TSV test grouping method based on neural network

Abstract

The invention belongs to the technical field of super-large-scale integrated circuit testing, and discloses a TSV test grouping method based on a neural network, and the method comprises the steps: writing a test constraint into a loss function through a method of combining a self-designed neural network and the loss function; continuously optimizing the weight by using a gradient descent method to solve a specific test grouping scheme under a fixed test grouping number; and the minimum test grouping number is solved by using the improved dichotomy, so that the robustness of the grouping algorithm is improved. Experimental analysis proves that the method provided by the invention has excellent grouping accuracy and operation efficiency under the irregular TSV layout.

Description

Technical Field

[0001] The invention belongs to the technical field of ultra-large-scale integrated circuit testing, and in particular relates to a TSV testing grouping method based on a neural network. Background Art

[0002] In recent years, through silicon vias (TSV) have become a key technology for achieving efficient chip stacking. Through dense vertical connections, high-bandwidth and low-latency communication can be achieved between chips, but correspondingly, the overly dense TSV layout exacerbates the impact of crosstalk, resulting in crosstalk failures, which seriously affect the quality of TSV transmission signals. Crosstalk failures are when certain transitions or states of the victim TSV are affected by the crosstalk noise caused by the adjacent attacking TSV, which has a significant impact on timing and signal integrity. Therefore, testing TSV crosstalk failures is critical to ensuring quality and improving manufacturing yield.

[0003] In the crosstalk fault model, adjacent TSVs have the greatest crosstalk impact on the victim TSV, but TSVs farther away also affect the victim TSV. In order to effectively test all TSVs, it is necessary to group TSVs that have no crosstalk relationship into the same test iteration, so that all TSVs in the same test iteration can be tested in parallel to shorten the test time.

[0004] When grouping TSVs, not only the impact distance should be considered, but also the TSV layout topology, such as regular and irregular TSV layouts; for regular TSV layouts, there are existing methods that cover the TSV array by basic shapes and group the vertices into the same test group, proving that the grouping result of the hexagonal structure TSV layout is optimal, but this method cannot be applied to irregular TSV layouts because it is impossible to use basic graphics to arrange in sequence to cover all irregular TSV arrays. Irregular TSV layouts are more complex, and existing methods such as grouping methods based on greedy algorithms cannot guarantee a global optimal solution, especially when the layout is complex. Usually, after each group of TSV tests is completed, the test responses need to be removed for observation, and the time required is proportional to the number of TSVs. Therefore, test time is an important factor, and reducing the number of test groups is crucial to minimize the total test time. Although the grouping method for regular TSV layouts has been well explored, the grouping method for irregular TSV layouts still has certain limitations, so it is urgent to develop a TSV test grouping method designed specifically for irregular layouts. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a TSV test grouping method based on neural network. By adopting a method combining binary search method and neural network, the invention purpose of obtaining the minimum test iteration and corresponding grouping scheme that meets the test constraints is achieved. At the same time, it has a stronger ability to fit nonlinear relationships, and is therefore more suitable for more complex and larger-scale TSV layouts, and can solve the technical problems of poor efficiency and low precision of existing TSV grouping methods.

[0006] The present invention provides a TSV test grouping method based on a neural network, comprising the following steps: S1. Use the binary method to initialize the number of groups of the TSV to be tested; S2, constructing a neural network according to the number of TSVs to be tested, and randomly initializing and grouping the TSVs to be tested; S3, construct an adjacency matrix to characterize whether there is crosstalk between any two TSVs; S4, define the loss function, and calculate the current loss function value according to the probability that each TSV is assigned to different test groups; S5, updating the weight of the neural network according to the result of S4, and obtaining an updated TSV grouping scheme; S6. Determine whether the updated grouping scheme meets the TSV test constraint. If it is correct, use binary search to determine whether the minimum number of groups is found. If it is found, end. Otherwise, return to S1 and update the maximum number of groups. If it does not meet the TSV test constraint, repeat S3-S6.

