Parallel broadcast communication method of Mesh structure wavelength division multiplexing on-chip optical network

Abstract

The invention provides a parallel broadcast communication method of a Mesh structure wavelength division multiplexing on-chip optical network, which comprises the following steps of: firstly, dividing an n * m mesh structure into four blocks, including two row blocks, nodes of which communicate in rows, and each row being a linear array with (or) nodes; there are two column blocks whose nodes communicate in columns, each column being a linear array with (or) nodes. Then, carrying out parallel communication on each row or column linear array in the block by adopting a bit algorithm; the bit algorithm comprises the following steps: firstly, carrying out lambda + 1 system coding on nodes in a linear array, wherein lambda is an available wavelength number; then multiple rounds of communication are carried out according to the coded bits in sequence, a target node set corresponding to the source node is determined according to values on the coded bits in each round of communication, and wavelengths are distributed according to the values on the coded bits; and finally, carrying out conflict-free parallel communication according to the formulated path and wavelength allocation rule.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of parallel broadcast communication technology, and particularly relates to a parallel broadcast communication method for a Mesh structure wavelength division multiplexing optical network-on-chip. BACKGROUND

[0002] In the traditional network-on-chip, an electrical interconnection method is adopted, which is referred to as an electrical network-on-chip (ENoC). With the increase in the number of on-chip processors and the improvement in the computing speed, the problems such as the limitation of the integration level on the chip, the large electrical signal interference, the large power consumption, and the difficulty in heat dissipation gradually become the bottleneck restricting the performance improvement of the electrical interconnection network. In order to break through the bottleneck of the traditional electrical interconnection network, scholars have proposed a solution of an optical network-on-chip (ONoC). In the ONoC using the WDM technology, the bandwidth of the optical waveguide is divided into multiple wavelengths, and then optical channels are established using different wavelengths. The multiple data streams are simultaneously transmitted in the same optical waveguide by using the optical channels, which enables the network-on-chip to simultaneously process multiple data streams, greatly improving the communication efficiency and the data transmission capacity. The wavelength division multiplexing enables a single physical link to simultaneously transmit multiple independent channels, which provides a parallel basis for the on-chip broadcast communication, especially in the Mesh topology structure. The regular node layout of the Mesh topology structure further enhances the potential of improving the parallel efficiency by using the wavelength resources.

[0003] The wavelength division multiplexing technology provides multiple wavelength parallel channels for parallel transmission, but most communication algorithms still use the dynamic routing logic in the electrical network, and then use a dynamic arbitration mechanism to compete for the wavelength resources. This process not only causes extra energy consumption, but also weakens the parallel gain of the wavelength division multiplexing due to the arbitration delay. Based on the parallel broadcast communication algorithm of the wavelength division multiplexing, the wavelength-path joint coding mechanism can enable the physical link to realize the conflict-free concurrent transmission relying on the wavelength isolation characteristics. By analyzing the symmetry of the Mesh topology and the characteristics of the broadcast communication, the algorithm pre-establishes a static wavelength-path mapping rule, and plans the communication path and the wavelength allocation at the beginning of the broadcast communication. Such a static routing and wavelength allocation mechanism avoids the wavelength arbitration and conversion hop by hop, and the node only needs to perform the low-complexity wavelength allocation and data transmission according to the preset rule. Therefore, the present application provides a parallel broadcast communication method for a Mesh structure wavelength division multiplexing optical network-on-chip to solve the above problems. SUMMARY

[0004] The present application aims at solving the problems in the prior art, and provides a parallel broadcast communication method for a Mesh structure wavelength division multiplexing optical network-on-chip.

[0005] In order to achieve the above object, the present application adopts the following technical scheme:

[0006] A parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network, comprising the following steps:

[0007] S1: dividing an n*m mesh structure on-chip optical network into four blocks, i.e. two row blocks and two column blocks;

[0008] S2: respectively performing natural numbering on each row and column after the division, so that each row and column becomes a linear array topology;

[0009] S3: respectively performing parallel communication on each row and column by calling a bitwise parallel broadcast communication algorithm Bitwise of a linear array;

[0010] The parallel broadcast communication algorithm Bitwise comprises the following steps:

[0011] S3.1: encoding nodes in the linear array by using Λ+1 base, Λ being the number of available wavelengths, and the set S i representing a node set receiving data after the end of the i-th round of communication, the initial source node being (0.......0);

[0012] S3.2: determining a communication path, the source node set being the node set having received data, and the target node set being determined according to a specific rule of the encoding bits;

[0013] S3.3: allocating a wavelength to the path if the path exists;

[0014] S3.4: performing parallel communication and updating the set S

[0015] S3.5: updating the value of i, and repeating steps 3.2 to 3.4 until all nodes receive data.

