A mixed height cell layout method to avoid circuit minimum injection area limitation

By adopting a hybrid height unit aggregation method based on the optimal region in the circuit design, combined with DLX shaping and density-aware search, the problem of increasing line length caused by unit aggregation is solved, high-quality circuit layout is achieved, and the performance of chip design is improved .

CN114997099BActive Publication Date: 2024-11-19SOUTHWEAT UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210459006.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-04-27
Publication Date
2024-11-19
Estimated Expiration
2042-04-27

AI Technical Summary

Technical Problem

When dealing with the minimum injection area limitation, the existing unit aggregation technology leads to an increase in line length and poor quality, which cannot effectively avoid the minimum injection area limitation, affecting the performance of the circuit design.

Method used

Using the hybrid height unit aggregation method based on the optimal region, the DLX shaping algorithm and legalization algorithm are combined with the multi-line height unit aggregation algorithm and density-aware search to form a cluster and perform refinement processing to ensure that the units are aggregated in the optimal region. , reduce line length and improve layout quality.

Benefits of technology

Significantly reduce or reduce wire length while maintaining high-quality circuit design, improving the performance and layout quality of computer chip-assisted design software.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114997099B_ABST
    Figure CN114997099B_ABST
Patent Text Reader

Abstract

The present invention discloses a mixed height unit layout method for avoiding the minimum injection area restriction of a circuit, comprising the steps of: inputting the global layout result of the circuit; aggregating the mixed height units with MIA violation based on the optimal area to form a cluster; using the DLX shaping algorithm to obtain the order of all units in the cluster; using the legalization algorithm to perform legalization processing; performing refinement processing, and outputting a result without MIA violation. The present invention is based on the optimal area mixed height unit aggregation, so that it will not cause a significant increase in line length and may even reduce the line length, so that the line length is minimized while the standard unit can be placed in its optimal area.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of circuit design, and in particular relates to a mixed height unit layout method for avoiding the limitation of the minimum injection area of ​​a circuit. Background Art

[0002] In advanced power-aware high-performance circuit design, applying multiple threshold voltages (VT) to standard cells is a popular approach to optimize both timing and power simultaneously. However, this approach leads to additional design constraint restrictions, namely, the minimum injection area (MIA) restriction. Cell clustering is an effective way to eliminate this limitation.

[0003] The performance of existing unit clustering technology is unsatisfactory. Although the existing unit clustering technology can handle MIA constraints, it will get a poor quality result and there will be some loss in line length. Specifically, the existing unit clustering technology is to cluster these adjacent units that violate the MIA constraint after legalization. The biggest disadvantage of this clustering technology that clusters adjacent units together is that it will cause additional quality loss, because these units are clustered together only because they are physically adjacent. Clustering these units together may cause additional line length increase. Summary of the invention

[0004] In order to solve the above problems, the present invention proposes a mixed-height cell layout method that avoids the limitation of the minimum injection area of ​​the circuit. The mixed-height cells are clustered based on the optimal area, which will not cause a significant increase in line length and may even reduce the line length. While minimizing the line length, the standard cells can be placed in their optimal area.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a mixed height cell layout method that avoids the limitation of the minimum injection area of ​​the circuit, comprising the steps of:

[0006] S10, global layout result of input circuit;

[0007] S20, aggregate the mixed height units with MIA default based on the optimal region to form a cluster; S30, use the DLX shaping algorithm to obtain the ranking of all units in the cluster;

[0008] S40, performing legalization processing using a legalization algorithm;

[0009] S50, performing refinement processing and outputting a result without MIA default.

[0010] Furthermore, in the highly aggregated stage of the mixing unit, the steps are included:

[0011] Get the optimal area of ​​all cells that cause MIA violations within the row;

[0012] According to the overlap of the optimal regions, the units with overlapping optimal regions are gathered together to form a cluster.

