Iterative model building method and simulation method for cross-point storage array

Through the iterative model and voltage update method, the problem of high time complexity in cross-point storage array simulation is solved, the simulation efficiency is improved and the overhead is reduced, and accurate current and voltage relationship simulation is achieved.

CN114818573BActive Publication Date: 2024-11-12HUAZHONG UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210473740.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-29
Publication Date
2024-11-12
Estimated Expiration
2042-04-29

AI Technical Summary

Technical Problem

The simulation methods of existing intersection storage arrays have high time complexity and spatial complexity, resulting in low simulation efficiency and high overhead.

Method used

The iterative model establishment method is adopted, and based on Kirchoff's voltage law, a model of the intersection storage array is established through iterative sequences, and the node voltage and current are updated during the simulation process to reduce the time complexity and space complexity.

Benefits of technology

It effectively reduces the time and space complexity of the simulation cross-point array, improves the simulation efficiency and reduces the simulation overhead, and realizes accurate current and voltage relationship simulation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114818573B_ABST
    Figure CN114818573B_ABST
Patent Text Reader

Abstract

The present invention discloses an iterative model building method and a simulation method for a crosspoint storage array, belonging to the field of information storage and storage-computing integration. The model is: #imgabs0##imgabs1##imgabs2# The simulation method comprises: (S1) initializing the voltage of each node in the array, and forming a voltage matrix V by the voltage of the upper node and the lower node of each storage unit. up and V down ; (S2) According to I arr[k‑1] =G⊙(V up ‑V down ) calculate the current of each storage cell in the k-1th iteration, and calculate the total current flowing through each word line and each bit line; (S3) update the upper node voltage and the lower node voltage of each storage cell in the kth iteration to update V up and V down ; (S4) iteratively execute (S2) to (S3) until the error of the kth iteration compared with the previous iteration is less than ε, or the maximum number of iterations is reached. The present invention can reduce the time and space complexity of simulating a crosspoint array.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of information storage and storage-computing integration, and more specifically, relates to an iterative model building method and a simulation method for a cross-point storage array. Background Art

[0002] Resistive memory, magnetoresistive memory, phase change memory, etc. use the high and low values ​​of the resistance of the storage cell to represent the stored information. In these memories, the storage cell is sandwiched between the word line and the bit line to construct a cross-point storage array. Among them, resistive memory is the most commonly used cross-point storage array. Applying a positive or reverse pulse exceeding a certain amplitude to both ends of the storage cell of the cross-point storage array can change the resistance value of the storage cell, and applying a smaller read voltage to it can determine the stored resistance value by the magnitude of the outflowing current. Due to the existence of interconnection line resistance, whether writing, reading or calculating operations on the cross-point array, the voltage applied by the word line will be divided by the interconnection line resistance, resulting in failure of the write and read operations or too large errors in the results of the calculation operation. It is necessary to model the cross-point array considering the influence of the line resistance and accurately simulate the influence of the line resistance. The existing simulation methods have problems of large time complexity and space complexity. The following is an example of resistive memory.

[0003] Resistive Random-Access Memory (ReRAM) is a type of two-terminal passive memory device, such as Figure 1 As shown in (a) of FIG. 1 , applying a forward or reverse pulse of a certain amplitude to both ends of the ReRAM cell can change the resistance value of the cell. Applying a smaller read voltage to it can determine the stored resistance value by the magnitude of the outflowing current.

[0004] like Figure 1 As shown in (b), the ReRAM cell is sandwiched between the word line and the bit line to form a cross-point memory array. The area of ​​the ReRAM cell in the cross-point array is only 4F. 2 (F is the feature size), which is an effective way to build a large-capacity, high-density storage and energy-efficient storage and computing integrated architecture. The following introduces the writing, reading and computing operations of the crosspoint array.

[0005] like Figure 1 As shown in (c) in FIG. 1 , a half-bias write voltage scheme is generally used to write to the target cell of the cross-point array, where the target cell is in a full write voltage bias, and the cells in the same row and column as the target cell are in a half-select state, that is, the voltage across the cell is half of the write voltage, so only the target cell is written and other cells are not modified. Figure 1As shown in (d), the traditional double-terminal grounded read scheme is used for the cross-point array, that is, the word line where the target cell is located is applied with a read voltage v rd , all other word lines and bit lines are grounded, and the read operation is completed by detecting the current size of the bit line where the target cell is located. At the same time, the cross-point array also has computing power, and can complete the matrix-vector multiplication calculation of analog quantities in situ in the array. Figure 1 As shown in (e) in the figure, a voltage involved in the calculation is applied to the word line And all bit lines are grounded. Assume that the conductance of the ReRAM cell on a bit line is Then the current flowing out of the bit line can be expressed as Completed a voltage vector Conductivity vector with ReRAM cell The current output of all bit lines in the entire array completes the matrix-vector multiplication operation, where the matrix is ​​the conductance matrix of the ReRAM unit in the cross-point array and the vector is the voltage vector applied on the word line.

