WO2023015536A1

WO2023015536A1 - Method for placing macrocells of integrated circuit

Info

Publication number: WO2023015536A1
Application number: PCT/CN2021/112360
Authority: WO
Inventors: 焦润; 沈嘉华; 张锐
Original assignee: 华为技术有限公司
Priority date: 2021-08-12
Filing date: 2021-08-12
Publication date: 2023-02-16
Also published as: CN117561514A

Abstract

Embodiments of the present disclosure relate to a method for placing macrocells of an integrated circuit, comprising: selecting a region to be partitioned in a placement space; providing multiple candidate tangent lines in the region to be partitioned; and on the basis of initial positions of multiple macrocells of the region to be partitioned in the placement space and center-of-gravity positions of clusters obtained by clustering the multiple macrocells, selecting a target tangent line from the multiple candidate tangent lines so as to partition the region to be partitioned into two subregions, each subregion being used for placing at least one cluster of macrocells. According to the embodiments of the present disclosure, a placement region can be adaptively divided by using the target tangent line according to actual placement requirements of the macrocells in the clusters, thereby implementing parallel placement of multiple modules.

Description

Method for placing macrocells of an integrated circuit

technical field

Embodiments of the present disclosure generally relate to the technical field of computer-aided design of integrated circuits, and more particularly, relate to a method for laying out macrocells of an integrated circuit, an electronic device, a computer-readable storage medium, and a computer program product.

Background technique

As design and manufacturing processes continue to evolve, integrated circuits continue to grow in size and may even include billions of transistors. In order to reduce the complexity of the design, modern large-scale integrated circuit design often reuses a large number of IP (Intellectual property). In chip design, layout is divided into three parts: top-level layout, macrocell layout, and standard cell placement. The quality of the macro cell layout will directly affect the placement effect of subsequent standard cells, and has a greater impact on the performance, power consumption, and area (PPA) of the integrated circuit. For example, simply global placement of macrocells and standard cells can lead to severe overflow and congestion problems. In the actual product design process, the macro cells are often placed first, and then the standard cells are placed with the positions of the macro cells fixed. There are already many practical tools to realize the placement of standard units. Therefore, for this common chip design where large macrocells and standard cells coexist, an efficient macrocell layout method is very important.

The current layout of macrocells mainly relies on manual layout planning, and more practical constraints need to be considered when placing macrocells, such as macrocells need to be placed on the border of the die, distance constraints between macrocells, and design levels structure etc. The placement of macro cells is very dependent on the experience of engineers, and has a great impact on the subsequent global layout and routing. Poor macrocell placement results lead to repeated iterations, rework.

Contents of the invention

Embodiments of the present disclosure provide a method, an electronic device, a computer-readable storage medium, and a computer program product for laying out macrocells of an integrated circuit, aiming to solve the above-mentioned problems and others existing in conventional macrocell layout schemes Potential problems.

According to a first aspect of the present disclosure, there is provided a method for laying out a macrocell of an integrated circuit, comprising: selecting a region to be divided in a layout space; setting a plurality of candidate tangent lines in the region to be divided; and Selecting a target tangent from the plurality of candidate tangents based on the initial positions of the plurality of macro-units in the region to be segmented in the layout space and the positions of centers of gravity of each cluster obtained by clustering the plurality of macro-units The area to be divided is divided into two sub-areas, and each sub-area is used to place at least one cluster of macro units.

In the embodiment according to the present disclosure, the target tangent can be used to adaptively divide the layout area according to the actual placement requirements of the macro units in each cluster, so as to realize parallel placement of multiple modules.

In some embodiments, selecting the target tangent from the plurality of candidate tangents includes: for each candidate tangent, based on the initial positions of the plurality of macro units in the layout space and the The center of gravity position of each cluster obtained by unit clustering, calculating the classification loss used to reflect the mixing degree of the macrounits on both sides of the candidate tangent in the region to be segmented; and selecting the candidate tangent with the smallest classification loss as the target tangent. In such an embodiment, the target tangent can be accurately selected from multiple candidate tangents based on the classification loss.

In some embodiments, for each candidate tangent, calculating the classification loss used to reflect the mixing degree of the macro-units located on both sides of the candidate tangent in the region to be segmented includes: The clusters on both sides of the candidate tangent are assigned different classification labels; the macro units whose initial positions are located on both sides of the candidate tangent in the region to be divided are assigned different prediction labels; for each macro in the region to be divided Unit, based on whether the predicted label of the macrounit matches the classification label of the cluster to which the macrounit belongs and the distance between the macrounit and the candidate tangent, calculate the loss function related to the macrounit; and for each macrounit The loss functions are weighted and summed to determine the classification loss. In such an embodiment, the loss function related to the macro-unit can be determined based on whether the predicted label matches the classification label and the distance between the macro-unit and the candidate tangent, so as to further determine the classification loss.

In some embodiments, the loss function includes at least one of the following: an exponential loss function, a hinge loss function, a 0-1 loss function, or a Sigmoid loss function.

In some embodiments, the distance comprises a Euclidean distance.

In some embodiments, assigning different classification labels to the clusters whose center of gravity positions are located on both sides of the candidate tangent in the region to be segmented comprises: assigning the clusters whose center of gravity positions are located on the first side of the candidate tangent in the region to be segmented The clusters of the to-be-segmented region are assigned a first classification label, and the clusters whose barycenter position is located on the second side of the candidate tangent in the region to be segmented are assigned a second classification label, and the second classification label is different from the first classification label; Assigning different prediction labels to the macro-units in the region to be segmented whose initial positions are on both sides of the candidate tangent includes: assigning the macro-units in the region to be segmented whose initial positions are on the first side of the candidate tangent A prediction label, and assigning a second prediction label to the macro-unit whose initial position is on the second side of the candidate tangent in the region to be divided, the second prediction label is different from the first prediction label; wherein the The first predicted label matches the first class label, and the second predicted label matches the second class label. In such an embodiment, a method of assigning classification labels to clusters and prediction labels to macro-units is constructed. Using such classification labels and prediction labels, the target tangent can be accurately found in the subsequent process.

In some embodiments, the method further includes: after selecting the target tangent, adding perturbation to the target tangent to obtain one or more scrambled tangents near the target tangent; and the one or more scrambling tangents, arranging the plurality of macro-units; and selecting an actual tangent from the target tangent and the one or more scrambling tangents based on the arrangement result, the actual The tangent divides the area to be divided into two adjusted sub-areas. In such an embodiment, an actual tangent that is more in line with actual placement requirements can be selected from the target tangent and the corresponding scrambling tangent, so that the placement effect of the macro unit can be further optimized.

In some embodiments, the region to be divided is initially selected as the layout space. In such an embodiment, the entire layout space can be adaptively divided according to the actual placement requirements of the macro units in each cluster.

