DRC automatic repair strategy for standard cell layout

Abstract

The invention discloses a DRC automatic repair strategy for a standard cell layout, and relates to electric digital data processing, and the strategy comprises the following steps: identifying a DRC violation in the standard cell layout; extracting graphic information of a GDS file of the standard unit layout, and taking the image information as a current solution x; and dynamically adjusting a temperature attenuation coefficient of a simulated annealing algorithm according to the current solution x and the number of iterations, so as to optimize the current solution x, and taking an optimal solution obtained through optimization as a DRC violation repair parameter of the standard cell layout to repair the DRC violation. The method has the remarkable advantages in the aspects of efficiency, accuracy, stability and the like, the defects of a traditional manual repairing mode can be effectively overcome, and powerful support is provided for integrated circuit layout design.

Description

Technical Field

[0001] The present invention relates to electrical digital data processing, and more particularly to a DRC automatic repair strategy oriented to standard cell layout. Background Art

[0002] Standard cell libraries are a crucial building block for digital integrated circuits. Layout design is the most complex and time-consuming part of their construction. This is primarily due to the complex set of design rules that must be followed, each of which is interconnected and constrained. Currently, existing solutions for resolving DRC violations in standard cell layouts mostly employ manual methods, relying primarily on engineers' experience to identify and manually adjust each violation point in the layout. This approach is not only inefficient for large-scale layouts but is also susceptible to human interference, resulting in inconsistent repair quality. Summary of the Invention

[0003] The technical problem to be solved by this invention is to address the shortcomings of existing technologies and provide an automatic DRC repair strategy for standard cell layouts. This strategy aims to effectively overcome the many difficulties of DRC violations in standard cell layout design. Through the innovative application of an improved simulated annealing algorithm, it achieves efficient and accurate automatic repair of DRC violations in standard cell layouts. This strategy can intelligently identify various complex violations and fully consider the overall structure of the layout and various design rules during the repair process, ensuring that while eliminating DRC violations, the original performance and functionality of the layout are maintained to the greatest extent possible, providing a more reliable and efficient solution for standard cell layout design.

[0004] The present invention discloses an automatic DRC repair strategy for a standard cell layout. The strategy comprises the following steps: identifying DRC violations in the standard cell layout; extracting graphic information from the GDS file of the standard cell layout and using the image information as the current solution x; dynamically adjusting the temperature attenuation coefficient of the simulated annealing algorithm according to the current solution x and the number of iterations to optimize the current solution x, and using the optimal solution obtained as the DRC violation repair parameter of the standard cell layout to repair the DRC violation.

[0005] Preferably, the specific steps of dynamically adjusting the temperature attenuation coefficient of the simulated annealing algorithm are:

[0006] S101, initializing the initial temperature T0 and the termination temperature T of the simulated annealing algorithm end , initial temperature attenuation coefficient α0, maximum number of iterations iter max ;

[0007] S102, randomly generate an initial solution x0;

[0008] S103, calculating the objective function value f(x0) of the initial solution x0;

[0009] S104, assuming the initial solution x0 is the historical optimal solution x best , and initialize the historical optimal solution x best The objective function of

[0010] S105: Determine whether the current temperature T is greater than the termination temperature T end And whether the current number of iterations iter is less than the maximum number of iterations iter max If yes, proceed to S106; otherwise, perform DRC violation repair judgment;

[0011] S106, according to the initial temperature attenuation coefficient α0, the current number of iterations iter, the maximum number of iterations iter max 、Historical optimal solution x best The objective function calculates the new temperature attenuation coefficient α new ; At the same time, the current temperature T is updated to α new ×T;

[0012] S107, heuristically generate domain solution x';

[0013] S108, calculating the objective function value f(x') of the domain solution x' and the objective function value f(x) of the current solution x;

[0014] S109, determining whether to accept the domain solution x' as a new current solution based on the objective function value f(x') of the domain solution x' and the objective function value f(x) of the current solution x, and if accepted, executing S110;

[0015] S110, determine whether the objective function value f(x') of the domain solution x' is less than the historical optimal solution x best The objective function value f(x best ), if yes, execute S111; otherwise execute S112;

[0016] S111, the historical optimal solution x best And its objective function value f(x best ) is updated to the domain solution x' and its objective function value f(x');