[0006] Traditional complete model building methods such as Figure 1 As shown in (f) in the figure, an m×n cross-point array has 2mn+m+n nodes, where the mn+m nodes on the word line are on the upper plane of the cross-point array, and the mn+n nodes on the bit line are on the lower plane of the cross-point array. Each word line can be regarded as a line resistor r connecting the nodes. wl In the same way, each bit line can be regarded as a line resistance r connecting the nodes. bl Whether it is writing, reading or computing operations, they can be regarded as applying voltage to the word line. and applying voltage to the bit line For example, in calculation operations is the voltage vector involved in the calculation, That is, a vector of all zeros is used to represent grounding. Then, through Kirchhoff's current law, equations can be established for the voltages at all nodes to form a linear system of equations: in is the voltage vector formed by all nodes, A is the coefficient matrix of the linear equations established by Kirchhoff’s current law, is the result vector of the system of equations. Finally, the system of equations is solved to obtain the voltage of each node. At the same time, the current flowing out of the bit line can be calculated through the node voltage, so as to accurately evaluate whether the write and read operations are successful and calculate the precise value of the operation. Since the number of variables in the linear system of equations is 2mn+m+n, the space complexity of solving the system of equations is O(m 2 n 2 ), the time complexity is O(m 3 n 3 ).

[0007] In general, the traditional method establishes a model of the crosspoint storage array based on a set of linear equations, and performs simulation based on the model, which has high time complexity and space complexity, low simulation efficiency, and high simulation overhead. Summary of the invention

[0008] In view of the defects and improvement needs of the prior art, the present invention provides an iterative model building method and simulation method for a cross-point storage array, the purpose of which is to reduce the time and space complexity of simulating the cross-point storage array, so as to improve the simulation efficiency and reduce the simulation overhead.

[0009] To achieve the above object, according to one aspect of the present invention, an iterative model building method for a cross-point memory array is provided. The cross-point memory array uses high and low values ​​of memory cell resistance to represent stored information. The total number of word lines m and the total number of bit lines n of the cross-point memory array satisfy:

[0010] The iterative model building method of the cross-point memory array includes: for a target memory cell located at (x, y) in the cross-point memory array, building the following iterative model based on Kirchhoff's voltage law:

[0011]

[0012] Wherein, x and y represent the numbers of the word line and bit line where the target storage unit is located, respectively, 1≤x≤m, 1≤y≤n; Represents the voltage applied to the word line where the target memory cell is located. Represents the voltage applied to the bit line where the target memory cell is located; r wl Represents the line resistance between two adjacent nodes on the word line, r bl Represents the line resistance between two adjacent nodes on the bit line; R x,y Represents the resistance of the memory cell; G max I represents the maximum conductance in the conductance range that the storage unit can select; arr A matrix representing currents flowing through each memory cell in a cross-point memory array; represents the current of the storage unit located at (p, q) in the rth iteration, the subscript r represents the number of iterations, and the superscript (p, q) represents the position of the storage unit.

[0013] According to another aspect of the present invention, a simulation method for a cross-point storage array based on the above-mentioned iterative model building method for a cross-point storage array is provided, comprising the following steps:

[0014] (S1) Initialize the voltage of each node in the cross-point memory array according to the word line voltage and bit line voltage to be applied, and form a voltage matrix V by the voltage of the upper node of each memory cell.up , the voltage matrix V is formed by the voltage of the lower node of each storage unit down , and initialize the current number of iterations to k = 1;

[0015] The upper node of the memory cell is a node formed by the intersection of the memory cell and the word line where it is located, and the lower node of the memory cell is a node formed by the intersection of the memory cell and the bit line where it is located;

[0016] (S2) According to I arr[k-1] =G⊙(V up -V down ) calculating the current flowing through each memory cell in the k-1th iteration, and calculating the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration based on the current flowing through each memory cell;

[0017] G represents the conductivity matrix of each storage unit, ⊙ represents the Hadamard product of the corresponding elements of the matrix;

[0018] (S3) updating the upper node voltage and the lower node voltage of each storage cell in the kth iteration according to the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration, thereby updating the voltage matrix V up And the voltage matrix V down Make updates;

[0019] (S4) iteratively executing steps (S2) to (S3) until the difference between the voltage of each node in the kth iteration and the voltage of each node in the k-1th iteration is less than a preset error ε, or the number of iterations reaches a preset maximum number of iterations K;

[0020] (S5) The voltage of each node in the last iteration is taken as the simulation result, and the simulation ends.