In some embodiments, the method further includes: selecting a sub-region containing two or more clusters in the two sub-regions as a new region to be segmented; setting a plurality of bars in the new region to be segmented a new candidate tangent; and based on the initial position of the macro-unit in the new region to be divided in the layout space and the position of the center of gravity of each cluster contained in the new region to be divided, from the plurality of new Select a new target tangent from the candidate tangents to divide the new region to be segmented into two new sub-regions. In such an embodiment, the layout space can be sequentially partitioned into multiple sub-regions using target tangents.

In some embodiments, the method further includes: after dividing a plurality of sub-regions in the layout space by using a plurality of target tangent lines, respectively placing the macrocells in each cluster in corresponding sub-regions.

In some embodiments, placing the macrocells in each cluster in the corresponding sub-areas includes: for each cluster, selecting macrocells with the same device type and a hierarchical similarity above a first threshold; for each cluster , form the selected macrocells into an array; for each cluster, place the array in the corresponding subregion; and for each cluster, place the macrocells other than the macrocells in the array in the corresponding in the subregion. In such an embodiment, by forming an array of macrocells of the same device type and similar in hierarchy, the layout regularity of the macrocells can be improved

In some embodiments, placing the macro-units in each cluster in corresponding sub-regions further includes: for each cluster, limiting the number of macro-units in the array to not exceed a second threshold. In such an embodiment, the size of the array can be reasonably adjusted according to design requirements.

In some embodiments, for each cluster, forming the selected macrocells into an array includes: for each cluster, calculating the total area of the selected macrocells; for each cluster, based on the total area and the corresponding The boundary length of the region determines a desired height of the array; and based on the desired height, a target number of rows for the array is determined. In such an embodiment, the shape of the array can be reasonably selected and placed according to the size of the layout area, so as to further improve the regularity of the layout of the macro-units.

In some embodiments, the method further includes: evaluating the layout result of the macro-unit based on the difference between the shape of the core area except the plurality of sub-areas in the layout space and a predetermined shape. In such an embodiment, evaluation can be easily performed on the layout results of the macro-cells, and a reasonable layout can be realized based on such evaluation results.

In some embodiments, the predetermined shape includes an oval or a rectangle.

In some embodiments, the method further includes: evaluating the layout result of the macrocell based on the boundary utilization of the layout space.

In some embodiments, the method further includes: for each sub-area, adjusting the orientation of the macro-units contained therein so that the pin positions of the macro-units close to the boundary of the sub-area are oriented in the layout space except for the The core area outside the multiple sub-areas, and the pin positions of the remaining macrocells are arranged back-to-back. By arranging the pins of the macrocell in this way, it is possible to reduce the congestion caused by layout.

In some embodiments, the method further includes: performing mixed-size layout on the macro cells and standard cells in the layout space at the same time, by optimizing the line length between the macro cells and the standard cells in the layout space and the The layout density of the macro-units in the layout space is determined, and the initial position of each macro-unit in the layout space is determined. In such an embodiment, the initial position of the macrocell in the layout space can be easily determined.

In some embodiments, the method further includes: based on the design hierarchy information of the macrocells in the layout space, the distance between the macrocells in the layout space and/or the distance between the cells in the layout space The interconnection number of , the macrocells in the layout space are clustered into multiple clusters. In such an embodiment, the macrocells can be accurately clustered by considering the design hierarchy information of the macrocells, the distance between the macrocells and the number of interconnections.

In some embodiments, the plurality of candidate tangents includes a plurality of first candidate tangents along a vertical direction and/or a plurality of second candidate tangents along a horizontal direction.

In some embodiments, the plurality of first candidate tangent lines are arranged at equal intervals.

In some embodiments, the plurality of second candidate tangent lines are arranged at equal intervals.

According to a second aspect of the present disclosure, there is provided an electronic device comprising: a processor; and a memory coupled to the processor and including instructions stored thereon, which when executed by the processor cause The electronic device executes the method according to the first aspect of the present disclosure.

According to a third aspect of the present disclosure, a computer-readable storage medium is provided, the computer-readable storage medium stores machine-executable instructions, when the machine-executable instructions are executed by at least one processor, such that The at least one processor implements the method according to the first aspect of the present disclosure.

According to a fourth aspect of the present disclosure there is provided a computer program product tangibly stored on a computer-readable storage medium and comprising machine-executable instructions which when executed by a device The device is caused to perform the method according to the first aspect of the present disclosure.

This Summary is provided to introduce a selection of concepts in a simplified form that are further described below in the Detailed Description. This Summary is not intended to identify key features or principal characteristics of the disclosure, nor is it intended to limit the scope of the disclosure.

Description of drawings

The above and other objects, features and advantages of embodiments of the present disclosure will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the drawings, several embodiments of the present disclosure are shown by way of example and not limitation.

FIG. 1 shows a flowchart of a method for laying out macrocells of an integrated circuit according to an embodiment of the disclosure.

Figure 2 shows the initial positions of some macrocells in a partial layout space.

Fig. 3 shows a flowchart for determining target tangents in layout space.

FIG. 4A shows multiple candidate tangent lines set for a region to be segmented.

FIG. 4B shows a target tangent determined for a region to be segmented.

Fig. 5 shows multiple candidate tangent lines set for another region to be segmented.

Fig. 6 shows a target tangent determined for another region to be segmented.

Fig. 7 shows a plurality of scrambling tangents arranged around the target tangent as shown in Fig. 4B.

FIG. 8 shows an array of macrocells arranged in a sub-area.

FIG. 9 shows a plurality of sub-regions divided in the layout space.

Figure 10 shows an ellipse constructed in layout space.

FIG. 11 shows the tree structure obtained when each region to be divided is divided by different target tangents.

FIG. 12 shows an exemplary arrangement of pins of three columns of macrocells.

Figure 13 shows a schematic block diagram of an example device that may be used to implement embodiments of the present disclosure.

In the respective drawings, the same or corresponding reference numerals denote the same or corresponding parts.

Detailed ways

Preferred embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although preferred embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

As used herein, the term "comprise" and its variants mean open inclusion, ie "including but not limited to". The term "or" means "and/or" unless otherwise stated. The term "based on" means "based at least in part on". The terms "one example embodiment" and "one embodiment" mean "at least one example embodiment." The term "another embodiment" means "at least one further embodiment". The terms "upper", "lower", "front", "rear" and other words indicating placement or positional relationship are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the principles of the present disclosure, rather than indicating or implying References to elements must have a particular orientation, be constructed, or operate in a particular orientation, and thus should not be construed as limiting the disclosure.

