Macro layout increment optimization method based on time sequence optimization

Through the Macro layout incremental optimization method based on timing optimization, detect and move defective macro to chip boundary, optimize line length and build constraint diagrams, the problem of inability to take into account both line length and time sequence optimization in the existing technology is solved, and better Macro layout results and timing performance.

CN119047404APending Publication Date: 2024-11-29INSTITUTE OF ELECTRONIC DESIGN AUTOMATION PEKINGUNIVERSITY WUXI
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410940389.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing macro layout optimization solution cannot take into account the optimization of line length and timing, which affects the final performance of chip physical design.

Method used

Adopt the Macro layout incremental optimization method based on timing optimization. By establishing a kd-tree to detect defective macros, moving it to the chip boundary, optimizing line length, and building a constraint diagram for rational solution, ensuring the relative positional relationship and legal layout between macros result.

Benefits of technology

To maximize the time series optimization effect achieved by mixed-size layout, while reducing congestion, it is suitable for layout results of different layouters, solving the problem of macro size diversity.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119047404A_ABST
    Figure CN119047404A_ABST
Patent Text Reader

Abstract

The invention discloses a Macro layout increment optimization method based on time sequence optimization, and the method comprises the following steps: building a kd-tree according to the layout information of a place prototype, and detecting a defect macro by using the kd-tree; moving the defect macro, optimizing the line length and moving the defect macro to the boundary of the chip; the method comprises the following steps of: constructing a constraint graph according to a relative position relationship between macro in a planar prototype, and carrying out reasonable solution according to the constraint graph to obtain a legitimate macro layout result. According to the Macro layout increment optimization method based on time sequence optimization provided by the invention, after the hybrid size layout is carried out to obtain the time sequence optimized placement type, the macro causing congestion is circularly moved to the boundary of the chip, and meanwhile, the relative position relationship between the macro in the placement type is maintained, and proxy targets such as line length are optimized, so that the probability of congestion of the chip is reduced, and the service life of the chip is prolonged. The method can be integrated into different layout processes, and can be used for optimizing layout results of different layout devices (black boxes or white boxes).
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of integrated circuit design automation, and in particular to a Macro layout incremental optimization method based on timing optimization. Background Art

[0002] Placement is an important step in the field of chip physical design. This step involves solving the positions of the chip's standard cells and macros (usually IP cores, SRAM, etc.) to meet layout legalization constraints (i.e., there should be no overlap between chip devices) and optimize line length and other goals. Usually in layout tasks, although the number of macros accounts for a small proportion of all chip devices, the macro area is huge, and the total area of ​​macros often accounts for a large proportion of the area of ​​all devices. Through research, it is found that the placement of macros has a great impact on the overall quality of chip layout. In traditional analytical layout methods, macros are often placed in the center of the chip, causing congestion problems in standard cell layout and signal routing, and negatively affecting the final timing performance. Improving the layout optimization solution for macros has always been one of the bottlenecks faced by placers.

[0003] The three-stage macro placement algorithm is widely used in macro-intensive chip layout tasks. The algorithm is divided into three steps: the first step is mixed-size placement, which uses a global layouter and analytical placement strategies to uniformly place macros and standard cells; the second step, macro placement, optimizes the coordinates of the macro based on the results of the first step, and reduces congestion by moving the macro to the chip boundary; the third step, standard cell placement, places standard devices and obtains legal device locations based on the macro coordinates. Then, the second step (macro placement) of the algorithm usually destroys the line length and timing optimization achieved in the first step (mixed-size placement) when moving the macro to the edge of the chip, affecting the final performance of the chip physical design. Summary of the invention

[0004] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is that the existing macro layout optimization scheme cannot take into account the optimization of line length and timing, which affects the final performance of the chip physical design. Therefore, the present invention provides a Macro layout incremental optimization method based on timing optimization. After the mixed-size placement (mixed-size placement) obtains the placement prototype with timing optimization, the macros that cause congestion are cyclically moved to the boundary of the chip, while maintaining the relative position relationship between macros in the placement prototype and optimizing proxy targets such as line length. The method of the present invention can be integrated into different layout processes and can be used to optimize the layout results of different layouters (black box or white box).

[0005] To achieve the above object, the present invention provides a Macro layout incremental optimization method based on timing optimization, comprising the following steps:

[0006] According to the layout information of the placement prototype, a kd-tree is established, and the defective macro is detected using the kd-tree;

[0007] Move the defect macro, optimize the line length and move it to the edge of the chip;

[0008] A constraint graph is constructed based on the relative position relationship between macros in the placement prototype, and a rational solution is performed based on the constraint graph to obtain a legal macro layout result without overlap between macros. Since the constraint graph constructed by the placement prototype is followed in the rational solution process, the relative position relationship between macros is preserved, and the macros are moved to the chip boundary.

[0009] Furthermore, based on the layout information of the placement prototype, a kd-tree is established and the defective macro is detected using the kd-tree, including:

[0010] Define macros that are distributed in the center of the chip and may cause congestion as defective macros, and the rest of the macros that are distributed at the edge of the chip and will not cause congestion as normal macros;

[0011] Calculate the distance between each macro and the nearest boundary;

[0012] Use kd-tree to search the four macros closest to it, and determine whether these four neighbors can completely surround the macro to determine whether it is a defective macro.