[0016] Preferably, the division mode is described as follows:

[0017] Row block one, containing nodes in the first row to the (n-1)-th row satisfying the following condition: wherein wherein

[0018] Row block two, containing nodes in the second row to the n-th row satisfying the following condition: wherein wherein

[0019] Column block one, containing nodes in the first column to the (m-1)-th column satisfying the following condition: wherein wherein

[0020] Column block two, containing nodes in the second column to the m-th column satisfying the following condition: The nodes in the column to the (m-1)th column satisfying the following conditions Wherein

[0021] Preferably, the specific process of calling "bit algorithm" for parallel communication in S3 is as follows:

[0022] Initialization, the source node is (0, 0);

[0023] Call the bitwise algorithm Bitwise (m, A) for communication in the 0th row; call the bitwise algorithm Bitwise (n, A) for communication in the 0th column;

[0024] When the ith node in the 0th column receives data, the row where the node is located starts to call the bitwise algorithm for communication, wherein the node parameter value is the number of nodes in each row in the row-oriented block one and the row-oriented block two in step 2;

[0025] When the jth node in the 0th row receives data, the column where the node is located starts to call the bitwise algorithm for communication, wherein the node parameter value is the number of nodes in each column in the column-oriented block one and the column-oriented block two in step 2.

[0026] Preferably, determining the communication path includes that the source node set is the set of nodes that have received data, denoted as S i ;

[0027] The source node (s0s1...s i-1 0s i+1 ...s k )∈S i The nodes in the corresponding target node set D i are selected according to the following requirements: first, the values of the 0th bit to the (i-1)th bit of the target node code (d0...d i-1 ld i+1 ...d k ) should be the same as the values on the corresponding bits of the source node, and the values of the (i+1)th bit to the (l)th bit should be 0;

[0028] Second, the value of the lth bit is a positive integer greater than or equal to 1 and less than or equal to A; finally, all the nodes in the vertex set satisfying the above conditions constitute the target node set;

[0029] The communication path from each source node (s0s1...s i-1 0s i+1 ...s k )∈S i to the target node (d0...d i-1 ld i+1 ...d k )∈D i can be represented as P l= < (s0s1...s i-1 0s i+1 ...s k ), (d0...d i-1 ld i+1 ...d k ) >.

[0030] Preferably, the assigned wavelength comprises:

[0031] If (s0s1...s i-1 0s i+1 ...s k ) e V and (d0...d i-1 ld i+1 ...d k ) e V, then the path exists and is assigned a wavelength λ l .

[0032] Preferably, the parallel communication update set comprises:

[0033] The communication path P l communicates simultaneously using the assigned wavelength, and after the communication is over, the set F i+1 is updated to F i ∪ D i .

[0034] Preferably, S3.5 comprises updating the value of i (from 0 to k) and repeating the above steps 2 to 4 until all nodes have received the data, i.e., |F i+1 | = n.

[0035] Preferably, the source node set is S i ;

[0036] S i = { (s0s1...s i-1 0s i+1 ...s k ) | (s0s1...s i-1 0s i+1 ...s k ) e F i and (s i+1 ...s k ) = (0...0)}.

[0037] Preferably, the destination node set is D i ;

[0038] D i = { (d0...d i-1 ld i+1 ...d k ) e V | (d0...d i-1) = (s0s1...s i-1 ) and (d i+1 ...d k ) = (0..0), 1≤l≤Λ}.

[0039] Advantages of the present application:

[0040] 1. Improve communication efficiency: By dividing the Mesh structure into four sub-grid blocks, the block parallel broadcast communication is realized, and data transmission can be carried out simultaneously in each sub-grid block, greatly shortening the time of overall broadcast communication. At the same time, by using wavelength division multiplexing technology (WDM), a single physical link can transmit multiple independent channels at the same time, providing multiple wavelength parallel channels for parallel transmission, further improving the efficiency and capacity of data transmission.

