generating a control signal for connecting a communication path between an input port and an output port in response to a request of each of the plurality of input units; And And a crossbar fabric configured to establish a communication path between an input port and an output port in response to the control signal. The method of claim 7, wherein each of the plurality of input unit is If only the source routing information is stored in the packet, the output port number is identified by the source routing method, and if only the (X, Y) coordinates are stored, the output port number is identified by the XY routing method, and the source routing information and the (X, Y) A mesh-star mixed on-chip network system, characterized in that if the coordinates are stored at the same time, an output port number for outputting an input packet is determined through a source routing algorithm. The method of claim 1, wherein the star IP core is Mesh-star mixed on-chip network system, characterized in that for setting the routing information set according to the source routing method in the packet, and then transmitting and receiving the packet in fleet units. The method of claim 1 wherein the mesh IP core is Mesh-star mixed on-chip network system, characterized in that for setting the routing information set according to the X-Y routing method in the packet, and then transmitting and receiving the packet in fleet units. The method of claim 1, wherein the star switch Mesh-star mixed on-chip network system, characterized in that it can be connected to two or more bridges. The method of claim 1, wherein the star switch A mesh-star mixed on-chip network system, which can be connected to a star switch different from itself. (X, y) coordinate information is inserted into a packet transmitted from a star network and transmitted to a mesh network, and the mesh network transfers the packet to a destination IP core using the (x, y) coordinate information. Mesh transmission step; And A mesh including a mesh / star transmission step of inserting source routing information into the packet transmitted from the mesh network and transmitting the packet to the star network, wherein the star network delivers the packet to a destination IP core through the source routing information. -Star mixed on-chip network communication method. The method of claim 13, wherein the star / mesh transfer step If a packet including source routing information is transmitted from a star IP core, the star switch switching the packet according to a source routing scheme; Receiving a packet transmitted from the star switch, acquiring (X, Y) coordinate information corresponding to a destination address of the packet and inserting the packet into the packet; And And transmitting the packet using the (X, Y) coordinate information inserted into the packet. The method of claim 13, wherein the mesh / star transmission step If a packet in which (X, Y) coordinate information is inserted is transmitted from a mesh IP core, a mesh switch switching the packet using the (X, Y) coordinate information; Receiving a packet transmitted from the mesh switch, acquiring and inserting source routing information corresponding to a destination address of the packet into the packet; And And the star switch includes forwarding the packet using source routing information inserted into the packet.

Images