[0017] S112, the historical optimal solution x best The minimum value of the objective function is assigned to f min =min(f min ,f(x')), the maximum value is assigned to f max =max(f max ,f(x'));

[0018] S113, calculate the new solution acceptance probability P of the new current solution, and determine whether the new current solution is better than the historical optimal solution x best Difference, and judge whether the acceptance probability P of the new solution is less than the set minimum probability threshold; if the new current solution is better than the historical optimal solution x best If the new solution acceptance probability P is less than the minimum probability threshold, then the historical optimal solution x is restored. best is the current solution and is taken as the optimal solution, then the current number of iterations iter is updated and S105 is returned; otherwise, the domain solution x' is taken as the optimal solution, the current number of iterations iter is updated and S105 is returned.

[0019] Preferably, in S105, the DRC violation repair judgment is specifically:

[0020] Check the repair result of the DRC violation. If the DRC violation still exists, execute S101. If the DRC violation does not exist, end the repair of the DRC violation.

[0021] Preferably, in S104, the historical optimal solution x is initialized. best The objective functions include:

[0022] The calculated objective function value f(x0) of the initial solution x0 is assigned to the historical optimal solution x best The objective function value f(x best ), historical optimal solution x best The objective function value f(x best ) minimum value f min and the maximum value f max .

[0023] Preferably, in S106, the new temperature attenuation coefficient α new Calculated by the following formula:

[0024]

[0025] Where f(x)=f(x0); β and λ are adjustment parameters.

[0026] Preferably, in S107, the method for heuristically generating the domain solution x' is:

[0027] The number of conflicts between each graphic and other graphics in the standard cell layout is calculated, the graphics are sorted from large to small according to the number of conflicts, the graphic with the largest number of conflicts is selected for position adjustment, and a domain solution x' is generated.

[0028] Preferably, the method for calculating the number of conflicts is:

[0029] Obtain the distance d between two figures ij , set a minimum distance constraint d min , if d ij < d min , it is determined that there is a conflict between these two images, and the conflict count of both figures is incremented by 1.

[0030] Preferably, in S108, the objective function value f(x) of the current solution x is calculated by the following formula:

[0031]

[0032] In the formula, n is the number of figures in the standard cell layout, and d ij is the distance between figure i and figure j.

[0033] Preferably, in S109, the specific method for determining whether to accept the neighborhood solution x' as the new current solution is:

[0034] If f(x') < f(x), then accept the neighborhood solution x' as the new current solution;

[0035] If f(x') ≥ f(x), then calculate the new solution acceptance probability P and generate a random number γ at the same time; if γ < P, then accept the neighborhood solution x' as the new current solution; otherwise, the current solution x remains unchanged.

[0036] Preferably, the calculation formula for the new solution acceptance probability P is:

[0037]

[0038] Beneficial effects

[0039] The advantages of the present invention are as follows:

[0040] 1. The repair efficiency is greatly improved:

[0041] The present invention adopts an automated repair strategy and uses an improved simulated annealing algorithm to quickly identify and handle violation problems. The algorithm efficiently searches for the optimal solution in the solution space through an adaptive temperature decay strategy and a heuristic neighborhood solution generation method, greatly shortening the repair time and significantly improving the repair efficiency.

[0042] 2. The repair accuracy is higher:

[0043] This method automatically identifies various complex violations and ensures accurate repairs based on precise objective function calculations and strict solution acceptance criteria. For example, by calculating the number of conflicts between graphics and prioritizing adjustments to severely conflicting graphics, it can more effectively address critical issues and effectively avoid negative impacts on other parts of the layout during the repair process, ensuring that the repaired layout is more compliant with design rule requirements and improving repair accuracy.

[0044] 3. Reduce human interference:

[0045] The automatic repair strategy of the present invention is not affected by human emotions, fatigue and other factors. As long as the algorithm parameters are set reasonably, it can run stably according to established rules and processes. Each repair follows the same scientific method, thereby ensuring the stability and reliability of the repair results and providing more reliable protection for standard cell layout design. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flow chart of the DRC automatic repair strategy of the present invention;

[0047] Figure 2 Flow chart of the improved simulated annealing algorithm of the present invention;

[0048] Figure 3 Schematic diagram of repairing DRC violations in a standard cell layout according to the present invention. DETAILED DESCRIPTION

[0049] The present invention will be further described below in conjunction with the embodiments, but this does not constitute any limitation to the present invention. Any limited number of modifications made by anyone within the scope of the claims of the present invention are still within the scope of the claims of the present invention.