[0007] Furthermore, S1 is specifically: Assume there are N TSVs to be tested, set the maximum number of groups , minimum number of groups , initialize the number of TSV groups according to the binary method .

[0008] Further, in S2, the neural network includes an input layer, a fully connected layer, a softmax layer and an output layer, The first layer is the input layer, where the input data for The dimension constant is 1, which represents TSVs; The second layer is a fully connected layer, which outputs a one-dimensional matrix , ,express TSV is divided into Group data; the output data of this layer is composed of the input data of the first layer and the weights of the neural network The solution for each set of data is shown in formula (1). TSV, the preset number of groups is situation, After the calculation is completed, it is reshaped A real matrix of : , The third layer is the softmax layer, and its output The real matrix of ; Indicates TSVs are divided into The probability of a group: , The fourth layer is the output layer, which converts the output of the softmax layer into a one-hot encoding and outputs the data , to determine which test iteration each TSV is ultimately assigned to.

[0009] Furthermore, in S2, the neural network weights are initialized by uniform distribution to obtain an initialization grouping scheme.

[0010] Furthermore, in S3, in order to characterize whether there is crosstalk between any two TSVs, a The adjacency matrix of ; If the number is and number There is crosstalk between the two TSVs, and the adjacency matrix In and The value of two positions and are all 1, otherwise they are 0; , , and m≠n.

[0011] Furthermore, S4 is specifically: Define the collision loss function , the adjacency matrix constructed by calling Calculate as shown in formula (3): , in, represents the probability that any two TSVs are assigned to the same test iteration, and the factor The role of is to keep the collision loss at the same level for different TSV numbers; Defining the imbalance loss function , whose value is the mean square error of the number of TSVs in different test groups, is defined as follows: ; Total loss function is defined as the weighted sum of collision loss and misalignment loss, and The weights of collision loss and misalignment loss are respectively, and the constraint is to avoid collision and ensure that TSVs with crosstalk are not assigned to the same test iteration: ; The current loss function value is calculated using the probability distribution solved by the softmax layer.

[0012] Furthermore, S5 specifically includes: Update the neural network weights to reduce the loss function. The weight update adopts the gradient descent method, as shown in formula (6), where Represents the learning rate, which is a preset fixed value; , After the update is completed, the loss function is recalculated; Repeat the operation of updating the weights and the loss function multiple times until the weights are stable and the output data of the neural network is stable.

[0013] Furthermore, S6 is specifically: Determine whether the specific grouping scheme (output data of the neural network) meets the test constraints of TSV, that is, any two TSVs with crosstalk relationship are not grouped into the same test group; If the grouping scheme is correct, return to S1 and update the maximum number of groups to ; If the grouping scheme is incorrect, in order to avoid misclassification due to weak robustness of the neural network, repeat steps S3-S6. If the correct solution is not found after multiple attempts, the minimum number of groups is updated. ; like , which means finding the minimum number of groups. The minimum number of groups is , otherwise go to S1.

[0014] The beneficial effects of the present invention are: (1) Compared with the traditional grouping method based on greedy algorithm, the method of the present invention can efficiently explore the solution space, continuously optimize the solution according to the loss function to explore the possibility of solving a better grouping solution, avoid the problem of poor solution caused by local optimal goals, and reduce the number of test groups while ensuring the test coverage, thereby reducing the test time and improving the test efficiency; (2) The method of the present invention reduces the unbalanced consumption of hardware resources by balancing the size of the test group, thereby avoiding the waste of resources caused by an overly unbalanced test group; (3) Strong adaptability: The method described in the present invention can dynamically adjust parameters such as learning rate based on different circuit scales. The neural network can adjust the grouping strategy according to actual conditions and has better adaptability to meet the TSV test grouping requirements under large-scale and complex layouts. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a flow chart of the method of the present invention; Figure 2 Schematic diagram of the neural network described in the present invention. DETAILED DESCRIPTION

[0016] In order to make the contents of the present invention more clearly understood, the present invention is further described in detail below based on specific embodiments in conjunction with the accompanying drawings.