[0021] In some optional embodiments, in step (S1), initializing the voltage of each node in the cross-point memory array according to the word line voltage and the bit line voltage to be applied includes:

[0022] The voltage of each node on the word line is initialized to the voltage applied on the corresponding word line; the voltage of each node on the bit line is initialized to the voltage applied on the corresponding bit line.

[0023] In some optional embodiments, in step (S1), initializing the voltage of each node in the cross-point memory array according to the word line voltage and the bit line voltage to be applied includes:

[0024] According to the word line voltage and bit line voltage to be applied, and the conductivity matrix G of each memory cell, the corresponding node initial voltage matrix is ​​searched from the pre-established mapping table, and the voltage of each node is initialized according to the node initial voltage matrix;

[0025] The node initial voltage matrix is ​​used to record the initial voltage value of each node during the simulation process, and the mapping table is used to record the corresponding relationship between the combination of word line voltage, bit line voltage and conductance matrix and the node initial voltage matrix.

[0026] Furthermore, the maximum number of iterations K is set as follows:

[0027] When the conductance of each storage unit is the maximum conductance, step (S1) is executed, and after initialization is completed, steps (S2) to (S3) are iteratively executed until a preset relative error δ is reached;

[0028] Assign the value of the last iteration number k to the maximum iteration number K;

[0029] The relative error is the error of each node voltage obtained by simulation relative to the actual voltage of each node.

[0030] Further, in step (S2), the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration are calculated based on the current flowing through each memory cell, including:

[0031] according to Calculate the total current flowing through the pth word line, 1≤p≤m;

[0032] according to Calculate the total current flowing through the qth bit line, 1≤q≤n.

[0033] Further, in step (S4), the upper node voltage and the lower node voltage of each memory cell in the kth iteration are updated according to the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration, including:

[0034] The upper node voltages of the storage cells on the first to nth bit lines are updated sequentially, including: After updating the voltage of the upper node of the memory cell on the first bit line, Update word line current; according to After updating the voltage of the upper node of the memory cell on the jth bit line, Update word line current; j = 2, 3, ... n;

[0035] The lower node voltages of the storage cells on the mth to 1st word lines are updated sequentially, including: After the lower node voltage of the memory cell on the mth word line is calculated, Update the bit line current; according to After the lower node voltage of the memory cell on the i-th word line is calculated, Update bit line current; i = m-1, m-2, ..., 1;

[0036] Among them, V wl is the voltage vector composed of the voltage applied on each word line, I wl is the current vector composed of the current flowing through each word line calculated in step (S2); V bl is the voltage vector composed of the voltage applied to each line, I bl is a current vector composed of the current flowing through each bit line calculated in step (S2); is the voltage vector composed of the upper node voltage of the memory cell on the first bit line, is the voltage vector composed of the upper node voltage of the storage cell on the jth bit line, is the current vector consisting of the current flowing through the storage cell on the first bit line in the k-1th iteration, is the current vector consisting of the current flowing through the storage cell on the j-th bit line in the k-1-th iteration; is the voltage vector composed of the lower node voltages of the memory cells on the mth word line, is the voltage vector composed of the lower node voltages of the memory cells on the i-th word line, is the current vector consisting of the current flowing through the storage cell on the mth word line in the k-1th iteration, is the current vector consisting of the current flowing through the storage cell on the i-th word line in the k-1-th iteration.

[0037] According to another aspect of the present invention, a computer-readable storage medium is provided, comprising a stored computer program; when the computer-readable storage medium is executed by a processor, the device where the computer-readable storage medium is located is controlled to execute the iterative model building method for the cross-point storage array provided by the present invention, and / or the simulation method for the cross-point storage array provided by the present invention.