[0050] See Figure 1 The present invention provides a DRC automatic repair strategy for standard cell layout, which mainly includes the following three steps.

[0051] S1. Identify DRC violations in standard cell layout.

[0052] In the standard cell layout design, design rule checking (DRC) is a key step to ensure that the layout meets the manufacturing process requirements. Professional automated tools (such as Calibre) are used to perform comprehensive DRC violation checks on the standard cell layout. These tools will check the graphic elements in the layout one by one according to a series of pre-set design rules (such as line width, spacing, coverage, etc.). After identifying all violation types, they are classified. Common violation types include: (1) Spacing violation: The distance between graphics is less than the minimum spacing specified by the design rules. (2) Line width violation: The width of the wire is less than or greater than the range allowed by the design rules. (3) Area violation: The area of the graphic does not meet the requirements of the design rules. Figure 3 As shown in the figure, (a) the distance between the contact hole and the metal; (b) the distance between metal and metal; (c) the distance between the contact hole and the active area.

[0053] S2. Extract the graphic information from the GDS file of the standard cell layout and use the image information as the current solution x. Specifically, the GDS file of the standard cell layout is viewed using Klayout software, and the graphic information is extracted to form specific data point coordinates. This data point coordinate form is used as the current solution x to meet the input parameter requirements of the simulated annealing algorithm.

[0054] S3. Dynamically adjust the temperature attenuation coefficient of the simulated annealing algorithm according to the current solution x and the number of iterations to optimize the current solution, and use the optimal solution as the DRC violation repair parameter of the standard cell layout to repair the DRC violation.

[0055] The simulated annealing algorithm mimics the solid annealing process. By randomly searching the solution space, it accepts inferior solutions with a certain probability, thus avoiding local optimal solutions and ultimately finding the global optimal solution. In standard cell layout repair, the goal is to minimize conflicts between graphics and ensure that the distances between all graphics meet the user-defined minimum distance constraint.

[0056] In the traditional simulated annealing algorithm, the temperature decreases linearly according to a fixed attenuation coefficient during the search process, and the generation of neighborhood solutions is completely random. In the standard cell layout repair problem, this fixed temperature attenuation and random neighborhood solution generation method may lead to slow algorithm convergence or a high risk of falling into a local optimal solution in the later stages of the search. Therefore, the present invention proposes an improved simulated annealing algorithm that adopts an adaptive temperature attenuation strategy and dynamically adjusts the temperature attenuation coefficient according to the current solution x and the number of search iterations, thereby better automatically repairing DRC errors in standard cells.

[0057] For improved simulated annealing algorithms, such as Figure 2 As shown, the specific implementation steps are as follows:

[0058] S101, initialize the initial temperature T0 and the termination temperature T of the simulated annealing algorithm end , initial temperature attenuation coefficient α0, maximum number of iterations iter max For example, T0 is set to 100, T end Set to 0.1, α0 to 0.8-0.99, iter max Set to 100-1000, and the number of iterations L at each temperature is set to 10-100.

[0059] Here is an explanation of the important parameters:

[0060] Initial temperature T0: The initial temperature is a key parameter in the simulated annealing algorithm, determining the randomness of the algorithm's initial search. A higher initial temperature allows the algorithm to search more extensively in the solution space, avoiding premature regression into local optima. Generally, an appropriate initial temperature can be selected based on the scale and complexity of the problem. For example, for a smaller layout repair problem, T0 can be set to 100.

[0061] Termination temperature T end : The termination temperature indicates the condition for the algorithm to stop searching. When the temperature drops below the termination temperature, the algorithm considers that a better solution has been found and stops iterating. end Set it to a small value, such as 0.1.

[0062] Temperature attenuation coefficient α: The temperature attenuation coefficient controls the speed at which the temperature decreases. Its value is usually between 0 and 1. For example, α = 0.9 means that the temperature decreases to 90% of the original temperature after each iteration.