The devices in the chip are roughly divided into two types: macro cells and standard cells. The size of a macrocell is large compared to the size of a standard cell, for example, the size of a macrocell may be several thousand times larger than that of a standard cell. Because large-sized devices have a great influence on the overall layout of the chip, the macro cells are usually placed first, and then the standard cells are placed in the chip layout. After the macro cell and the standard cell are placed, the macro cell and the standard cell are wound to connect the macro cell and the standard cell.

Macrocell placement has practical constraints. For example, macro-units cannot overlap. In this case, using coordinates to represent the layout is not ideal, because it is difficult to define the rules that macro-units cannot overlap based on coordinates, which makes subsequent optimization inconvenient. Therefore, the existing macrocell layout schemes mainly focus on how to encode the layout, instead of using coordinates to encode the layout, but to use a special data structure to encode the layout. For example, one layout may be placed against the left side of the layout space, corresponding to one data structure, and another layout may be placed against the right side of the layout space, corresponding to a different data structure. According to these two data structures, the real layout can be obtained through the calculation process, that is, the layout can be optimized by means of encoding.

A conventional macrocell layout scheme is a tree-based coding method. The nodes of the tree are arranged in a certain order, and after some calculations, the tree can be translated into a specific layout. Therefore, what needs to be done is to optimize the structure of the tree to obtain the optimal tree, so that the optimal layout result can be obtained after translation. This solution can handle pre-positioning better. For example, in actual engineering design, it is required that some modules must be placed in a specific position (such as the lower left corner) and cannot be placed in other positions. This tree-based scheme can handle this situation well. However, this tree-based layout scheme also has some disadvantages, such as too large a search space and relatively low layout optimization efficiency. For example, for a tree with a search space size of 100 (100 candidates), twenty, thirty or more iterations may be required. The inventors found that the main reason for this problem is that this conventional layout scheme encodes the entire layout without considering the module concept. However, in the actual placement process, there is a very clear module concept. For example, a large module can be subdivided into several smaller modules. When placing them, the engineer first determines the position of the small modules, and then places the small modules to form a large module. Such an implementation process is difficult to achieve in conventional layout schemes.

There are also some macro-unit layout schemes that propose modular placement, and use design level information to help measure the closeness between macro-units. For example, macro-units under the same sub-module can be placed closer together. In this scheme, it can be judged which two macro-units can be placed relatively close by extracting design information. When dividing the layout area, this conventional layout method usually takes the center of gravity of the module as the center, and allocates a layout area for each module according to a given area and a fixed aspect ratio, so this is a deterministic layout method. If there is not enough space in the currently selected area, this layout method will expand the current area according to a certain ratio, and then search for a position where the macro unit can be placed in the new area. However, the layout area is only divided according to the area, and the aspect ratio is fixed, which cannot meet the needs of changing macrocell shapes. For example, many macro cells are long strips, in which case it is unreasonable to allocate the layout area according to the area.

There are also some macrocell layout schemes that do not use the tree structure for encoding, but use the contour line structure for encoding. After placing a macro unit each time, the layout space extends to the middle edge to a new corner. Subsequent macro units can continue to be placed according to the new corners, so as to achieve non-overlapping placement of macro units. However, such a layout scheme does not consider the hierarchical information of the macro-units, the search space is too large, and the layout optimization efficiency is low.

Embodiments of the present disclosure provide a method for laying out macrocells of an integrated circuit, which can cluster macrocells that need to be laid out into multiple clusters according to hierarchical information, and can Adaptively use the target tangent to divide the layout area according to the demand, so as to realize the parallel placement of multiple modules. Hereinafter, the principle of the present disclosure will be described in detail with reference to the accompanying drawings and exemplary embodiments.

FIG. 1 shows a flowchart of a method 100 for laying out macrocells of an integrated circuit according to an embodiment of the disclosure. As shown in FIG. 1 , the method 100 includes: at 101, determining the initial positions of multiple macro-units in the layout space; at 102, clustering the multiple macro-units into multiple clusters; at 103, based on the initial position of each macro unit and the center of gravity position of each cluster, multiple target tangents are determined in the layout space, and the multiple target tangents divide a plurality of sub-regions in the layout space, and each sub-region is used for placing the corresponding clusters; and at 104, placing the macro-units of each cluster in corresponding sub-regions.

At 101 , in order to determine the initial positions of multiple macro cells in the layout space, multiple macro cells and standard cells can be simultaneously placed with mixed sizes.

In some embodiments, the initial positions of the multiple macro cells in the layout space can be determined by optimizing the line length between the multiple macro cells and the standard cell, so as to avoid the disorder of the data flow. In November 2019 by Jai-Ming Lin, You-Lun Deng, Ya-Chu Yang, Jia-Jian Chen and Yao-Chieh Chen et al. in "IEEE/ACM International Conference on Computer-Aided Design, Digest of Technical Papers "A Novel Macro Placement Approach based on Simulated Evolution Algorithm" published on ICCAD discloses a method for determining the initial position of a macrocell by optimizing the line length between multiple macrocells and standard cells. The entire content of this document is incorporated herein by reference.

In some other embodiments, the initial positions of the multiple macro cells in the layout space may be determined by optimizing the line length between the multiple macro cells and the standard cell and the layout density of the multiple macro cells. In such an embodiment, the layout density of the macro-units is further considered, so that the macro-units can be separated as much as possible and the data flow disorder can be avoided.

In other embodiments, other methods may also be used to determine the initial positions of the multiple macro-units in the layout space, and the scope of the present disclosure is not limited in this regard.

The initial position of the macro unit determined at 101 may not be very precise, but it can reflect the information that needs to be physically considered during the layout, such as roughly which direction the macro unit needs to be placed in the layout space, such as on the left, right, and bottom wait. FIG. 2 shows the determined initial positions of some macrocells in the layout space. Fig. 2 only shows a part of the layout space, but not the whole layout space. As shown in FIG. 2 , rectangular blocks are used to represent macro units, some rectangular blocks are filled with vertical line patterns, some rectangular blocks are filled with horizontal line patterns, and some rectangular blocks are filled with oblique line patterns. Different fill patterns respectively indicate macrocells in different clusters generated at 102, the macrocells in the first cluster are represented by 201, the macrocells in the second cluster are represented by 202, and the macrocells in the third cluster are represented by 203 , which will be further elaborated below.

At 102, the plurality of macrocells are clustered into a plurality of clusters based on the design hierarchy information of the plurality of macrocells. The level information refers to the level of division of different modules, and the sub-modules belonging to the same module should be physically placed together. In this way, a shorter winding length can be achieved. For example, the current module A for realizing computing functions can be divided into several sub-modules B1, B2 and B3. Since module A is functionally cohesive, all sub-modules B1, B2, and B3 within this module are used to implement computing functions, so these sub-modules should also be physically placed together, which can reduce communication overhead.