[0013] Furthermore, when performing mixed aggregation:

[0014] Select all cells with the same voltage type and whose defect width is smaller than the minimum injection width;

[0015] Use the multiple row height unit clustering algorithm repeatedly to build a single clustering algorithm to cluster all units to form clusters.

[0016] Furthermore, in the DLX shaping stage, the steps include:

[0017] The DLX algorithm is used to reshape the cells in each cluster and the order of all cells in the cluster is obtained so that the cluster has the smallest area.

[0018] Furthermore, in the legalization stage, the steps are included: searching around the initial position of the unit and determining it as its legalization position.

[0019] Furthermore, in the legalization stage, according to the multi-stage decision theory, each search is regarded as a decision; density perception is added to the legalization, and searches are performed sequentially from low-density areas to high-density areas, and finally all units are legalized.

[0020] Furthermore, in the refinement processing stage, the remaining inter-row MIA violations are removed by using the unit translation method, and finally a result without MIA violations is output.

[0021] Furthermore, in the unit translation method, by pre-calculating the feasible positions of the unit, the position with the smallest Manhattan displacement is selected, and a simple translation or exchange is used on the unit, and finally a result without MIA violation is output.

[0022] The beneficial effects of adopting this technical solution are:

[0023] The present invention gathers these standard units together according to the overlap of the optimal area, gathers suitable units together with the core of the optimal area, so that the gathered units are placed in their optimal area as much as possible, and then after shaping and subsequent processing, it will not cause a significant increase in line length and may even reduce the line length, while minimizing the line length and placing the standard unit in its optimal area, which will not affect the quality of the solution.

[0024] The present invention uses a unit clustering algorithm based on the optimal region to form a cluster, wherein a rectangular overlapping related unit clustering algorithm is used, the optimal region is regarded as a matrix, and a single clustering algorithm is established according to repeated use in the multiple row height unit clustering algorithm, and multiple matrices are divided into several groups, and one group constitutes a cluster, and finally the sum of the matrix distances in all groups is minimized. The present invention uses a density-aware or similar unit search algorithm to sequentially search for the final position of the unit. The present invention uses a unit search algorithm based on dynamic programming decision theory for legalization, including using coordinate sorting of the unit so that the result of each search reduces the impact on subsequent searches, thereby achieving global optimization. The present invention uses a greedy unit shifting method to solve the inter-row MIA violation algorithm with the minimum displacement. Thereby significantly improving the quality of the detailed layout of the connection, especially in terms of line length and time, the detailed layout is an important part of the entire computer chip-aided design software (EDA), which can improve the quality of the computer chip-aided design software, so that the performance of the chip designed by the computer chip-aided design software can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic flow chart of a mixed height cell layout method for avoiding the limitation of the minimum injection area of ​​a circuit according to the present invention;

[0026] Figure 2 This is a flow chart of a method for training a user tag model in an embodiment of the present invention;

[0027] Figure 3 This is a flow chart of a method for training a user tag model in an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of BOW data conversion in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described below with reference to the accompanying drawings.

[0030] In this embodiment, see Figure 1 As shown, the present invention proposes a mixed height cell layout method that avoids the limitation of the minimum injection area of ​​the circuit, comprising the steps of:

[0031] S10, global layout result of input circuit;

[0032] S20, the mixed height units with MIA default are aggregated based on the optimal area to form clusters;

[0033] S30, using the DLX shaping algorithm to obtain the ranking of all units in the cluster;

[0034] S40, performing legalization processing using a legalization algorithm;

[0035] S50, performing refinement processing and outputting a result without MIA default.

[0036] As an optimization scheme of the above embodiment, in the highly aggregated stage of the mixing unit, the steps are included:

[0037] Get the optimal area of ​​all cells that cause MIA violations within the row;

[0038] According to the overlap of the optimal regions, the units with overlapping optimal regions are gathered together to form a cluster.