[0038] In general, through the above technical solution conceived by the present invention, a model of a cross-point memory array is established by establishing an iterative sequence. Based on the model, when simulating the cross-point memory array, the influence of the line resistance can be fully considered, and the time complexity of updating the current flowing through the memory cell and the time complexity of updating the node voltage are both O(mn) each time. In the worst case, that is, when the conductance of each memory cell is the maximum conductance, the logarithm needs to be executed. γ δ iterations, where represents the convergence rate of the iterative sequence, so the time complexity of the simulation method is O(mn·log γδ); Since each simulation only needs to record the value of the node voltage, the space complexity of the simulation method is O(mn). Compared with the traditional modeling and simulation method, the present invention can effectively reduce the time and space complexity of simulating the cross-point array, thereby improving the simulation efficiency and reducing the simulation overhead. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 Schematic diagram of the structure and related principles of existing resistive random access memory; (a) is a diagram showing the variation of ReRAM unit resistance with pulses, (b) is a diagram showing the ReRAM cross-point storage array, (c) is a diagram showing the write operation of half-biased write voltage, (d) is a diagram showing the full ground read operation, (e) is a diagram showing the matrix-vector multiplication operation, and (f) is a complete model of the cross-point array;

[0040] Figure 2 A schematic diagram of an iterative model building method for a cross-point storage array provided by an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of node voltage initialization in a simulation method for a cross-point memory array provided by an embodiment of the present invention;

[0042] Figure 4 A schematic diagram of a simulation method for a cross-point storage array provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0044] In the present invention, the terms "first", "second", etc. (if any) in the present invention and the drawings are used to distinguish similar objects but not necessarily to describe a specific order or sequence.

[0045] In order to solve the technical problems of high time complexity and high space complexity in the existing simulation methods of cross-point storage arrays, the present invention provides an iterative model establishment method and a simulation method for a cross-point storage array, the overall idea of ​​which is: under the condition of satisfying the model convergence conditions, a model of the cross-point storage array is established by establishing an iterative sequence, while ensuring that the established model can accurately reflect the relationship between current and voltage in the cross-point storage array, the time complexity and space complexity of subsequent simulation based on the model are effectively reduced.

[0046] The following are examples.

[0047] Embodiment 1:

[0048] A method for establishing an iterative model of a cross-point memory array; in this embodiment, the cross-point memory array uses the high and low values ​​of the memory cell resistance to represent the stored information, and the total number of word lines m and the total number of bit lines n of the cross-point memory array satisfy:

[0049] In this embodiment, the iterative model building method of the cross-point memory array includes: for a target memory cell located at (x, y) in the cross-point memory array, building the following iterative model based on Kirchhoff's voltage law:

[0050]

[0051] Wherein, x and y represent the numbers of the word line and bit line where the target storage unit is located, respectively, 1≤x≤m, 1≤y≤n; Represents the voltage applied to the word line where the target memory cell is located. Represents the voltage applied to the bit line where the target memory cell is located; r wl Represents the line resistance between two adjacent nodes on the word line, r bl Represents the line resistance between two adjacent nodes on the bit line; R x,y Represents the resistance of the memory cell; G max I represents the maximum conductance in the conductance range that the storage unit can select; arr A matrix representing currents flowing through each memory cell in a cross-point memory array; represents the current of the storage unit located at (p, q) in the rth iteration, the subscript r represents the number of iterations, and the superscript (p, q) represents the position of the storage unit.

[0052] The cross-point storage array model established in this embodiment is an iterative sequence. Based on Kirchhoff's voltage law, the model assumes that the current passing through each storage cell in the current iteration is derived from the storage cell current in the previous iteration. The convergence of the model established in this embodiment is verified in combination with relevant principles.

[0053] The current matrix I flowing through each memory cell in the cross-point memory array is real is a variable, Represents the current flowing through the storage cell at position (i, j); complete modeling based on Kirchhoff's voltage law: for the target cell at position (x, y) in the m×n cross-point storage array and its word line and bit line, such as Figure 2 As shown, the following Kirchhoff voltage equation can be established:

[0054]

[0055] The above Kirchhoff voltage equation can be written in matrix form as follows:

[0056]

[0057] Among them, Λ x,y and I real are all m×n matrices, It is a convolution operation of two matrices of the same size; the following equation can be established for all storage units:

[0058]