[0041] 2. Reduce energy consumption and delay: Using a static wavelength-path mapping rule, the communication path and wavelength allocation are planned at the beginning of broadcast communication, avoiding the wavelength arbitration and conversion process in traditional dynamic routing logic hop by hop. This not only reduces the extra energy consumption caused by the dynamic arbitration mechanism competing for wavelength resources, but also eliminates the arbitration delay, fully utilizing the parallel gain of wavelength division multiplexing.

[0042] 3. Reduce transmission conflict: By using "bit algorithm" to encode the nodes in the linear array, and determining the target node set and allocating wavelength according to the encoding bit rule, combined with the wavelength isolation feature, the physical link is realized without conflict concurrent transmission. In the sub-grid structure after blocking, the source node is located at the top position, and the communication path and wavelength allocation in each sub-grid block are relatively independent, further reducing the transmission conflict between different blocks.

[0043] 4. Reduce node complexity: The node only needs to allocate wavelength and transmit data with low complexity according to the preset wavelength-path mapping rule, without complex dynamic routing calculation and arbitration logic, simplifying the design and implementation of the node, which is conducive to reducing the overall complexity and cost of the optical network on chip.

[0044] 5. Adapt to the characteristics of Mesh topology: Make full use of the node layout characteristics of Mesh topology structure rule, and transform the large-scale Mesh grid into small-scale sub-grid through blocking strategy, so that the parallel broadcast communication is more suitable for the symmetry and regularity of Mesh structure, and the potential of improving parallel efficiency by using wavelength resources is enhanced, especially suitable for the scene of optical network on chip with a large number of processors. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is an 8x8 grid topology structure example of the present application;

[0046] Figure 2 is the row direction block after blocking and its ternary encoding;

[0047] Figure 3 is the column direction block after being divided and its 3-ary coding;

[0048] Figure 4 is the route and wavelength allocation of the first round of communication on the 8x8 mesh topology;

[0049] Figure 5 is the route and wavelength allocation of the second round of communication on the 8x8 mesh topology;

[0050] Figure 6 is the route and wavelength allocation of the third round of communication on the 8x8 mesh topology;

[0051] Figure 7 is the route and wavelength allocation of the fourth round of communication on the 8x8 mesh topology;

[0052] Figure 8 is the flow chart of the present application. DETAILED DESCRIPTION

[0053] The following will be described in detail with reference to the accompanying drawings. Figures 1-8 The present application will be further described in detail.

[0054] The embodiment of the present application discloses a parallel broadcast communication method of a mesh structure wavelength division multiplexing optical network on chip.

[0055] Embodiment one

[0056] Referring to Figure 8 A parallel broadcast communication method of a mesh structure wavelength division multiplexing optical network on chip, comprising the following steps:

[0057] S1: Dividing the n x m mesh structure optical network on chip into four blocks, i.e. two row direction blocks and two column direction blocks;

[0058] S2: Carrying out natural numbering on each row and column after being divided, so that each row and column becomes a linear array topology;

[0059] S3: Calling the bitwise parallel broadcast communication algorithm Bitwise of the linear array to carry out parallel communication on each row and column respectively;

[0060] The parallel broadcast communication algorithm Bitwise comprises the following steps:

[0061] S3.1: Coding the nodes in the linear array by using Λ+1-ary, Λ being the number of available wavelengths, and the set S i Indicates the node set receiving data after the end of the i-th round of communication, and the initial source node is (0.......0);

[0062] S3.2: Determine the communication path, the source node set is the node set that has received data, and the target node set is determined according to a specific rule of the encoding bits;

[0063] S3.3: Assign a wavelength to the path if it exists;

[0064] S3.4: Communicate in parallel and update the set;

[0065] S3.5: Update the value of i, repeat steps 3.2 to 3.4 until all nodes have received data.