[0017] Example 1

[0018] like Figure 1 As shown, a TSV test grouping method based on a neural network according to the present invention comprises the following steps: S1. Use the binary method to initialize the number of groups of the TSV to be tested; S2, constructing a neural network according to the number of TSVs to be tested, and randomly initializing and grouping the TSVs to be tested; S3, construct an adjacency matrix to characterize whether there is crosstalk between any two TSVs; S4, define the loss function, and calculate the current loss function value according to the probability that each TSV is assigned to different test groups; S5, updating the weight of the neural network according to the result of S4, and obtaining an updated TSV grouping scheme; S6. Determine whether the updated grouping scheme meets the TSV test constraint. If it is correct, use binary search to determine whether the minimum number of groups is found. If it is found, end. Otherwise, return to S1 and update the maximum number of groups. If it does not meet the TSV test constraint, repeat S3-S6.

[0019] In S1, assuming there are N TSVs to be tested, set the maximum number of groups , minimum number of groups , initialize the number of TSV groups according to the binary method .

[0020] Figure 2 is a neural network structure diagram of a TSV. If there is TSV, then the input layer is dimension, the fully connected layer, softmax layer, and output layer are dimension.

[0021] The neural network includes an input layer, a fully connected layer, a softmax layer and an output layer; The first layer is the input layer, where the input data for The dimension constant is 1, which represents TSVs; The second layer is a fully connected layer, and the output data of the fully connected layer is a one-dimensional matrix ( ,express TSV is divided into Each group of data consists of groups), the output data of this layer is composed of the first layer input data and the weights of the neural network The solution for each set of data is shown in formula (1). TSV, the preset number of groups is situation, After the calculation is completed, it is reshaped A real matrix of : , The third layer is the softmax layer, and the output data of the softmax layer is a matrix , for The real matrix of Indicates TSVs are divided into The softmax function is applied to the output of the fully connected layer to convert the output feature vector of the fully connected layer into a probability distribution, which represents the probability of each TSV being assigned to a different test group: , The fourth layer is the output layer. The output data of the output layer , to convert the output of the softmax layer into one-hot encoding, and assign each TSV to the group whose test iteration is set to 1 according to the one-hot encoding to determine which test iteration each TSV is finally assigned to.

[0022] The neural network weights are initialized by uniform distribution to obtain an initialization grouping scheme.

[0023] In S3, in order to characterize whether there is crosstalk between any two TSVs, a The adjacency matrix of , if the number is and number ( , ) has crosstalk between the two TSVs, and the adjacency matrix In and The value of two positions and 1 if both are true, otherwise 0.

[0024] In S4, to ensure the correct grouping of TSVs, a collision loss function is defined: , the adjacency matrix constructed by calling Calculation is performed, as shown in formula (3), the higher the probability that TSVs with crosstalk relationship are classified into the same test iteration, the greater the value of the collision loss function; , in, represents the probability that any two TSVs are assigned to the same test iteration, and the factor The role of is to keep the collision loss at the same level with different TSV numbers.

[0025] In addition, the diagnostic circuit needs to balance the number of TSVs in different test groups in each test iteration to minimize hardware overhead; since unevenness consumes more resources, the offset loss function is defined , whose value is the mean square error of the number of TSVs in different test groups, is defined as follows: , in, is the output of the softmax layer, i.e. TSVs are divided into The probability of a test group.

[0026] The total loss function is defined as the weighted sum of the collision loss and the misalignment loss. The primary constraint is to avoid collisions and ensure that TSVs with crosstalk are not assigned to the same test iteration; therefore, the weight assigned to the collision loss function must greatly exceed the weight of the misalignment loss. The total loss function is defined as follows: ; The current loss function value is calculated using the probability distribution solved by the softmax layer.