In some embodiments, in order to cluster multiple macro-units into multiple clusters, a scoring function score(x, y) can be defined first, represented by the following formula (1):

score(x, y) = a·Distance(x, y)+b·Hierarch_Diff(x, y)+c·Connectivity(x, y) (1).

The scoring function score(x,y) is used to measure the closeness of the association between any two macrounits x and y, where Distance(x,y) represents the distance between the initially arranged macrounits x and y in the layout space Distance, Hierarchy_Diff(x,y) indicates the similarity between macrounit x and y at the design level, Connectivity(x,y) indicates the interconnection between macrounit x and y, a, b and c are weights respectively coefficient. The larger the Distance(x,y) is, the larger the distance between the two macro-units is, and the smaller the Distance(x,y) is, the smaller the distance between the two macro-units is. The larger the Hierarchy_Diff(x, y), the higher the similarity between the two macro-units at the design level, and the smaller the Hierarchy_Diff(x, y), the lower the similarity between the two macro-units at the design level. The larger the Connectivity (x, y), the more interconnections between two macro-units, and the smaller the Connectivity (x, y), the less interconnections between two macro-units. The scoring function score(x,y) can be obtained by weighting and summing Distance(x,y), Hierarchy_Diff(x,y) and Connectivity(x,y).

Based on the above scoring function score(x, y), the higher the score of the macrounit x and y, the higher the closeness of the two macrounits. Such macrounits need to be combined to form modules and placed together. By continuously merging similar macrocells from the bottom up, the macrocells in the layout space can be clustered into multiple clusters. In such an embodiment, the scoring function score(x, y) further considers the distance between macro-units and the number of interconnections on the basis of considering the design level information of multiple macro-units, so that it can more accurately evaluate Macrounits are clustered.

Three clusters of macrocells are shown in FIG. 2 . The macrocells in the first cluster are indicated by 201 , shown as rectangles filled with a vertical line pattern, comprising a total of five macrocells. The macrocells in the second cluster are indicated by 202 and are shown as rectangles filled with a horizontal line pattern, comprising a total of six macrocells. The macrocells in the third cluster are denoted by 203 and are shown as rectangles filled with an oblique line pattern, comprising a total of two macrocells.

In other embodiments, macrocells can also be clustered in other ways, such as only based on the design hierarchy information of macrocells, based on both the design hierarchy information of macrocells and the distance between macrocells, and based on the design Both the hierarchy information and the number of interconnections between macrocells, etc., the scope of the present disclosure is not limited in this respect.

For each cluster generated at 102, it needs to be allocated an actual layout area in the layout space. To this end, at 103, based on the initial position of each macrocell and the position of the center of gravity of each cluster, multiple target tangent lines are determined in the layout space, so as to divide the layout space into multiple sub-regions for placing each cluster.

Fig. 3 shows a flowchart for determining target tangents in layout space. In one embodiment, as shown in FIG. 3 , the determination of the target tangent in the layout space includes: at 1031, selecting a region to be divided in the layout space; at 1032, setting a plurality of candidate tangents in the region to be divided and at 1033, based on the initial positions of the plurality of macro-units in the region to be divided in the layout space and the positions of the centers of gravity of each cluster obtained by clustering the plurality of macro-units, from the plurality of A target tangent is selected from the candidate tangents to divide the area to be segmented into two sub-areas, and each sub-area is used to place at least one cluster of macro-units.

In some embodiments, selecting the target tangent from the plurality of candidate tangents includes: for each candidate tangent, based on the initial position of the macro-units in the region to be segmented in the layout space and the clustering of the macro-units in the region to be segmented to obtain The position of the center of gravity of each cluster of , calculate the classification loss used to reflect the mixing degree of the macrounits on both sides of the candidate tangent in the region to be segmented; and select the candidate tangent with the smallest classification loss as the target tangent. In such an embodiment, the target tangent can be accurately selected from multiple candidate tangents based on the classification loss. The higher the mixing degree of the macrounits on both sides of the candidate tangent, the worse the classification effect of the candidate tangent (for example, the macrounits in the same cluster are classified on both sides of the candidate tangent), and thus the greater the classification loss. The lower the mixing degree of the macrounits on both sides of the candidate tangent, the better the classification effect of the candidate tangent (for example, the macrounits in the same cluster are basically classified on the same side of the candidate tangent), so the classification loss is smaller . In such an embodiment, the target tangent can be accurately selected from multiple candidate tangents based on the classification loss.

FIG. 4A shows multiple candidate tangent lines 301 set for a region 300 to be segmented. As shown in FIG. 4A , the region to be divided 300 includes three clusters, and the macro-units of each cluster are respectively represented by rectangles with different filling patterns. The macrocells in the first cluster are indicated by 201 and are shown as rectangles filled with a vertical line pattern, comprising five macrocells in total. The macrocells in the second cluster are indicated by 202 and are shown as rectangles filled with a horizontal line pattern, comprising a total of six macrocells. The macrocells in the third cluster are indicated by 203 and are shown as rectangles filled with an oblique line pattern, comprising a total of two macrocells. A plurality of candidate tangent lines 301 are set in the region to be segmented 300 .

In some embodiments, as shown in FIG. 4A , the plurality of candidate tangent lines 301 includes a plurality of first candidate tangent lines extending along a vertical direction. In some embodiments, the plurality of candidate tangent lines 301 may further include a plurality of second candidate tangent lines extending along the horizontal direction. Using such candidate tangents 301 , the region to be divided 300 can be divided along the vertical and/or horizontal direction.

In some embodiments, multiple first candidate tangent lines may be arranged at equal intervals. Similarly, multiple second candidate tangents can also be set at equal intervals. In some other embodiments, the multiple first candidate tangent lines may not be arranged at equal intervals. Similarly, the multiple second candidate tangents may not be equally spaced. The scope of the present disclosure is not limited in this regard.

In some embodiments, the candidate tangent 301 may not extend along the vertical or horizontal direction, but extend along an oblique direction, and such a tangent can also divide the region 300 to be divided. In addition, it should be understood that in other embodiments, instead of a straight line, the candidate tangent 301 may also be a curved line or a broken line, and such a tangent can also divide the region 300 to be divided.

As shown in FIG. 4A , 10 candidate tangent lines 301 are set in the region to be segmented 300 . It should be understood that the number of candidate tangent lines 301 may be more or less, which may be set according to specific design requirements, and the scope of the present disclosure is not limited in this regard.