[0066] In this embodiment, the block mode is described as follows:

[0067] Row-oriented block one, containing nodes are nodes in the first row to the nth-1 row that satisfy the following conditions: Nodes in the row that satisfy the following conditions: Wherein

[0068] Row-oriented block two, containing nodes are nodes in the first row to the nth-1 row that satisfy the following conditions: Nodes in the row that satisfy the following conditions: Wherein

[0069] Column-oriented block one, containing nodes are nodes in the first column to the mth-1 column that satisfy the following conditions: Nodes in the column that satisfy the following conditions: Wherein

[0070] Column-oriented block two, containing nodes are nodes in the first column to the mth-1 column that satisfy the following conditions: Nodes in the column that satisfy the following conditions: Wherein

[0071] In this embodiment, the specific process of calling "bit algorithm" for parallel communication in S3 is as follows:

[0072] Initialization, the source node is (0, 0);

[0073] Call the bitwise algorithm Bitwise(m, Λ) in the 0th row for communication; call the bitwise algorithm Bitwise(n, Λ) in the 0th column for communication;

[0074] When the ith node in the 0th column receives data, the row where the node is located starts to call the bitwise algorithm for communication, wherein the node parameter value is the number of nodes in each row in the row-oriented block one and the row-oriented block two in step 2;

[0075] When the jth node in the 0th row receives data, the column where the node is located starts to call the bitwise algorithm for communication, wherein the node parameter value is the number of nodes in each column in the column-oriented block one and the column-oriented block two in step 2.

[0076] In this embodiment, the determination of the communication path includes that the source node set is the set of nodes that have received data, denoted as S i ;

[0077] The source node (s0s1...s i-1 0s i+1 ...s k )∈S i The node selection requirement in the corresponding target node set D i is as follows: first, the value of the 0th to the i-1th bit of the target node code (d0...d i-1 ld i+1 ...d k ) should be the same as the value on the corresponding bit of the source node, and the value of the i+1th to the bit should be 0;

[0078] Second, the value of the bit is a positive integer greater than or equal to 1 and less than or equal to Λ; finally, all nodes in the vertex set that satisfy the above conditions constitute the target node set;

[0079] The communication path from each source node (s0s1...s i-1 0s i+1 ...s k )∈S i to the target node (d0...d i-1 ld i+1 ...d k )∈D i can be represented as P l =<(s0s1...s i-1 0s i+1 ...s k ),(d0...d i-1 ld i+1 ...d k )>.

[0080] In this embodiment, the wavelength allocation includes:

[0081] If (s0s1...s i-1 0s i+1 ...s k )∈V and (d0...d i-1 ld i+1 ...d k )∈V, then the path exists, and the wavelength λ l is allocated to the path.

[0082] In this embodiment, the parallel communication update set includes:

[0083] The communication path P lSimultaneously communicate using the assigned wavelengths, and update the set F after the communication is over i+1 = F i ∪ D i .

[0084] In this embodiment, S3.5 includes updating the value of i (from 0 to k), repeating the above steps 2 to 4 until all nodes have received the data, i.e. |F i+1 | = n.

[0085] In this embodiment, the source node set is S i ;

[0086] S i = {(s0s1...s i-1 0s i+1 ...s k ) | (s0s1...s i-1 0s i+1 ...s k ) ∈ F i and (s i+1 ...s k ) = (0...0)}.

[0087] In this embodiment, the target node set D i ;

[0088] D i = {(d0...d i-1 ld i+1 ...d k ) ∈ V | (d0...d i-1 ) = (s0s1...s i-1 ) and (d i+1 ...d k ) = (0..0), 1≤l≤Λ}.

[0089] Embodiment Two

[0090] Referring to Figures 1-7 , a parallel broadcast communication method for a Mesh structure wavelength division multiplexing optical network on a chip, the optical network on a chip is an 8x8 mesh structure, and the number of available wavelengths is 2, as shown in Figure 1 , which shows an 8x8 mesh structure, wherein the numbers in the nodes are the natural numbers of the processors, so the natural numbers of the nodes are 0-63, and the numbers in the brackets on the nodes are the row and column coordinates of the processors, i.e. the X-axis and Y-axis coordinates, so the source node can be represented as (0, 0); since the number of available wavelengths is 2, the available wavelength set can be represented as {λ1, λ2};

[0091] Step 1: divide the mesh structure optical network on a chip M 8*8The nodes of the matrix are divided into four blocks, two of which are row direction blocks as shown in Figure 2 and two of which are column direction blocks as shown in Figure 3 .