[0059] In equation (1), Λ is (m 2 )×(n 2 ) size, consisting of the matrix Λ corresponding to all storage units x,y Composition; V represents the matrix composed of the voltage differences across all storage cells; is a convolution operation with a step window of (m,n); here, the model is written in the form of a convolution operation rather than a linear equation system to facilitate subsequent model derivation;

[0060] Based on the complete model of the convolution operation, that is, the above formula (1), the iterative sequence is established by transforming the formula so that the limit of the sequence is the simulation result of the complete model, and the iterative formula of the sequence is derived. Still for the target cell at position (x, y) in the m×n crosspoint array and its word line and bit line, the following iterative model can be established:

[0061]

[0062] In each iteration, the current matrix I flowing through the cells in the cross-point memory array arr will be updated, so I arr In the above formula, the serial number in “[]” indicates the number of iterations, which is also the number of items in the sequence; accordingly, I arr[k] is the kth item of the sequence, I arr[k-1] is the k-1th item of the sequence; in the above iterative model, the current passing through each storage unit in the current iteration (i.e., the kth iteration) is derived from the storage unit current in the previous iteration (i.e., the k-1th iteration);

[0063] The above iterative model is written in matrix form as follows:

[0064]

[0065] Ψ x,y =[R x,y ]

[0066] Among them,x,y is an m×n matrix;

[0067] For all elements, the following equation is established:

[0068]

[0069] Among them, R represents the matrix composed of the resistance of all memory cells, and Ψ is the matrix Ψ corresponding to all memory cells. x,y Composition; ⊙ is the Hadamard product of the corresponding elements of the matrix;

[0070] The above equation (2) can be further transformed into

[0071]

[0072] Where G represents the matrix composed of the conductances of all storage cells. Combining equations (1) and (2), the following formula can be obtained:

[0073]

[0074] Subtract I from both sides real , we can get:

[0075]

[0076] It can be further simplified to:

[0077] I real -I arr[k] =(U-ρ)(I real -I arr[k-1] )

[0078] =(U-ρ) k (I real -I arr[0] )

[0079] Where U is the identity matrix, I arr[0] is the initialization of the current matrix of each unit in the crosspoint storage array, and the function ρ(X) is calculated Therefore Therefore, in order for the sequence to converge to I real , a sufficient but not necessary convergence condition is that, in any i-th iteration, the iterative current matrix I arr[i] Relative to the actual current matrix I real The error I real -I arr[i] It is reduced compared with the previous iteration, namely:

[0080]

[0081] The division, absolute value and less than symbol here are to divide each element in the matrix, find the absolute value and compare the corresponding elements. m,n is an m×n matrix of all ones.

[0082] In order to further simplify the convergence criterion formula, it is assumed that I real -I arr[i-1] Equal to ξ i ·J m,n , where ξ i is an arbitrarily small quantity; for the above convergence conditions, for any storage unit at any position (x, y) in any i-th iteration, we can obtain:

[0083]

[0084]

[0085] When the conductance of each unit is the maximum conductance, that is, the minimum resistance, the influence of line resistance is the greatest, which is called the worst case. Considering the worst case, the convergence condition discriminant of the sequence can be further simplified as:

[0086]

[0087] G max Represents the maximum conductance in the conductance range that the storage unit can select, R min Indicates that G max The corresponding minimum resistance, G max *R min =1. According to the above formula, the sequence convergence rate, i.e., the sequence convergence ratio, is γ. Therefore, in order to achieve an acceptable relative error δ, log γ δ iterations. In the non-worst case, the convergence ratio will be smaller, that is, the convergence speed will be faster; since in this embodiment, the crosspoint storage array satisfies That is, the convergence condition is met. Therefore, the iterative model established in this embodiment can eventually converge to I real ,Based on this model, the voltage of each node in the array can be accurately calculated.

[0088] Embodiment 2:

[0089] A simulation method for a cross-point storage array is proposed in this embodiment based on the iterative model establishment method for the cross-point storage array provided in the above embodiment 1. Figure 3 and Figure 4 , this embodiment includes the following steps:

[0090] (S1) Initialize the voltage of each node in the cross-point memory array according to the word line voltage and bit line voltage to be applied, and form a voltage matrix V by the voltage of the upper node of each memory cell. up , the voltage matrix V is formed by the voltage of the lower node of each storage unit down , and initialize the current number of iterations to k = 1;

[0091] The upper node of the memory cell is a node formed by the intersection of the memory cell and the word line where it is located, and the lower node of the memory cell is a node formed by the intersection of the memory cell and the bit line where it is located;