[0063] The number of iterations at each temperature, L: At each temperature, the algorithm performs L iterations to fully explore the solution space at the current temperature. The value of L can be adjusted based on the complexity of the problem and is generally set between 10 and 100.

[0064] S102: Randomly generate an initial solution x0. The initial solution x0 represents a layout of a pattern in a standard cell layout. The initial solution can be generated by randomly moving or rotating the pattern.

[0065] S103. Calculate the objective function value f(x0) of the initial solution x0.

[0066] S104, let the initial solution x0 be the historical optimal solution x best , and initialize the historical optimal solution x best In this step, the historical optimal solution x is initialized. best The objective functions include:

[0067] The objective function value f(x0) of the calculated initial solution x0 is assigned to the historical optimal solution x best The objective function value f(x best ), historical optimal solution x best The objective function value f(x best ) minimum value f min and the maximum value f max That is: x best =x0, f(x best )=f(x0),f min =f(x0),f max =f(x0).

[0068] S105: Determine whether the current temperature T is greater than the termination temperature T end And whether the current number of iterations iter is less than the maximum number of iterations iter max If so, proceed to S106; otherwise, perform DRC violation repair judgment to determine whether the DRC violation is completely repaired.

[0069] S106, adaptive temperature attenuation: according to the initial temperature attenuation coefficient α0, the current number of iterations iter, the maximum number of iterations iter max 、Historical optimal solution x best The objective function calculates the new temperature attenuation coefficient α new .

[0070] In this embodiment, the new temperature attenuation coefficient α new Calculated by the following formula:

[0071]

[0072] Where, f(x)=f(x0); β and λ are adjustment parameters, and their values are between 0.5-2.

[0073] Then update the current temperature T to α new ×T, that is, T=α new T.

[0074] S107, heuristically generate a domain solution x'. The method for generating the domain solution x' is as follows:

[0075] First, calculate the number of conflicts between each graphic in the standard cell layout and other graphics. For two graphics i and j, get the distance d between them. ij , if d ij <d min (d minIf it is the user-defined minimum distance constraint), it is considered that there is a conflict between these two graphics, and the conflict counts of graphics i and j are incremented by 1 respectively. Then, the graphics are sorted in descending order according to the conflict counts. In generating the neighborhood solution, the graphic with the most conflict counts is preferentially selected for position adjustment to generate the neighborhood solution x'. For example, a random small offset, such as a random number between [-0.1, 0.1], is added to the x and y coordinates of the graphic with the most conflict counts to achieve position adjustment.

[0076] S108. Calculate the objective function value f(x') of the neighborhood solution x' and the objective function value f(x) of the current solution x. In this embodiment, the objective function value f(x) is defined as the number of conflicts where the minimum distance constraint is not satisfied between the graphics:

[0077]

[0078] In the formula, n is the number of graphics in the standard cell layout, and d ij is the distance between graphic i and graphic j.

[0079] S109. Acceptance of the solution: Determine whether to accept the neighborhood solution x' as the new current solution based on the objective function value f(x') of the neighborhood solution x' and the objective function value f(x) of the current solution x. If accepted, execute S110.

[0080] Specifically, the specific method for determining whether to accept the neighborhood solution x' as the new current solution is as follows:

[0081] If f(x') < f(x), it means the new solution is better than the current solution, then accept the neighborhood solution x' as the new current solution, that is, update the current solution x to x = x';

[0082] If f(x') ≥ f(x), then calculate the new solution acceptance probability Meanwhile, generate a random number γ; where the value of γ is any random number between 0 and 1. If γ < P, then accept the neighborhood solution x' as the new current solution, that is, update the current solution x to x = x'; otherwise, the current solution x remains unchanged.

[0083] S110. Determine whether the objective function value f(x') of the neighborhood solution x' is less than the objective function value f(x best of the historical optimal solution x best ). If so, it means the new solution is the historical optimal solution, and execute S111; otherwise, execute S112.

[0084] S111. Update the historical optimal solution x best and its objective function value f(x best ) to the neighborhood solution x' and its objective function value f(x'), that is, x best = x', f(x best)=f(x').