[0092] Row direction block one, the nodes contained are the node set {(i, y) | y ∈ [0, 1, 2, 3], 1 ≤ i ≤ 3} from the first row to the third row;

[0093] Row direction block two, the nodes contained are the node set {(i, y) | y ∈ [0, 1, 2, 3], 4 ≤ i ≤ 7} from the fourth row to the seventh row;

[0094] Column direction block one, the nodes contained are the node set {(x, j) | x ∈ [0, 1, 2, 3], 1 ≤ j ≤ 3} from the first column to the third column;

[0095] Column direction block two, the nodes contained are the node set {(x, j) | x ∈ [0, 4, 5, 6, 7], 4 ≤ j ≤ 7} from the fourth column to the seventh column;

[0096] Step 2: Each row and column after the division is numbered naturally, and there are the following six numbering conditions.

[0097] Row 0, there are 8 nodes, numbered 0 to 7;

[0098] Row direction block one, rows 1 to 3, each row has 5 nodes, and each row is numbered 0 to 4, and the node set is represented as R1;

[0099] Row direction block two, from row 4 to row 7, each row has 4 nodes, and each row is numbered 0 to 3, and the node set is represented as R2;

[0100] Column 0, there are 8 nodes, numbered 0 to 7;

[0101] Column direction block one, from column 1 to column 3, each column has 4 nodes, and each column is numbered 0 to 3, and the node set is represented as C1;

[0102] Column direction block two, from column 4 to column 7, each column has 5 nodes, and each column is numbered 0 to 4, and the node set is represented as C2;

[0103] Step 3: Call the bitwise parallel broadcast communication method of the linear array to perform parallel communication on each row and column.

[0104] Initialization, the source node is (0, 0)

[0105] Call the bitwise algorithm Bitwise (8, 2) to communicate in row 0; call the bitwise algorithm Bitwise (8, 2) to communicate in column 0;

[0106] When the ith node in the 0th column receives data, if 1≤i≤3, the bitwise algorithm Bitwise(5, 2) is called; if 4≤i≤7, the bitwise algorithm Bitwise(4, 2) is called;

[0107] When the jth node in the 0th row receives data, if 1≤j≤3, the bitwise algorithm Bitwise(4, 2) is called; if 4≤j≤7, the bitwise algorithm Bitwise(5, 2) is called.

[0108] The implementation steps of the bitwise algorithm Bitwise are described as follows:

[0109] Step 3.1: The encoded rows and columns of the divided blocks are encoded in ternary, and there are three encoding cases as follows.

[0110] The encoding of the 0th column and the 0th row is {00, 01, 02, 10, 11, 12, 20, 21}, as shown in Figure 2 and Figure 3 .

[0111] In the row-to-block one, the encoding of each row is {00, 01, 02, 10, 11}; in the row-to-block two, the encoding of each row is {00, 01, 02, 10}, as shown in Figure 2 .

[0112] In the column-to-block one, the encoding of each column is {00, 01, 02, 10}; in the column-to-block two, the encoding of each column is {00, 01, 02, 10, 11}, as shown in Figure 3 .

[0113] Step 3.2: Determine the path, assign the wavelength, and update the iteration

[0114] Initialization, source node is (0, 0)

[0115] The first round of communication (as shown in Figure 4 .

[0116] The communication steps in the 0th row and the 0th column are as follows:

[0117] i takes the value of 0, the source node set is S0={(00)}, and the target node set is D0={(10), (20)}

[0118] Thus, there are two paths as follows

[0119] P1=<<(00), (10)>, and the wavelength λ1 is assigned;

[0120] P2=<<(00), (20)>, and the wavelength λ2 is assigned.

[0121] The above paths communicate in parallel, and the set F0 is updated to {(00), (10), (20)} and i is updated to 1.

[0122] The communication in the 3rd and 6th rows and columns is started.

[0123] The second round of communication (as shown in Figure 5

[0124] The communication steps in the 0th row and 0th column are as follows:

[0125] i is 1, the source node set is S0 = {(00), (10), (20)}, and the following two paths exist

[0126] The target node set corresponding to the source node (00) is {(01), (02)}, and the following two paths exist

[0127] P1 = <(00), (01)> and wavelength λ1 is assigned

[0128] P2 = <(00), (02)> and wavelength λ2 is assigned.