In some embodiments, for each candidate tangent 301 , calculating the classification loss used to reflect the mixing degree of the macrounits on both sides of the candidate tangent 301 in the region to be segmented 300 includes: The clusters on both sides of the candidate tangent 301 are given different classification labels; the macro units whose initial positions in the region 300 to be divided are located on both sides of the candidate tangent 301 are given different prediction labels; for each macro in the region 300 to be divided unit, based on whether the predicted label of the macrounit matches the classification label of the cluster to which the macrounit belongs and the distance of the macrounit from the candidate tangent line 301, calculates a loss function related to the macrounit; and related to each macrounit The weighted sum of the loss functions is used to determine the classification loss. In such an embodiment, the loss function related to the macro-unit can be determined based on whether the predicted label matches the classification label and the distance between the macro-unit and the candidate tangent 301 , so as to further determine the classification loss.

In one embodiment, for any candidate tangent 301, assign a first classification label (for example 0) to the cluster whose barycenter position is located on the first side (for example, the left side) of the candidate tangent 301 in the region 300 to be segmented, and assign The center of gravity position in the area to be divided 300 is located on the second side (for example, the right side) of the candidate tangent 301. The macro-units on the first side of the tangent line 301 are assigned a first prediction label (for example, 0), and the macro-units in the area to be divided 300 whose initial positions are on the second side of the candidate tangent line 301 are assigned a second prediction label (for example, 1). The first category label is different from the second category label, the first predicted label is different from the second predicted label, the first predicted label matches the first category label, and the second predicted label matches the second category label.

In another embodiment, for any candidate tangent 301, assign a first classification label (for example, 1) to the cluster whose barycenter position is located on the first side (for example, the left side) of the candidate tangent 301 in the region to be segmented 300, Assign a second classification label (such as 0) to the cluster whose center of gravity position is located on the second side (for example, the right side) of the candidate tangent line 301 in the region to be divided 300, and assign a second classification label (for example, 0) to the cluster whose initial position in the region to be divided 300 is located at the candidate tangent line The macro-units on the first side of 301 are assigned a first prediction label (for example, 1), and the macro-units whose initial positions are located on the second side of the candidate tangent line 301 in the area to be divided 300 are assigned a second prediction label (for example, 0) . The first category label is different from the second category label, the first predicted label is different from the second predicted label, the first predicted label matches the first category label, and the second predicted label matches the second category label.

It should be understood that, in other embodiments, for any candidate tangent line 301, other types of first classification labels and second classification labels may be assigned to the clusters in the region to be segmented 300, and the macrounits in the region to be segmented 300 may be assigned Other types of first predicted label and second predicted label, as long as the first class label is different from the second class label, the first predictive label is different from the second predictive label, the first predictive label matches the first class label, and the second It is sufficient that the predicted label matches the second classification label.

Using the first classification label, the second classification label, the first prediction label and the second prediction label, a loss function related to each macro-unit in the region to be segmented 300 can be determined. To this end, an exponential loss function l(z) expressed by Equation (2) is defined:

l(z)=exp(-fz) (2),

Among them, f controls the steepness of the loss function, and z indicates the quality of the sample classification result (that is, the result of classifying the macrounit). When calculating the loss specifically, for a candidate tangent line l, for a correctly classified macrounit (that is, a macrounit whose prediction label matches the classification label), make l(z) smaller, that is, z is a positive number; for a wrongly classified macrounit Units (i.e., macrounits whose predicted label does not match the class label), make l(z) large, ie z is a negative number.

In one embodiment, the distance from each macrounit x to the candidate tangent l is calculated as an index d(x,l) to measure the classification result of each macrounit, expressed by the following formula (3).

where l ₂ (x,l) represents the Euclidean distance from the center of the macrounit x to the candidate tangent l.

For each macro unit in the area to be divided 300, the loss function l(x)=exp(-f*d(x, l)) can be calculated by combining formula (2) and formula (3).

Subsequently, for the candidate tangent l, the following formula (4) is used to weight and sum the loss functions related to each macro-unit in the region to be segmented 300 to obtain the total classification loss for the candidate tangent l.

Where Wx represents a weighting coefficient for each macrounit x. For clusters with a large number of macro-units, the weighting coefficient Wx can be set smaller, and for clusters with a small number of macro-units, the weighting coefficient Wx can be set larger to solve the problem of unbalanced classification of macro-units.

Using the above method, the classification loss can be calculated for each candidate tangent 301 shown in FIG. 4A . On this basis, the candidate tangent 301 with the smallest classification loss can be selected as the target tangent. FIG. 4B shows the target tangent 302 determined for the region to be segmented 300 . As shown in FIG. 4B , the selected target tangent 302 is one of the candidate tangents 301 shown in FIG. 4A .

As shown in FIG. 4B , the target tangent line 302 divides the area to be divided into two sub-areas, that is, the first area 303 on the left and the second area 304 on the right. The first area 303 contains five macrocells 201 in the first cluster, and the second area 304 contains six macrocells 202 in the second cluster and two macrocells 203 in the third cluster. Since the second region 304 includes two clusters, the second region 304 can be used as a new region to be divided, and the process of determining the target tangent line as described above can be repeated. In such an embodiment, the layout space can be sequentially partitioned into multiple sub-regions using target tangents.

FIG. 5 shows a plurality of candidate tangent lines 305 set for the second region 304 . Similar to the candidate tangent 301 in the region to be segmented 300 as described above, for each candidate tangent 305, assign different classification labels to the clusters whose center of gravity positions in the second region 304 are located on both sides of the candidate tangent 305; In the second area 304, the macro-units whose initial positions are located on both sides of the candidate tangent line 305 are given different prediction labels; for each macro-unit in the second area 304, based on the prediction label of the macro-unit and the cluster to which the macro-unit belongs Whether the classification label matches and the distance of the macro-unit from the candidate tangent 305, use formula (2) and formula (3) to calculate the loss function related to the macro-unit; and use formula (4) to the relevant macro-unit The loss function performs a weighted sum to determine the classification loss. In this way, the classification loss can be calculated separately for each candidate tangent 305 shown in FIG. 5 . On this basis, the candidate tangent 305 with the smallest classification loss can be selected as the target tangent. FIG. 6 shows a target tangent 306 determined for the second region 304 . As shown in FIG. 6 , the selected target tangent 306 is one of the candidate tangents 305 shown in FIG. 5 .

By selecting the region to be divided in the layout space and determining the target tangent in each region to be divided in the above manner, multiple sub-regions can be divided in the layout space for arranging corresponding clusters. For the first target tangent among the plurality of target tangents, the region to be divided may be selected as the entire layout space. For any target tangent except the first target tangent among the plurality of target tangents, the region to be segmented is selected as the result obtained by cutting the corresponding region to be segmented by the previous target tangent, including two or more Subregions of multiple clusters. In this way, the layout space can be sequentially divided into a plurality of sub-regions using the target tangent.

In the embodiment according to the present disclosure, the macro-units that need layout can be clustered into multiple clusters according to the hierarchical information, and the layout area can be adaptively divided by the target tangent according to the actual placement requirements of the macro-units in each cluster, so that Realize parallel placement of multiple modules.