[0013] Furthermore, the defect macro is moved to optimize the line length and move it to the boundary of the chip, and the line length and boundary are used as optimization targets. The line length target is formulated using a weighted-average line length model.

[0014] Furthermore, for net e, the formula of WA line length model is:

[0015]

[0016] where v i Indicates a pin in nete, x i and i Respectively represent v i The horizontal and vertical coordinates of .

[0017] Furthermore, the model used when optimizing the boundary is as follows:

[0018]

[0019] in, and are the horizontal and vertical boundary items of the macro respectively; w core and h core Represent the width and height of the chip respectively.

[0020] Furthermore, a constraint graph is constructed according to the relative position relationship between macros in the placement prototype, and a rational solution is performed according to the constraint graph to obtain a legal macro layout result in which there is no overlap between macros, specifically including:

[0021] Create horizontal and vertical constraint graphs G based on the layout result of the placement prototype h and G v ,

[0022] Formulate linear programming subject to constraint graphs;

[0023] The first optimization objective, the second optimization objective, and the combined constraints are set to minimize the displacement of the macro before and after legalization and minimize the distance between the macro and the nearest boundary.

[0024] Technical Effects

[0025] The invention provides a Macro layout incremental optimization method based on timing optimization, which can meet two requirements in the process of macro layout: first, the macro is moved to the chip boundary as much as possible to reduce congestion, and second, the result of macro layout will maintain the relative position relationship between macros in the placement prototype, thereby retaining the timing optimization effect achieved by the mixed-size layout to the greatest extent. The method of the invention can solve the problem of macro size diversity in chip design. Even if a design contains macros with very different sizes, the method of the invention can still obtain a macro layout result with better effect. In addition, the method of the invention is not strongly coupled with any placer and can be integrated into different layout processes as an independent tool.

[0026] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings to fully understand the purpose, characteristics and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a flow chart of a Macro layout incremental optimization method based on timing optimization according to a preferred embodiment of the present invention;

[0028] Figure 2 It is a specific implementation diagram of defective macro detection of a Macro layout incremental optimization method based on timing optimization in a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the technical problems, technical solutions and beneficial effects to be solved by 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.

[0030] In the following description, specific details such as specific internal procedures and techniques are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0031] like Figure 1 As shown, the present invention provides a Macro layout incremental optimization method based on timing optimization, comprising the following steps:

[0032] Step 100, based on the layout information of the placement prototype, a kd-tree is established, and defective macros are detected using the kd-tree; macros that are distributed in the center of the chip and may cause congestion problems are defined as defective macros, and the rest of the macros that are arranged at the edge of the chip and will not cause congestion are defined as normal macros;

[0033] Calculate the distance between each macro and the nearest boundary;

[0034] Using the spatial characteristics of kd-tree, a recursive search method is used to perform nearest neighbor search: first, the nearest neighbor is initialized as the root node, and the maximum distance is infinite; then recursively perform the following steps:

[0035] 1. Calculate the distance between the query point and the current node. If this distance is closer than the nearest neighbor distance, update the nearest neighbor.

[0036] 2. Based on the coordinates of the query point at the current latitude, decide whether to recursively search the left subtree or the right subtree.

[0037] Using the nearest neighbor search function of kd-tree, search for the four macros closest to it, and determine whether these four neighbors can completely surround the macro to determine whether it is a defective macro.

[0038] A normal macro satisfies at least one of the following two conditions:

[0039] 1. The macro fits snugly around the edge of the chip.

[0040] 2. The four nearest macro neighbors of the macro can completely surround the macro. That is, the four neighbors of the macro are respectively above, below, to the left and to the right of the macro.

[0041] If a macro does not meet both of the above conditions, it will be identified as a defective macro. First, for the judgment of the first condition, we can calculate the distance between each macro and the nearest boundary to make a judgment. For the judgment of the second condition, we use the nearest neighbor search feature of kd-tree to search the four nearest macros for each macro using kd-tree, and judge whether these four neighbors can completely surround the macro.

[0042] Step 200, based on the differentiable wirelength model and the periphery model, the gradient of the horizontal and vertical movement of the macro is calculated, based on the gradient descent strategy, the coordinates of the defective macro are updated with the calculated gradient, the defective macro is moved, the wirelength is optimized and moved to the boundary of the chip, and the wirelength and the boundary are used as optimization targets, wherein the wirelength target is formulated using a weighted-average wirelength model.

[0043] For net e, the formula for the WA line length model is:

[0044]

[0045] where v i Indicates a pin in nete, x i and i Respectively represent v i The horizontal and vertical coordinates of .

[0046] The model used when optimizing the boundary is as follows:

[0047]

[0048] in, and are the horizontal and vertical boundary items of the macro respectively; w core and h core Represents the width and height of the chip respectively. The closer the macro is to the chip boundary, the smaller the boundary item overhead.

[0049] Step 300, constructing a constraint graph according to the relative position relationship between macros in the placement prototype, performing a rational solution according to the constraint graph, and obtaining a legal macro layout result without overlapping between macros.