[0039] Specifically, Figure 2 As shown, when mixed aggregation is performed:

[0040] Select all cells with the same voltage type and whose defect width is smaller than the minimum injection width;

[0041] Use the multiple row height unit clustering algorithm repeatedly to build a single clustering algorithm to cluster all units to form clusters.

[0042] Algorithm 1 Multiple row height cell aggregation algorithm:

[0043] Input: a set of cells C[1…] with the same voltage (VT) and the target width W of the cluster;

[0044] Output: a series of clusters U[1…];

[0045] i=1, start looping when length[C]>0, U[i]=build a single cluster (C,W);

[0046] i=i+1, end the loop; return U.

[0047] Algorithm 2 builds a single cluster algorithm:

[0048] Input: a set of cells C[1…] with the same voltage (VT) and the target W of the cluster;

[0049] Output: a cluster u;

[0050] c = Get a cell with the smallest optimal area, let the initial coordinates of the cluster be the center of the optimal area of ​​cell c, the height of the cluster hu = the height of cell c, add cell c to cluster u, and the width of the cluster wu = the area of ​​the cluster / hu;

[0051] Start looping when length[C]>0 and wu <W;

[0052] c = Get a cell from C with the smallest Manhattan distance to cluster u;

[0053] If hu < hc, set hu = hc, add cell c to cluster u, and update the area of cluster u as the sum of the area of cluster u and the area of cell c; the width of the cluster wu = the area of the cluster / hu;

[0054] End the loop;

[0055] Return u.

[0056] As an optimized solution of the above embodiment, in the DLX shaping stage, it includes the steps of:

[0057] Use the DLX algorithm to shape the cells within each cluster to obtain the sorting of all cells within the cluster, so that the cluster has the smallest area.

[0058] As Figure 3 , for example, Figure 3 in a, cluster u1 consists of cells c1, c2, and c3. Figure 3 In b, the area of cluster u1 can be divided into 6 sub-regions, namely p1, p2,..., p6. There are two placement schemes for cell c3, which are respectively represented as q c3 1 and q c3 2. Among them, q c3 1 means placing cell c3 in the region composed of sub-regions p1, p3, and p5, while q c3 2 means placing cell c3 in the region composed of p2, p4, and p6. Figure 3 c shows a matrix, where each row represents a possible placement scheme of a cell. The first 6 columns represent the 6 grid regions of cluster u1, and the last 3 columns represent the three cells. For each column, there can only be one value. As Figure 3 in c, if q c3 1 is selected, q c1 1 and q c2 1 cannot be selected because p1 will overlap. Selecting the appropriate rows (and each column must have a value) to obtain the arrangement of cells in the cluster can be solved by the exact cover problem (DLX), such as q c3 1, q c2 2, and q c1 6.

[0059] This stage aims to make all clusters have a smaller area.

[0060] As an optimized solution of the above embodiment, in the合法化 stage, it includes the steps of: Search around the initial position of the cell and determine its legalized position.

[0061] Specifically, in the legalization stage, according to the multi-stage decision theory, each search is regarded as a decision; density perception is added to the legalization, and searches are performed sequentially from low-density areas to high-density areas, and finally all units are legalized.

[0062] In the cluster-based legalization stage, the main content is legalization, and the legalization position of the unit is determined by searching. Specifically, the search is performed around the initial position of the unit, and a legal position is searched, including constraints based on row alignment, and is determined as its legalization position. According to the multi-stage decision theory, the present invention regards each search as a decision, and tries not to affect or less affect the subsequent decisions after each decision. Therefore, density perception is added to the legalization to improve the quality of the solution, and searches are performed sequentially from low-density areas to high-density areas, so that all units are finally legalized.

[0063] Figure 4 This is a simple example. The left side shows c2 before legalization, and the right side shows it after legalization.

[0064] The benefit of density awareness is that it pushes cells in high-density areas to low-density areas without significantly affecting the quality of the solution. After this stage, all cells are legalized (rows are aligned and cells do not overlap).