[0027] In S5, the neural network weights are updated to reduce the loss function. The weights are updated using the gradient descent method, as shown in formula (6), where Represents the learning rate, which is a preset fixed value; , After the update is completed, the loss function is recalculated; the weights are updated repeatedly, and the loss function is updated multiple times until the weights are stable, the loss function value changes very little, and the neural network output data is stable.

[0028] S6 is specifically: determining whether the updated grouping scheme (output data of the neural network) satisfies the test constraints of the TSV, that is, any two TSVs with a crosstalk relationship are not grouped into the same test group; If the grouping scheme is correct, return to S1 and update the maximum number of groups to ; If the grouping scheme is incorrect, in order to avoid misclassification due to weak robustness of the neural network, repeat steps S3-S6. If the correct solution is not found after multiple attempts, the minimum number of groups is updated. ; Among them, if , which means finding the minimum number of groups. The minimum number of groups is , otherwise go to S1.

[0029] In order to better verify and illustrate the technical effects used in the method of the present invention, this embodiment selects the grouping method of the traditional greedy algorithm to compare with the method of the present invention, compares the test results by means of scientific demonstration, and verifies the real effect of the method of the present invention.

[0030] As shown in Table 1, compared with the grouping method of the traditional greedy algorithm, the present invention greatly reduces the number of test iterations while increasing a small amount of computing time, and the number of victim TSVs in each test group is relatively uniform.

[0031] Table 1 Performance comparison between the grouping method of the traditional greedy algorithm and the method of the present invention

[0032] Example 2

[0033] In order to further compare the neural network-based method with the greedy algorithm-based grouping method, the number of groupings solved by the two methods under TSV layouts of different scales and complexities is compared.

[0034] For small and medium scale, the number of TSVs is set to 25, 50, 100, 150 and 200. Each configuration is simulated 80 times in different crosstalk impact ranges, and a total of 400 simulations are performed. The crosstalk impact range is set to 0% ~ 50%, and the statistical step size is 12.5%. For large scale, the number of TSVs is set to 300, 400, 500, 600, 700, 800 and 900. Each configuration is simulated 48 times in different crosstalk impact ranges, and a total of 336 simulations are performed. The crosstalk impact range is set to 0% ~ 30%, and the statistical step size is 7.5%.

[0035] As the crosstalk impact range increases, more and more TSVs are affected by crosstalk, making parallel testing impossible and increasing the complexity of the solution. Therefore, different crosstalk impact ranges represent different levels of complexity.

[0036] The comparison of the number of test iterations obtained on small and medium-scale TSV layouts using the method of the present invention is shown in Table 2. The number of TSVs is set to 25, 50, 100, 150 and 200. The values ​​in Table 2 represent the total difference in the number of groups under different degrees of crosstalk between the two methods. Since the number of groups based on the greedy algorithm is always greater than the number of groups based on the method of the present invention, the difference is always positive. In all scenarios, the number of groups of the neural network in the present invention is always less than that of the greedy algorithm, showing excellent performance. In addition, with the increase in complexity or the expansion of the TSV scale, the advantages of the neural network solution method are more obvious.

[0037] Table 2 Comparison results of the number of groupings in the greedy algorithm grouping method and the method of the present invention in the small and medium-scale TSV layout solution test

[0038] In large-scale TSV layout, as shown in Table 3, the method of the present invention maintains strong performance, and the number of groupings is better than the grouping method based on the greedy algorithm. As the complexity increases, its advantages become more and more obvious.

[0039] Table 3 Comparison results of the number of groupings between the greedy algorithm grouping method and the method of the present invention in large-scale TSV layout solution test

[0040] The above description is only a preferred embodiment of the present invention and is not intended to be a further limitation of the present invention. All equivalent changes made using the contents of the present specification and drawings are within the protection scope of the present invention.