[0129] The target node set corresponding to the source node (10) is {(11), (12)}, and the following two paths exist

[0130] P1 = <(10), (11)> and wavelength λ1 is assigned

[0131] P2 = <(10), (12)> and wavelength λ2 is assigned.

[0132] The target node set corresponding to the source node (20) is {(21)}, and the following one path exists

[0133] P1 = <(20), (21)> and wavelength λ1 is assigned

[0134] The above paths communicate in parallel, and the set F0 is updated.

[0135] F0 = {(00), (01), (02), (10), (11), (12), (20), (21)}, since |F0| = 8, i.e., all nodes in these rows and columns have received data, the communication in these rows and columns is ended.

[0136] The communication in the remaining rows and columns (i.e., the 1st, 2nd, 4th, 5th, and 7th rows and columns) is started.

[0137] The communication steps in the 3rd and 6th rows and columns are as follows:

[0138] i is 0, the source node set is S0 = {(00)}, the target node set is D0 = {(10)}, and the following one path exists

[0139] ​P1 = < (00), (10) >, assign wavelength λ1;

[0140] The above paths communicate in parallel, and the set F0 = {(00), (10)} is updated, and i = 1;

[0141] The third round of communication (as shown in Figure 6

[0142] The communication steps in the 3rd, 6th row and 3rd, 6th column are as follows:

[0143] i = 1, the source node set S0 = {(00), (10)}, then

[0144] The target node set corresponding to the source node (00) is {(01), (02)}, so there are the following two paths

[0145] P1 = < (00), (01) >, assign wavelength λ1;

[0146] P2 = < (00), (02) >, assign wavelength λ2.

[0147] Only in the 3rd row and 6th column source node (10) has a corresponding target node set {(11)}, so there is the following path

[0148] P1 = < (10), (11) >, assign wavelength λ1.

[0149] The above paths communicate in parallel, and the set F0 = {(00), (01), (02), (10), (11)} is updated in the 3rd row and 6th column

[0150]

[0151] Since |F0| = 5, i.e. all nodes on these rows and columns have received the end of data communication.

[0152] The set F0 = {(00), (01), (02), (10)} is updated in the 6th row and 3rd column

[0153]

[0154] Since |F0| = 4, i.e. all nodes on these rows and columns have received the end of data communication.

[0155] The communication steps in the remaining rows and columns (i.e. 1st, 2nd, 4th, 5th, 7th rows and columns) are as follows:

[0156] i = 0, the source node set S0 = {(00)}, the target node set D0 = {(10)}, so there is the following path

[0157] P1 = < (00), (10) >, assign wavelength λ1; ​​​

[0158] The above paths communicate in parallel, and the set F0 is updated to {(00), (10)}, and i is updated to 1;

[0159] The fourth round of communication (as shown in Figure 7

[0160] The communication steps in the first, second, fourth, fifth, and seventh rows and columns are as follows:

[0161] i is 1, and the source node set is S0 = {(00), (10)}, so

[0162] The target node set corresponding to the source node (00) is {(01), (02)}, so there are the following two paths

[0163] P1 = <(00), (01)> and wavelength λ1 is assigned;

[0164] P2 = <(00), (02)> and wavelength λ2 is assigned.

[0165] Only in the first and second rows and the fourth, fifth, and seventh columns, the source node (10) has a corresponding target node set {(11)}, so there is the following path

[0166] P1 = <(10), (11)> and wavelength λ1 is assigned.

[0167] The above paths communicate in parallel, and the set F0 is updated to {(00), (01), (02), (10), (11)} in the first, second, fourth, fifth, and seventh rows and columns.

[0168]

[0169] Since |F0| = 5, all nodes on these rows and columns have received the end of data communication.

[0170] The set F0 is updated to {(00), (01), (02), (10)} in the fourth, fifth, and seventh rows and the first and second columns.

[0171]

[0172] Since |F0| = 4, all nodes on these rows and columns have received the end of data communication.