[0085] S112, the historical optimal solution x best The minimum value of the objective function is assigned to f min =min(f min ,f(x')), the maximum value is assigned to f max =max(f max ,f(x')).

[0086] After assigning the current solution x through the above steps, we obtain a new current solution. The new current solution may be the original current solution x or the domain solution x'. However, whether the new current solution is the optimal solution still needs further comparison and determination.

[0087] S113. First, we need to determine whether the objective function value of the new current solution is better than the historical optimal solution x best The objective function value f(x best ) difference, that is, if f(x)≥f(x best ), it means that the new current solution is better than the historical optimal solution x best Then we need to judge the relationship between the acceptance probability P of the new solution and the set minimum probability threshold. If f(x)≥f(x best ), and if the acceptance probability P of the new solution is less than the set minimum probability threshold, then the historical optimal solution x is restored best is the current solution, and the historical optimal solution x is restored best The current solution is taken as the optimal solution, and then the current iteration number iter is updated, that is, iter = iter + 1, and the process returns to S105. best ) and the acceptance probability P of the new solution is less than the set minimum probability threshold. If either of the two judgment conditions is not met, the domain solution x' is used as the optimal solution, the current iteration number iter is updated, that is, iter = iter + 1, and the process returns to S105. After each iteration, the randomness of the algorithm gradually decreases, allowing the algorithm to gradually converge to a better solution.

[0088] In the above improved simulated annealing algorithm, the termination condition is: when T≤T end or iter ≥ iter max The algorithm terminates when . That is, in the above S105, if the previous temperature T is not greater than the termination temperature T end And the current number of iterations iter is not less than the maximum number of iterations iter max, the algorithm terminates and the DRC violation repair process continues. This process involves checking the DRC violation repair results. If a DRC violation still exists, the process proceeds to S101. If no DRC violation exists, the DRC violation repair process ends. In practical applications, a maximum number of iterations can be set to prevent the algorithm from falling into an infinite loop.

[0089] In this embodiment, the time complexity of the improved simulated annealing algorithm mainly depends on the number of iterations and the computational complexity of the objective function. In each iteration, the objective function value and the number of conflicts of each graph need to be calculated. The computation of the objective function needs to traverse all graph pairs, and the time complexity is O(n 2 ), where n is the number of graphs. The total number of iterations depends on the initial temperature, the end temperature and the maximum number of iterations, which is generally O(iter max ). Therefore, the overall time complexity of the improved simulated annealing algorithm is O(n 2 iter max ). In terms of space complexity, the main space overhead lies in storing graphic information, the number of conflicts, and the historical optimal solution, so the space complexity is O(n), where n is the number of graphics. It can be seen that the method of dynamically adjusting the temperature attenuation coefficient based on the current solution and the number of search iterations in the present invention is the key to distinguishing it from traditional algorithms and improving the effect of layout DRC repair; and by prioritizing the position of graphics with the number of conflicts to generate neighborhood solutions, it helps to improve the efficiency and accuracy of the repair algorithm.

[0090] The above is only a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, several modifications and improvements can be made without departing from the structure of the present invention. These modifications and improvements will not affect the effect of the implementation of the present invention and the practicality of the patent.

Claims

1. A DRC automatic repair strategy for standard cell layout, characterized by: The strategy is as follows: identify DRC violations in the layout of the standard cell; extract the graphic information of the GDS file of the layout of the standard cell, and use the graphic information as the current solution x; Dynamically adjust the temperature decay coefficient of the simulated annealing algorithm according to the current solution x and the number of iterations to optimize the current solution x, and use the optimal solution obtained by optimization as the DRC violation repair parameter of the layout of the standard cell to repair the DRC violation.