[0092] As an optional implementation, in this embodiment, it is assumed that the initial moment is an ideal situation with no line resistance. Accordingly, the voltage of each node on the word line is initialized to the voltage applied on the corresponding word line, and the voltage of each node on the bit line is initialized to the voltage applied on the corresponding bit line, such as Figure 3 As shown;

[0093] (S2) According to I arr[k-1] =G⊙(V up -V down ) calculating the current flowing through each memory cell in the k-1th iteration, and calculating the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration based on the current flowing through each memory cell;

[0094] G represents the conductivity matrix of each storage unit, ⊙ represents the Hadamard product of the corresponding elements of the matrix;

[0095] After finding the current flowing through each memory cell, the total current flowing through the word line and the bit line is calculated by:

[0096] For any p-th word line, the current of all storage cells on the word line is summed in the column direction to obtain the total current flowing through the word line Right now 1≤p≤m;

[0097] For any qth bit line, the current of all storage cells on the bit line is summed in the row direction to obtain the total current flowing through the bit line. Right now 1≤q≤n;

[0098] The above process of calculating word line current and bit line current can be abbreviated as I wl =sum(I arr[k-1] ,0) and I bl =sum(I arr[k-1] ,1), where sum(I arr[k-1] ,0) indicates I arr[k-1] The column direction of sum(I arr[k-1] ,1) indicates Iarr[k-1] The row direction of ;

[0099] (S3) updating the upper node voltage and the lower node voltage of each storage cell in the kth iteration according to the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration, thereby updating the voltage matrix V up And the voltage matrix V down Make updates;

[0100] Based on Kirchhoff's voltage law and Kirchhoff's current law, for any p-th word line, the voltage of the first node on it is The voltage applied to the word line Subtract line resistance voltage divider Get it, then update for Thus, the cumulative sum of subsequent currents of the word line is obtained, so Repeat this process to update all node voltages of the word line in the pth row;

[0101] In the same way, the voltage of each node on the bit line can be updated;

[0102] Based on the above analysis, if Figure 4 As shown, in this embodiment, the method of updating the node voltage includes:

[0103] The upper node voltages of the storage cells on the first to nth bit lines are updated sequentially, including: After updating the voltage of the upper node of the memory cell on the first bit line, Update word line current; according to After updating the voltage of the upper node of the memory cell on the jth bit line, Update word line current; j = 2, 3, ... n;

[0104] The lower node voltages of the storage cells on the mth to 1st word lines are updated sequentially, including: After the lower node voltage of the memory cell on the mth word line is calculated, Update the bit line current; according to After the lower node voltage of the memory cell on the i-th word line is calculated, Update bit line current; i = m-1, m-2, ..., 1;

[0105] Among them, V wl is the voltage vector composed of the voltage applied on each word line, I wl is the current vector composed of the current flowing through each word line calculated in step (S2); V bl is the voltage vector composed of the voltage applied to each line, Ibl is a current vector composed of the current flowing through each bit line calculated in step (S2); is the voltage vector composed of the upper node voltage of the memory cell on the first bit line, is the voltage vector composed of the upper node voltage of the storage cell on the jth bit line, is the current vector consisting of the current flowing through the storage cell on the first bit line in the k-1th iteration, is the current vector consisting of the current flowing through the storage cell on the j-th bit line in the k-1-th iteration; is the voltage vector composed of the lower node voltages of the memory cells on the mth word line, is the voltage vector composed of the lower node voltages of the memory cells on the i-th word line, is the current vector consisting of the current flowing through the storage cell on the mth word line in the k-1th iteration, is the current vector consisting of the current flowing through the storage cell on the i-th word line in the k-1-th iteration.

[0106] (S4) iteratively executing steps (S2) to (S3) until the difference between the voltage of each node in the kth iteration and the voltage of each node in the k-1th iteration is less than a preset error ε, or the number of iterations reaches a preset maximum number of iterations K;

[0107] (S5) The voltage of each node in the last iteration is taken as the simulation result, and the simulation ends;

[0108] As a preferred implementation, in the embodiment, the maximum number of iterations K is set as follows:

[0109] When the conductance of each storage unit is the maximum conductance, step (S1) is executed, and after initialization is completed, steps (S2) to (S3) are iteratively executed until the relative error is less than δ;

[0110] Assign the value of the iteration number k of the last iteration to the maximum iteration number K.