In the above embodiments, the division of the layout space is described by taking the exponential loss function as an example of the loss function. It should be understood that in other embodiments, other loss functions may also be used to calculate the classification loss, such as a hinge loss function, a 0-1 loss function, or a Sigmoid loss function. Using these loss functions, it is also possible to determine the target tangents in the region to be segmented.

In the above embodiments, the Euclidean distance from the center of the macrounit x to the candidate tangent l is used as the index d(x,l) to measure the classification result of each macrounit. It should be understood that, in other embodiments, other distances than the Euclidean distance may also be used as the index d(x,l) to measure the classification result of each macro unit.

As mentioned above, FIG. 4B shows the target tangent 302 determined for the region to be segmented 300 . However, the target tangent 302 may not be the optimal tangent for the region 300 to be segmented. To this end, as shown in FIG. 7 , after the target tangent 302 is determined and before the next target tangent is determined, a disturbance can be added to the target tangent 302 to obtain one or more scrambled tangents 307 near the target tangent 302 . Subsequently, for the target tangent 302 and the scrambling tangent 307, a plurality of macro-units in the region to be divided 300 are placed respectively, and based on the placement result, the one with the best actual placement effect is selected from the target tangent 302 and the scrambling tangent 307. Actual tangents to split the area to be segmented into two adjusted sub-areas. The scrambling tangent 307 may be selected randomly, or may be selected at a predetermined distance from the target tangent 302 , and the scope of the present disclosure is not limited in this respect.

For each target tangent determined in the layout space, the scrambling process described above for the target tangent 302 may be performed to obtain the corresponding actual tangent, thereby dividing the layout space into multiple adjusted sub-regions. In this way, an actual tangent that is more in line with actual placement requirements can be selected from the target tangent and the corresponding scrambling tangent, so that the placement effect of the macro-unit can be further optimized.

As shown in FIG. 1 , after the corresponding sub-regions for arranging each cluster are determined in the layout space, at 104 , the macro-units of each cluster are respectively placed in the corresponding sub-regions. In order to have better regularity in the layout area after placement, macro cells of the same type and with a high degree of similarity in design levels can be formed into an array for placement. Since all macrocells have unique numbers, the similarity between two macrocells can be determined based on the macrocell numbers. To this end, another scoring function score2(x, y) represented by the following formula (5) can be defined to determine the hierarchical similarity between two macro-units x and y.

score2(x, y) = hierarchy_Diff(x, y) (5),

Hierarchy_Diff(x,y) indicates the similarity between macrounit x and y at the design level. The larger the Hierarchy_Diff(x,y), the higher the similarity between the two macrounits at the design level. Hierarchy_Diff(x,y) The smaller the , the lower the similarity between the two macrocells at the design level.

For example, suppose macrounit x is numbered /A/B/C and macrounit y is numbered /A/D/E, then the hierarchical similarity between these two macrounits is 1, because macrounits x and y Share /A this level. Similarly, assuming that the number of macrounit x is /A/B/C and the number of macrounit y is /A/B/E, the hierarchical similarity between two macrounits is 2, because macrounits x and y Share the two levels of /A/B.

For each cluster, macrocells with the same device type and a hierarchical similarity above a first threshold may be selected to form an array. In some embodiments, the first threshold may be set to 1. In this case, macrocells with the same device type and a hierarchical similarity of more than 1 are selected to form an array. In some embodiments, the first threshold may be set to two. In this case, select macrocells with the same device type and a hierarchical similarity of 2 or more to form an array. In other embodiments, the first threshold can be set higher, which can be determined according to design requirements. In addition, the first threshold may be set to be the same or different for each cluster, and the scope of the present disclosure is not limited in this regard.

When actually forming an array, a constraint on the size of the array may be added, for example, a second threshold is set so that the size of the array (that is, the number of macro-units in the array) cannot exceed the second threshold. In this way, the size of the array can be reasonably adjusted according to design requirements.

For each sub-region, after determining the macro-units that need to form the array, it is also necessary to determine the shape of the array. For example, for an array containing 8 macro-units, it can be arranged in a 4x2 array or a 2x4 array.

In one embodiment, the shape of the array can be determined according to the area of the macro-units that need to form the array and the boundary lengths of the corresponding sub-regions where these macro-units will be placed. To this end, the ideal height ideaHeight of the array is first defined, expressed by the following formula (6):

Among them, total_macro_area indicates the total area of macro cells that need to form an array in a sub-area, boundary_length indicates the boundary length of the sub-area where the macro cells will be placed, and scale is a coefficient greater than or equal to 1 used to control the height of the array. The specific value of the coefficient scale can be set according to design requirements.

Using formula (6), the ideal height ideaHeight of the array can be calculated. Subsequently, according to the ideal height ideaHeight, the optimal number of rows best_row can be determined according to the following formula (7).

best_row=argmin _m abs(ideaHeight-h(m)) (7),

Among them, h(m) indicates the height of the array when the number of rows is m; abs is the absolute value function, and argmin _m is the value of m when abs(ideaHeight-h(m)) is the minimum value.

According to formula (7), the number of rows m of the array when the height h(m) of the array is closest to the ideal height ideaHeight can be taken as the optimal number of rows best_row.

Then, for each cluster, the selected macrocells are arrayed and placed in the corresponding subregion, and the macrocells other than the macrocells in the array are also placed in the corresponding subregion. FIG. 8 shows a macrocell array 309 arranged in a sub-area 308, the array includes three rows of macrocells, and each row has six macrocells arranged. By forming an array of macrocells of the same device type and similar layers, the layout regularity of the macrocells can be improved.

In some embodiments, for each sub-area, after determining the macro units that need to form the array, the shape of the array can also be determined to be close to a square or other shape, so that the macro units that need to form the array can be placed therein. In other embodiments, the shape of the array can also be determined in other ways, and the scope of the present disclosure is not limited in this respect.

After forming a plurality of subregions in the layout space, the layout results of the subregions can be evaluated. Since there are few connections between macrocells and the location of the standard cells is unknown at this time, it is not possible to accurately calculate the line length. According to expert experience, the higher the regularity of the remaining core area after placing the macro-units, the better. FIG. 9 shows a plurality of sub-regions 308 divided in the layout space 400 . In the layout space 400 , the area except the sub-area 308 is the core area 310 , and the core area 310 is used for arranging standard units.

In some embodiments, the method 100 further includes: evaluating the layout result of the macro-unit based on the difference between the shape of the core region 310 except the plurality of sub-regions 308 in the layout space 400 and a predetermined shape. For example, assuming that the area of the core area 310 is areaA, the center of the core area 310 is (xA, yA), and the ratio of the height to width of the core area is γ, then an ellipse B represented by the following formula (8) can be constructed:

in,

The area of the ellipse B is made equal to the area areaA of the core area 310 .