[0173] The above are preferred embodiments of the present application, and are not intended to limit the protection scope of the present application, so: any equivalent changes made according to the structure, shape, principle of the present application should be covered within the protection scope of the present application.​​​

Claims

1. A parallel broadcast communication method for a mesh-structured wavelength division multiplexing on-chip optical network, characterized in that: The following steps are involved: S1: Divide the n×m mesh structure on-chip optical network into four blocks, namely two row blocks and two column blocks; S2: naturally number each row and column after the blocks, so that each row and column has a linear array topology; S3: Call the Bitwise bitwise parallel broadcast communication algorithm of the linear array to perform parallel communication on each row and column; The Bitwise parallel broadcast communication algorithm includes the following steps: S3.1: Use Λ+1 base to encode the nodes in the linear array, where Λ is the number of available wavelengths, and the set S i It represents the set of nodes that receive data after the end of round i of communication, and the initial source node is (0.......0); S3.2: Determine the communication path. The source node set is the set of nodes that have received the data, and the target node set is determined according to the specific rules of the coding bit; S3.3: Allocate wavelength. If the path exists, allocate the corresponding wavelength to it. S3.4: Communicate and update collections in parallel; S3.5: Update the value of i and repeat steps 3.2 to 3.4 until all nodes receive the data.

2. The parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: The block division method is described as follows: Row-wise block 1, containing nodes from row 1 to row Nodes in the row that meet the following conditions in Row to block 2, which contains nodes from Go to the node in row n-1 that satisfies the following conditions in Column-wise block 1, containing nodes from column 1 to column Nodes in the column that meet the following conditions in Column-wise block 2, containing nodes from Column m-1 to the node that satisfies the following conditions in 3. The parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: The specific process of calling the "bit algorithm" for parallel communication in S3 is: Initialize, the source node is (0, 0); Call the bitwise algorithm Bitwise(m,Λ) for communication in row 0; call the bitwise algorithm Bitwise(n,Λ) for communication in column 0; When the i-th node in column 0 receives data, the row where the node is located starts to call the bit algorithm for communication, where the node parameter value is the number of nodes in each row of row-toward block 1 and row-toward block 2 in step 2; When the jth node in row 0 receives data, the column where the node is located starts to call the bit algorithm for communication, where the node parameter value is the number of nodes in each column in column-toward block 1 and column-toward block 2 in step 2.

4. The parallel broadcast communication method for a mesh-structured wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: Determine the communication path including the source node set as the set of nodes that have received the data, denoted as S i ; Source node (s0s1...s i-1 0s i+1 ...s k )∈S i The corresponding target node set D i The node selection requirements in are as follows: first, the target node code (d0...d i-1 ld i+1 ...d k The values ​​of bits 0 to i-1 of the node should be the same as those of the corresponding bits of the source node, and the values ​​of bits i+1 to i-1 should be 0; The value of the next position is a positive integer greater than or equal to 1 and less than or equal to Λ; finally, all the nodes in the vertex set that meet the above conditions constitute the target node set; Each source node (s0s1...s i-1 0s i+1 ...s k )∈S i To the target node (d0...d i-1 ld i+1 ...d k )∈D i The communication path can be expressed as P l =<(s0s1...s i-1 0s i+1 ...s k ),(d0...d i-1 ld i+1 ...d k )>.

5. The parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: Assigned wavelengths include: If (s0s1...s i-1 0s i+1 ...s k )∈V and (d0...d i-1 ld i+1 ...d k )∈V, then the path exists and a wavelength λ is assigned to the path. l .

6. The parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: The set of parallel communication updates includes: Communication path P l Use the allocated wavelengths to communicate simultaneously, and update the set F after the communication is completed i+1 =F i ∪D i .

7. The parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: S3.5 includes updating the value of i (from 0 to k), repeating steps 2 to 4 above until all nodes have received the data, i.e. |F i+1 |=n.

8. The parallel broadcast communication method for a mesh-structured wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: The source node set is S i ; S i ={(s0s1...s i-1 0s i+1 ...s k )|(s0s1...s i-1 0s i+1 ...s k )∈F i And (s i+1 ...s k )=(0...0)}.

9. The parallel broadcast communication method of a mesh structure wavelength division multiplexing on-chip optical network according to claim 1, characterized in that: Target node set D i ; D i = {(d0...d i-1 ld i+1 ...d k ) ∈ V | (d0...d i-1 ) = (s0s1...s i-1 ) and (d i+1 ...d k ) = (0..0), 1 ≤ l ≤ Λ}.

Images