[0050] A constraint graph is constructed based on the relative position relationship between macros in the placement prototype, and a rational solution is performed based on the constraint graph to obtain a legal macro layout result without overlap between macros, including:

[0051] Create horizontal and vertical constraint graphs G based on the layout result of the placement prototype h and G v ,

[0052] Formulate the linear program subject to the constraints graph:

[0053]

[0054] For each macro i , x i / x i ′ Describe the horizontal axis (y before and after the macro is legalized) i / y i ′ The definition of is the same), w i / h i Indicates macro i Width / height of the macro. i , introduce four variables Assist in formula formulation.

[0055] Set the first optimization goal, the second optimization goal and the combined constraints to minimize the displacement of macro before and after legalization and minimize the distance between macro and the nearest boundary. Specifically, the first optimization goal of this linear programming is Combined Constraints The purpose of this optimization goal is to minimize the displacement of the macro before and after legalization. The second optimization goal is min(x i ,Wx i )+min(y i ,Hy i ), aims to minimize the distance between the macro and the nearest boundary.

[0056] constraint and Ensure that there is no overlap between the legalized macros, and constrain It ensures that the macro after legalization does not exceed the chip boundary.

[0057] like Figure 1 and Figure 2 As shown, the present invention provides a preferred embodiment. Figure 1 and Figure 2 In the layout diagram of , the purple squares represent defective macros (which may cause congestion), and the blue squares represent normal macros. The first step of the algorithm corresponds to Figure 1 kd-tree-based macro diagnosis (defect macro detection) step, Figure 2 A specific implementation diagram of defect macro detection is given: For the result of a macro layout ( Figure 2(a)), we first build a kd-tree ( Figure 2 (b); Then, using the nearest neighbor retrieval feature of kd-tree, we find the four nearest neighbors of each macro and determine whether the macro is defective based on its relative position to the neighboring macros (e.g. Figure 2 After the defect macro is retrieved, the second step of the algorithm (gradient-based macro shift) takes the x and y coordinates of the defect macro as variables, brings these variables into the line length model and boundary model, and then derives the x and y coordinates. The obtained gradient is directly added to the x and y coordinates to update the position of the defect macro. Figure 1 As shown in the layout diagram corresponding to the gradient-based macroshift, the defective macros will overlap with each other after the coordinates are updated by the gradient-based method. Finally, the constraint-graph-based lp for macro legalization step will establish horizontal and vertical constraint graphs based on the relative position relationship between the macros, and list a linear programming expression based on the constraint graph. The optimization target of this expression is the macro displacement and the distance of the macro from the chip boundary. In the macro layout solution obtained in this step, there is no overlap between the macros and they will be moved to the chip boundary.

[0058] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A Macro layout incremental optimization method based on timing optimization, characterized in that: The following steps are involved: According to the layout information of the placement prototype, a kd-tree is established, and the defective macro is detected using the kd-tree; Move the defect macro, optimize the line length and move it to the edge of the chip; A constraint graph is constructed according to the relative position relationship between macros in the placement prototype, and a rational solution is performed according to the constraint graph to obtain a legal macro layout result without overlap between macros.

2. The Macro layout incremental optimization method based on timing optimization according to claim 1, characterized in that: According to the layout information of the placement prototype, a kd-tree is established and the defective macro is detected using the kd-tree, including: Define macros that are distributed in the center of the chip and may cause congestion as defective macros, and the rest of the macros that are distributed at the edge of the chip and will not cause congestion as normal macros; Calculate the distance between each macro and the nearest boundary; Use kd-tree to search the four macros closest to it, and determine whether these four neighbors can completely surround the macro to determine whether it is a defective macro.

3. The Macro layout incremental optimization method based on timing optimization according to claim 1, characterized in that: Move the defect macro, optimize the line length and move it to the boundary of the chip, and use the line length and boundary as optimization targets. The line length target is formulated using the weighted-average line length model.

4. The Macro layout incremental optimization method based on timing optimization as claimed in claim 3, characterized in that: For net e, the formula for the WA line length model is: where v i Indicates a pin in nete, x i and i Respectively represent v i The horizontal and vertical coordinates of .

5. The Macro layout incremental optimization method based on timing optimization as claimed in claim 3, characterized in that: The model used when optimizing the boundary is as follows: in, and are the horizontal and vertical boundary items of the macro respectively; w core and h core Represent the width and height of the chip respectively.

6. The Macro layout incremental optimization method based on timing optimization according to claim 1, characterized in that: A constraint graph is constructed based on the relative position relationship between macros in the placement prototype, and a rational solution is performed based on the constraint graph to obtain a legal macro layout result without overlap between macros, including: Create horizontal and vertical constraint graphs G based on the layout result of the placement prototype h and G v , Formulate linear programming subject to constraint graphs; The first optimization objective, the second optimization objective, and the combined constraints are set to minimize the displacement of the macro before and after legalization and minimize the distance between the macro and the nearest boundary.

Citation Information

Cited By

  • Hierarchical transistor layout method for multiple rows of standard cells

    CN120524901A