In this way, an ellipse B having an area equal to that of the core region 310 is constructed, as shown in FIG. 10 . Subsequently, the difference loss2 between the core region 310 and the ellipse B is calculated according to the following formula (9).

loss2=∫l(A(θ), B(θ))dθ (9),

Wherein A(θ) and B(θ) represent respectively when the circumference angle is θ, the distance from the point on the core area 310 and the ellipse B to the center (xA, yA); l(A(θ), B(θ))= max(0, B(θ)-A(θ)), that is, only consider the case of A(θ)<B(θ). Graphically, the value of the difference loss2 between the core area 310 and the ellipse B corresponds to the shaded area in FIG. 10 . The smaller the value of loss2 is, the closer the core area 310 is to an ellipse, and the more regular the core area 310 is, the more favorable it is for the overall winding. In such an embodiment, evaluation can be easily performed on the layout results of the macro-cells, and a reasonable layout can be realized based on such evaluation results.

In some embodiments, it may be desirable for the shape of the core region 310 to approximate a rectangular shape. In this case, the layout result of the macrocell may be evaluated based on the difference between the shape of the core area 310 and the rectangular shape. In some embodiments, the layout results of the macrocells can also be evaluated based on the boundary utilization of the layout space. The higher the boundary utilization, the higher the regularity. In other embodiments, the layout results of the macrocells may also be evaluated in other ways, and the present disclosure is not limited in this respect.

In some cases, for each region to be divided, there may be more than one target tangent, and different target tangents correspond to different divisions. By continuing to make such divisions, a tree structure can be obtained, as shown in Figure 11. Node A represents the area to be segmented, nodes B and B' indicate that the area to be segmented is divided by using different horizontal target tangents, and nodes C and C' indicate that the area to be segmented is further divided by using different vertical target tangents. Nodes B and B' and nodes C and C' are legal target tangents. The layout results can be evaluated in the manner described above. In some embodiments, the result search may be performed using depth-first traversal (DFS). In other embodiments, other tree search methods, such as breadth-first traversal (BFS), or heuristic search methods, such as Monte Carlo Tree Search (MCTS), may be used for result search.

In some embodiments, the method 100 further includes: for each subregion, adjusting the orientation of the macrocells contained therein, so that the pin positions of the macrocells close to the boundary of the subregion are oriented to the layout space except for a plurality of subregions The core area of the macrocell, and the pin positions of the remaining macrocells are arranged back-to-back. FIG. 12 shows an exemplary arrangement of pins of three columns of macrocells. As shown in FIG. 12 , the pins 511 of the macrocells 51 in the first row near the border of the subregions are facing the core region except for a plurality of subregions in the layout space, and the pins 521 of the macrocells 52 in the second row and the third row The pins 531 of the macrocell 53 are arranged back to back. By arranging the pins of the macrocell in this way, it is possible to reduce the congestion caused by layout.

FIG. 13 shows a schematic block diagram of a device 1300 that can be used to implement embodiments of the present disclosure. As shown in FIG. 13 , device 1300 includes a central processing unit (CPU) 1301 that can be programmed according to computer program instructions stored in read only memory (ROM) 1302 or loaded from storage unit 1308 into random access memory (RAM) 1303. computer program instructions to perform various appropriate actions and processes. In the RAM 1303, various programs and data necessary for the operation of the device 1300 can also be stored. The CPU 1301, ROM 1302, and RAM 1303 are connected to each other through a bus 1304. An input/output (I/O) interface 1305 is also connected to the bus 1304 .

Multiple components in the device 1300 are connected to the I/O interface 1305, including: an input unit 1306, such as a keyboard, a mouse, etc.; an output unit 1307, such as various types of displays, speakers, etc.; a storage unit 1308, such as a magnetic disk, an optical disk, etc. ; and a communication unit 1309, such as a network card, a modem, a wireless communication transceiver, and the like. The communication unit 1309 allows the device 1300 to exchange information/data with other devices through a computer network such as the Internet and/or various telecommunication networks.

Various procedures and processes described above, such as the method 100 , can be executed by the processing unit 1301 . For example, in some embodiments, method 100 may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 1308 . In some embodiments, part or all of the computer program may be loaded and/or installed on the device 1300 via the ROM 1302 and/or the communication unit 1309. When a computer program is loaded into RAM 1303 and executed by CPU 1301, one or more steps of method 100 described above may be performed.

The present disclosure may be a method, apparatus, system and/or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions thereon for carrying out various aspects of the present disclosure.

A computer readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device. A computer readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory), static random access memory (SRAM), compact disc read only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanically encoded device, such as a printer with instructions stored thereon A hole card or a raised structure in a groove, and any suitable combination of the above. As used herein, computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., pulses of light through fiber optic cables), or transmitted electrical signals.

Computer readable program instructions described herein may be downloaded from a computer readable storage medium to a respective computing/processing device, or downloaded to an external computer or external storage device over a network, such as the Internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers. A network adapter card or a network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .

Computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or Source or object code written in any combination, including object-oriented programming languages—such as Smalltalk, C++, etc., and conventional procedural programming languages—such as the “C” language or similar programming languages. Computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement. In cases involving a remote computer, the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as via the Internet using an Internet service provider). connect). In some embodiments, an electronic circuit, such as a programmable logic circuit, field programmable gate array (FPGA), or programmable logic array (PLA), can be customized by utilizing state information of computer-readable program instructions, which can Various aspects of the present disclosure are implemented by executing computer readable program instructions.

Aspects of the present disclosure are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the disclosure. It should be understood that each block of the flowcharts and/or block diagrams, and combinations of blocks in the flowcharts and/or block diagrams, can be implemented by computer-readable program instructions.

These computer readable program instructions may be provided to a processing unit of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine such that when executed by the processing unit of the computer or other programmable data processing apparatus , producing an apparatus for realizing the functions/actions specified in one or more blocks in the flowchart and/or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause computers, programmable data processing devices and/or other devices to work in a specific way, so that the computer-readable medium storing instructions includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks in flowcharts and/or block diagrams.

It is also possible to load computer-readable program instructions into a computer, other programmable data processing device, or other equipment, so that a series of operational steps are performed on the computer, other programmable data processing device, or other equipment to produce a computer-implemented process , so that instructions executed on computers, other programmable data processing devices, or other devices implement the functions/actions specified in one or more blocks in the flowcharts and/or block diagrams.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, a portion of a program segment, or an instruction that includes one or more Executable instructions. In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions.