[0111] Among them, the relative error δ is the error of each node voltage obtained by simulation relative to the actual voltage of each node;

[0112] This embodiment sets the maximum number of iterations K in the iteration termination condition in the above manner. There are two final iteration termination conditions. One is that the difference between the simulation results obtained by two adjacent iterations is small (smaller than the error ε), and the other is that the error of the simulation result obtained by the current iteration relative to the true result is small (smaller than the relative error δ). Ultimately, it is ensured that the obtained simulation results have high accuracy, so that it can accurately evaluate whether the write and read operations are successful and the precise value of the calculation operation.

[0113] Embodiment 3:

[0114] A simulation method for a cross-point memory array, this embodiment is similar to the above-mentioned embodiment 2, except that in step (S1) of this embodiment, the voltage of each node in the cross-point memory array is initialized according to the word line voltage and the bit line voltage to be applied, including:

[0115] According to the word line voltage and bit line voltage to be applied, and the conductivity matrix G of each memory cell, the corresponding node initial voltage matrix is ​​searched from the pre-established mapping table, and the voltage of each node is initialized according to the node initial voltage matrix;

[0116] The node initial voltage matrix is ​​used to record the initial voltage value of each node during the simulation process, and the mapping table is used to record the corresponding relationship between the combination of word line voltage, bit line voltage and conductance matrix and the node initial voltage matrix.

[0117] Compared with the above-mentioned embodiment 2, this embodiment pre-establishes the corresponding relationship between the combination of different word line voltages, bit line voltages and conductivity matrices and the combination of upper node initial voltage matrix and lower node initial voltage matrix. During simulation, the corresponding upper node initial voltage matrix and lower node initial voltage matrix are first found by table lookup, and the node voltage is initialized according to the search result, so that at the initial moment, the node voltage is closer to the simulation result, thereby effectively reducing the number of iterations required for simulation and further improving simulation efficiency.

[0118] In this embodiment, the mapping table can be established by simulating different combinations of word line voltages, bit line voltages and conductance matrices using the simulation method described in the above embodiment 2, and recording the corresponding node voltage simulation results to complete the establishment of the mapping table.

[0119] Embodiment 4:

[0120] A computer-readable storage medium includes a stored computer program; when the computer-readable storage medium is executed by a processor, the device where the computer-readable storage medium is located is controlled to execute the iterative model building method for the cross-point storage array provided in the above-mentioned embodiment 1, and / or the simulation method for the cross-point storage array provided in any one of the above-mentioned embodiments 2 to 3.

[0121] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for iteratively modeling a cross-point memory array, wherein the cross-point memory array uses high and low values ​​of memory cell resistance to represent stored information, characterized in that: The total number of word lines m and the total number of bit lines n of the cross-point memory array satisfy: The iterative model building method of the cross-point storage array includes: for a target storage cell located at (x, y) in the cross-point storage array, building the following iterative model based on Kirchhoff's voltage law: Wherein, x and y represent the numbers of the word line and bit line where the target storage unit is located, respectively, 1≤x≤m, 1≤y≤n; represents the voltage applied to the word line where the target memory cell is located, represents the voltage applied to the bit line where the target memory cell is located; r wl Represents the line resistance between two adjacent nodes on the word line, r bl Represents the line resistance between two adjacent nodes on the bit line; R x,y Represents the resistance of the memory cell; G max I represents the maximum conductance in the conductance range that the storage unit can select; arr A matrix representing currents flowing through each memory cell in the cross-point memory array; represents the current of the storage unit located at (p, q) in the rth iteration, the subscript r represents the number of iterations, and the superscript (p, q) represents the position of the storage unit.

2. A simulation method for a cross-point storage array based on the iterative model building method for a cross-point storage array according to claim 1, characterized in that: The steps include: (S1) Initializing the voltage of each node in the cross-point memory array according to the word line voltage and bit line voltage to be applied, and forming a voltage matrix V by the voltage of the upper node of each memory cell up , the voltage matrix V is formed by the voltage of the lower node of each storage unit down , and initialize the current number of iterations to k = 1; The upper node of the memory cell is a node formed by the intersection of the memory cell and the word line where it is located, and the lower node of the memory cell is a node formed by the intersection of the memory cell and the bit line where it is located; (S2) According to I arr[k-1] =G⊙(V up -V down ) calculating the current flowing through each memory cell in the k-1th iteration, and calculating the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration based on the current flowing through each memory cell; G represents the conductivity matrix of each storage unit, ⊙ represents the Hadamard product of the corresponding elements of the matrix; (S3) updating the upper node voltage and the lower node voltage of each storage cell in the kth iteration according to the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration, thereby updating the voltage matrix V up And the voltage matrix V down Make updates; (S4) iteratively executing steps (S2) to (S3) until the difference between the voltage of each node in the kth iteration and the voltage of each node in the k-1th iteration is less than a preset error ε, or the number of iterations reaches a preset maximum number of iterations K; (S5) The voltage of each node in the last iteration is taken as the simulation result, and the simulation ends.