2. The DRC automatic repair strategy for standard cell layout according to claim 1, characterized in that: The specific steps for dynamically adjusting the temperature decay coefficient of the simulated annealing algorithm are as follows: S101, initializing the initial temperature T0 and the termination temperature T of the simulated annealing algorithm end , initial temperature attenuation coefficient α0, maximum number of iterations iter max ; S102. Randomly generate an initial solution x0; S103. Calculate the objective function value f(x0) of the initial solution x0; S104, assuming the initial solution x0 is the historical optimal solution x best , and initialize the historical optimal solution x best The objective function of S105: Determine whether the current temperature T is greater than the termination temperature T end And whether the current number of iterations iter is less than the maximum number of iterations iter max If yes, proceed to S106; otherwise, perform DRC violation repair judgment; S106, according to the initial temperature attenuation coefficient α0, the current number of iterations iter, the maximum number of iterations iter max 、Historical optimal solution x best The objective function calculates the new temperature attenuation coefficient α new ; At the same time, the current temperature T is updated to α new ×T; S107. Heuristically generate a neighborhood solution x'; S108. Calculate the objective function value f(x') of the neighborhood solution x' and the objective function value f(x) of the current solution x; S109. Decide whether to accept the neighborhood solution x' as the new current solution according to the objective function value f(x') of the neighborhood solution x' and the objective function value f(x) of the current solution x. If accepted, execute S110; S110, determine whether the objective function value f(x') of the domain solution x' is less than the historical optimal solution x best The objective function value f(x best ), if yes, execute S111; otherwise execute S112; S111, the historical optimal solution x best And its objective function value f(x best ) is updated to the domain solution x' and its objective function value f(x'); S112, the historical optimal solution x best The minimum value of the objective function is assigned to f min =min(f min ,f(x')), the maximum value is assigned to f max =max(f max ,f(x')); S113, calculate the new solution acceptance probability P of the new current solution, and determine whether the new current solution is better than the historical optimal solution x best Difference, and judge whether the acceptance probability P of the new solution is less than the set minimum probability threshold; if the new current solution is better than the historical optimal solution x best If the new solution acceptance probability P is less than the minimum probability threshold, then the historical optimal solution x is restored. best is the current solution and is taken as the optimal solution, then the current number of iterations iter is updated and S105 is returned; otherwise, the domain solution x' is taken as the optimal solution, the current number of iterations iter is updated and S105 is returned.

3. The DRC automatic repair strategy for standard cell layout according to claim 2, characterized in that: In S105, the DRC violation repair judgment is specifically as follows: Check the repair result of the DRC violation. If the DRC violation still exists, execute S101. If the DRC violation does not exist, end the repair of the DRC violation.

4. The DRC automatic repair strategy for standard cell layout according to claim 2, characterized in that: In S104, the historical optimal solution x is initialized. best The objective functions include: The calculated objective function value f(x0) of the initial solution x0 is assigned to the historical optimal solution x best The objective function value f(x best ), historical optimal solution x best The objective function value f(x best ) minimum value f min and the maximum value f max .

5. The DRC automatic repair strategy for standard cell layout according to claim 4, characterized in that: In S106, the new temperature attenuation coefficient α new Calculated by the following formula: In the formula, f(x) = f(x0); both β and λ are adjustment parameters.

6. The DRC automatic repair strategy for standard cell layout according to claim 2, characterized in that: In S107, the method for heuristically generating the neighborhood solution x' is as follows: Calculate the number of conflicts between each graphic in the layout of the standard cell and other graphics, sort the graphics in descending order according to the number of conflicts, and select the graphic with the most conflicts for position adjustment to generate the neighborhood solution x'.

7. The DRC automatic repair strategy for standard cell layout according to claim 6, characterized in that: The calculation method of the number of conflicts is as follows: Get the distance d between two figures ij , set a minimum distance constraint d min , if d ij <d min , then it is determined that there is a conflict between the two images, and the number of conflicts of the two graphics is increased by 1.

8. The DRC automatic repair strategy for standard cell layout according to claim 7, characterized in that: In S108, the objective function value f(x) of the current solution x is calculated by the following formula: Where n is the number of graphics in the standard cell layout, d ij is the distance between graph i and graph j.

9. The DRC automatic repair strategy for standard cell layout according to claim 2, characterized in that: In S109, the specific method for judging whether to accept the neighborhood solution x' as the new current solution is as follows: If f(x') < f(x), accept the neighborhood solution x' as the new current solution; If f(x') ≥ f(x), calculate the new solution acceptance probability P, and at the same time generate a random number γ; if γ < P, accept the neighborhood solution x' as the new current solution; otherwise, the current solution x remains unchanged.

10. The DRC automatic repair strategy for standard cell layout according to claim 2 or 9, characterized in that: The calculation formula of the new solution acceptance probability P is:

Images