Having described various embodiments of the present disclosure above, the foregoing description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and alterations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles of the various embodiments, practical applications or technical improvements over technologies in the market, or to enable other persons of ordinary skill in the art to understand the various embodiments disclosed herein.

Claims

A method for laying out macrocells of an integrated circuit, comprising:

Select the area to be divided in the layout space;

setting a plurality of candidate tangent lines in the region to be segmented; and

Selecting a target tangent from the plurality of candidate tangents based on the initial positions of the plurality of macro-units in the region to be segmented in the layout space and the positions of centers of gravity of each cluster obtained by clustering the plurality of macro-units The area to be divided is divided into two sub-areas, and each sub-area is used to place at least one cluster of macro units.
The method according to claim 1, wherein selecting the target tangent from the plurality of candidate tangents comprises:

For each candidate tangent, based on the initial positions of the plurality of macro-units in the layout space and the positions of the centers of gravity of each cluster obtained by clustering the plurality of macro-units, calculate the A classification loss for the degree of mixture of macrounits lying on both sides of the candidate tangent; and

The candidate tangent with the smallest classification loss is selected as the target tangent.
The method according to claim 2, wherein, for each candidate tangent, calculating the classification loss used to reflect the mixing degree of the macro-units located on both sides of the candidate tangent in the region to be segmented comprises:

Assigning different classification labels to the clusters whose barycenter positions are located on both sides of the candidate tangent line in the region to be segmented;

assigning different prediction labels to the macro-units whose initial positions are located on both sides of the candidate tangent in the region to be segmented;

For each macro-unit in the area to be segmented, based on whether the predicted label of the macro-unit matches the classification label of the cluster to which the macro-unit belongs and the distance between the macro-unit and the candidate tangent, calculate the loss function; and

A weighted sum of the loss functions associated with each macro-unit is performed to determine the classification loss.
The method according to claim 3, wherein the loss function includes at least one of the following: an exponential loss function, a hinge loss function, a 0-1 loss function, or a Sigmoid loss function.
The method of claim 3, wherein the distance comprises a Euclidean distance.
The method according to claim 3, wherein assigning different classification labels to the clusters whose center of gravity positions are located on both sides of the candidate tangent in the region to be segmented comprises: assigning the center of gravity position to the region to be segmented The clusters located on the first side of the candidate tangent are assigned a first classification label, and the clusters whose barycenter position is located on the second side of the candidate tangent in the region to be segmented are assigned a second classification label, and the second classification label is different from said first classification label;

Assigning different prediction labels to the macro-units in the region to be segmented whose initial positions are on both sides of the candidate tangent includes: assigning the macro-units in the region to be segmented whose initial positions are on the first side of the candidate tangent a prediction label, and assigning a second prediction label to a macrounit whose initial position is on the second side of the candidate tangent in the region to be divided, the second prediction label being different from the first prediction label;

wherein the first predicted label matches the first class label, and the second predicted label matches the second class label.
The method according to claim 1, further comprising:

After the target tangent is selected, adding a perturbation to the target tangent to obtain one or more scrambled tangents near the target tangent;

arranging the plurality of macrocells for the target tangent and the one or more scrambling tangents; and

An actual tangent is selected from the target tangent and the one or more scrambling tangents based on the arrangement result, and the actual tangent divides the area to be divided into two adjusted sub-areas.
The method according to claim 1, wherein the region to be divided is initially selected as the layout space.
The method according to claim 1, further comprising:

Selecting a sub-region containing two or more clusters in the two sub-regions as a new region to be divided;

setting a plurality of new candidate tangent lines in the new region to be segmented; and

Based on the initial position of the macro-unit in the new area to be divided in the layout space and the position of the center of gravity of each cluster contained in the new area to be divided, select a new tangent line from the plurality of new candidate tangents The target tangent to divide the new region to be segmented into two new sub-regions.
The method according to claim 9, further comprising:

After a plurality of sub-regions are divided in the layout space by using a plurality of target tangent lines, the macro-units in each cluster are respectively placed in corresponding sub-regions.
The method according to claim 10, wherein placing the macrocells in each cluster in corresponding sub-regions comprises:

For each cluster, selecting macrocells having the same device type and a hierarchical similarity above a first threshold;

For each cluster, forming an array of selected macrocells;

For each cluster, placing the array in a corresponding subregion; and

For each cluster, macrocells other than those in the array are placed in the corresponding sub-region.
The method according to claim 11, wherein placing the macrocells in each cluster in corresponding sub-regions further comprises:

For each cluster, the number of macrocells in the array is limited to not exceed a second threshold.
The method according to claim 11, wherein, for each cluster, forming an array of selected macrocells comprises:

For each cluster, calculate the total area of the selected macrocells;

for each cluster, determining a desired height of the array based on the total area and the boundary length of the corresponding sub-region; and

Based on the desired height, a target number of rows for the array is determined.
The method according to claim 10, further comprising:

The layout result of the macro-unit is evaluated based on the difference between the shape of the core area except the plurality of sub-areas in the layout space and a predetermined shape.
The method of claim 14, wherein the predetermined shape comprises an oval or a rectangle.
The method according to claim 10, further comprising:

The layout result of the macrocell is evaluated based on the boundary utilization rate of the layout space.
The method according to claim 10, further comprising:

For each sub-area, adjusting the orientation of the macro-units contained therein, so that the pin positions of the macro-units close to the boundary of the sub-area are oriented towards the core area in the layout space except for the multiple sub-areas, and the rest The pin locations of the macrocells are arranged back-to-back.
The method according to claim 1, further comprising:

Simultaneously perform mixed-size layout on the macrocells and standard cells in the layout space, and determine by optimizing the line length between the macrocells and standard cells in the layout space and the layout density of the macrocells in the layout space The initial position of each macrocell in the layout space.
The method according to claim 1, further comprising:

Based on the design hierarchy information of the macrocells in the layout space, the distance between the macrocells in the layout space and/or the number of interconnections between the cells in the layout space, the Macrounits cluster into multiple clusters.
The method according to claim 1, wherein the multiple candidate tangents include multiple first candidate tangents along the vertical direction and/or multiple second candidate tangents along the horizontal direction.
The method according to claim 20, wherein the plurality of first candidate tangent lines are arranged at equal intervals.
The method according to claim 20, wherein the plurality of second candidate tangent lines are arranged at equal intervals.
An electronic device comprising:

processor; and

A memory coupled to the processor and comprising instructions stored thereon which, when executed by the processor, cause the electronic device to perform the method of any one of claims 1-22.
A computer-readable storage medium, on which machine-executable instructions are stored, when the machine-executable instructions are executed by at least one processor, the at least one processor implements the The method of any one of 1-22.
A computer program product tangibly stored on a computer-readable storage medium and comprising machine-executable instructions which, when executed by a device, cause the device to perform the The method described in any one of 22.