Claims

1. A TSV test grouping method based on neural network, characterized in that: The following steps are involved: S1. Use the binary method to initialize the number of groups of the TSV to be tested; S2, constructing a neural network according to the number of TSVs to be tested, and randomly initializing and grouping the TSVs to be tested; S3, construct an adjacency matrix to characterize whether there is crosstalk between any two TSVs; S4, define the loss function, and calculate the current loss function value according to the probability that each TSV is assigned to different test groups; S5, updating the weight of the neural network according to the result of S4, and obtaining an updated TSV grouping scheme; S6. Determine whether the updated grouping scheme meets the TSV test constraint. If it is correct, use binary search to determine whether the minimum number of groups is found. If it is found, end. Otherwise, return to S1 and update the maximum number of groups. If it does not meet the TSV test constraint, repeat S3-S6.

2. A TSV test grouping method based on neural network according to claim 1, characterized in that: S1 is specifically: Assume there are N TSVs to be tested, set the maximum number of groups , minimum number of groups , initialize the number of TSV groups according to the binary method .

3. The TSV test grouping method based on neural network according to claim 2, characterized in that: In S2, the neural network includes an input layer, a fully connected layer, a softmax layer and an output layer. The first layer is the input layer, where the input data for The dimension constant is 1, which represents TSVs; The second layer is a fully connected layer, which outputs a one-dimensional matrix , ,express TSV is divided into Group data; the output data of this layer is composed of the input data of the first layer and the weights of the neural network The solution for each set of data is shown in formula (1). TSV, the preset number of groups is situation, After the calculation is completed, it is reshaped A real matrix of : ， The third layer is the softmax layer, and its output The real matrix of ; Indicates TSVs are divided into The probability of a group: ， The fourth layer is the output layer, which converts the output of the softmax layer into a one-hot encoding and outputs the data , to determine which test iteration each TSV is ultimately assigned to.

4. The TSV test grouping method based on neural network according to claim 3, characterized in that: In S2, the neural network weights are initialized by uniform distribution to obtain an initialization grouping scheme.

5. The TSV test grouping method based on neural network according to claim 3, characterized in that: In S3, in order to characterize whether there is crosstalk between any two TSVs, a The adjacency matrix of ; If the number is and number There is crosstalk between the two TSVs, and the adjacency matrix In and The value of two positions and are all 1, otherwise they are 0; , , and m≠n.

6. The TSV test grouping method based on neural network according to claim 5, characterized in that: S4 is specifically: Define the collision loss function , the adjacency matrix constructed by calling Calculate as shown in formula (3): ， in, represents the probability that any two TSVs are assigned to the same test iteration, and the factor The role of is to keep the collision loss at the same level for different TSV numbers; Defining the imbalance loss function , whose value is the mean square error of the number of TSVs in different test groups, is defined as follows: ， Total loss function is defined as the weighted sum of collision loss and misalignment loss, and The weights of collision loss and misalignment loss are respectively, and the constraint is to avoid collision and ensure that TSVs with crosstalk are not assigned to the same test iteration: 。 7. The TSV test grouping method based on neural network according to claim 6, characterized in that: S5 is as follows: Update the neural network weights to reduce the loss function. The weight update adopts the gradient descent method, as shown in formula (6), where represents the learning rate; ， After the update is completed, the loss function is recalculated; Repeat the operation of updating the weights and the loss function multiple times until the weights are stable and the output data of the neural network is stable.

8. The TSV test grouping method based on neural network according to claim 7, characterized in that: S6 is specifically: Determine whether the updated grouping scheme meets the test constraints of TSV, that is, any two TSVs with a crosstalk relationship are not grouped into the same test group; If the grouping scheme is correct, return to S1 and update the maximum number of groups to ; If the grouping scheme is incorrect, repeat steps S3-S6 q times; if the correct solution is not found after multiple attempts, update the minimum number of groups ; like , which means finding the minimum number of groups. The minimum number of groups is , otherwise go to S1.

Images