3. The cross-point storage array simulation method according to claim 2, characterized in that: In the step (S1), the voltage of each node in the cross-point memory array is initialized according to the word line voltage and the bit line voltage to be applied, including: The voltage of each node on the word line is initialized to the voltage applied on the corresponding word line; the voltage of each node on the bit line is initialized to the voltage applied on the corresponding bit line.

4. The cross-point storage array simulation method according to claim 2, characterized in that: In the step (S1), the voltage of each node in the cross-point memory array is initialized according to the word line voltage and the bit line voltage to be applied, including: According to the word line voltage and bit line voltage to be applied, and the conductivity matrix G of each memory cell, the corresponding node initial voltage matrix is ​​searched from a pre-established mapping table, and the voltage of each node is initialized according to the node initial voltage matrix; The node initial voltage matrix is ​​used to record the initial voltage value of each node during the simulation process, and the mapping table is used to record the corresponding relationship between the combination of word line voltage, bit line voltage and conductance matrix and the node initial voltage matrix.

5. The cross-point storage array simulation method according to claim 3 or 4, characterized in that: The maximum number of iterations K is set as follows: When the conductance of each storage unit is the maximum conductance, the step (S1) is executed, and after initialization is completed, the steps (S2) to (S3) are iteratively executed until a preset relative error δ is reached; Assign the value of the last iteration number k to the maximum iteration number K; The relative error is the error of each node voltage obtained by simulation relative to the actual voltage of each node.

6. The cross-point memory array simulation method according to claim 5, characterized in that: In the step (S2), the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration are calculated based on the current flowing through each memory cell, including: according to Calculate the total current flowing through the pth word line, 1≤p≤m; according to Calculate the total current flowing through the qth bit line, 1≤q≤n.

7. The cross-point memory array simulation method according to claim 6, characterized in that: In the step (S4), the upper node voltage and the lower node voltage of each memory cell in the kth iteration are updated according to the total current flowing through each word line and the total current flowing through each bit line in the k-1th iteration, including: The upper node voltages of the storage cells on the first to nth bit lines are updated sequentially, including: After updating the voltage of the upper node of the memory cell on the first bit line, Update word line current; according to After updating the voltage of the upper node of the memory cell on the jth bit line, Update word line current; j = 2, 3, ... n; The lower node voltages of the storage cells on the mth to 1st word lines are updated sequentially, including: After the lower node voltage of the memory cell on the mth word line is calculated, Update the bit line current; according to After the lower node voltage of the memory cell on the i-th word line is calculated, Update bit line current; i = m-1, m-2, ..., 1; Among them, V wl is the voltage vector composed of the voltage applied on each word line, I wl is the current vector composed of the current flowing through each word line calculated in step (S2); V bl is the voltage vector composed of the voltage applied to each line, I bl is a current vector composed of the current flowing through each bit line calculated in step (S2); is the voltage vector composed of the upper node voltage of the memory cell on the first bit line, is the voltage vector composed of the upper node voltage of the storage cell on the jth bit line, is the current vector consisting of the current flowing through the storage cell on the first bit line in the k-1th iteration, is the current vector consisting of the current flowing through the storage cell on the j-th bit line in the k-1-th iteration; is the voltage vector composed of the lower node voltages of the memory cells on the mth word line, is the voltage vector composed of the lower node voltages of the memory cells on the i-th word line, is the current vector consisting of the current flowing through the storage cell on the mth word line in the k-1th iteration, is the current vector consisting of the current flowing through the storage cell on the i-th word line in the k-1-th iteration.

8. A computer-readable storage medium, characterized in that: Comprising a stored computer program; when the computer-readable storage medium is executed by a processor, the device where the computer-readable storage medium is located is controlled to execute the iterative model building method for the cross-point storage array according to claim 1, and / or the simulation method for the cross-point storage array according to any one of claims 2 to 7.