[0065] As an optimization solution of the above embodiment, in the refinement processing stage, the remaining inter-row MIA violations are removed by using a unit translation method, and finally a result without MIA violations is output.

[0066] Specifically, in the unit translation method, by pre-calculating the feasible positions of the unit, the position with the smallest Manhattan displacement is selected, and a simple translation or exchange is used on the unit, and finally a result without MIA violation is output.

[0067] After legalization, there will be a small number of inter-row MIA violations. Through the inter-row MIA perception refinement stage, a greedy unit translation method is used to completely solve the inter-row MIA violation problem with the minimum displacement. By pre-calculating the feasible positions of the unit, greedily selecting the position with the smallest Manhattan displacement from the original position, using simple translation or exchange for the unit, and finally outputting the result without MIA violation. The result obtained after this stage is: no inter-row MIA violation. This algorithm is simple, low in complexity and can solve all inter-row MIA refinements.

[0068] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A hybrid-height cell layout method for avoiding the minimum injection area limitation of a circuit, characterized in that, it includes the steps: S10. Input the global layout result of the circuit; S20. Cluster the hybrid-height cells with MIA violations based on the optimal area to form clusters; When performing hybrid clustering: Select all cells with the same voltage type and a width smaller than the minimum injection width; Use the repeatedly used single-cluster building algorithm in the multi-row-height cell clustering algorithm to cluster all cells to form clusters; Algorithm 1 Multi-row-height cell clustering algorithm: Input: A series of cells C[1…] with the same voltage (VT) and the target width W of the cluster; Output: A series of clusters U[1…]; i = 1, start the loop while length[C]>0, U[i] = BuildSingleCluster(C, W); i = i + 1, end the loop; Return U; Algorithm 2 BuildSingleCluster algorithm: Input: A series of cells C[1…] with the same voltage (VT) and the target W of the cluster; Output: A cluster u; c = Get a cell with the smallest optimal area, let the initial coordinates of the cluster be the center of the optimal area of cell c, the height hu of the cluster = the height of cell c, add cell c to cluster u, and the width wu of the cluster = the area of the cluster / hu; Start the loop while length[C]>0 and wu<W; c = Get a cell from C with the smallest Manhattan distance to cluster u; If hu < hc, hu = hc, add cell c to cluster u, the area of cluster u = the area of cluster u + the area of cell c; the width wu of the cluster = the area of the cluster / hu; End the loop; Return u; S30. Use the DLX shaping algorithm to obtain the sorting of all cells within the cluster; S40. Perform legalization processing using the legalization algorithm; S50. Perform refinement processing and output the result without MIA violations.

2. The hybrid-height cell layout method for avoiding the minimum injection area limitation of a circuit according to claim 1, characterized in that, in the DLX shaping stage, it includes the steps: For the cells within each cluster, use the DLX algorithm to perform shaping to obtain the sorting of all cells within the cluster so that the cluster has the smallest area.

3. The hybrid-height cell layout method for avoiding the minimum injection area limitation of a circuit according to claim 1, characterized in that, in the legalization stage, it includes the steps: Search around the initial position of the cell and determine its legalization position.

4. The hybrid-height cell layout method for avoiding the minimum injection area limitation of a circuit according to claim 3, characterized in that, in the legalization stage, according to the multi-stage decision theory, regard each search as a decision; add density awareness in legalization, and perform sequential searches from low-density areas to high-density areas, and finally all cells are legalized.

5. The hybrid-height cell layout method for avoiding the minimum injection area limitation of a circuit according to claim 1, characterized in that, in the refinement processing stage, remove the remaining inter-row MIA violations by using the cell translation method, and finally output the result without MIA violations.

6. A mixed height cell layout method for avoiding circuit minimum injection area limitation according to claim 5, It is characterized in that In the cell translation method, the feasible positions of the cells are pre-calculated, the position with the smallest Manhattan displacement is selected, and a simple translation or exchange is used on the cells to finally output a result without MIA violation.