Fixed contour layout planning method considering right-angle polygon soft module

Abstract

The invention discloses a fixed contour layout planning method considering right-angle polygon soft modules, which comprises the following steps of: (1) according to a given circuit netlist, initially setting each module into a point shape; (2) if the area of the soft module does not reach the minimum required area, gradually increasing the area of the soft module; (3) optimizing overlapping among the modules by adopting a global layout planning model based on an electrostatic field until the density overflow value is reduced to a preset target value; (4) dividing the layout area into a plurality of grids, wherein the soft module competes for residual grid resources; (5) calculating a negotiation sequence of the soft modules; (6) updating the corresponding grid congestion penalty cost; and (7) updating the historical cost of the grids after completing the grid occupation of all modules. And (8) executing soft module movement to optimize the line length. According to the method, the position and the shape of the right-angle polygon soft module can be determined under the conditions of fixed contour constraint and existence of the pre-placed hard module, the line length is effectively reduced, and the requirements of the current layout planning stage can be met.

Description

Technical Field

[0001] The present invention relates to the field of integrated circuit design automation, and in particular, to a fixed-outline layout planning method considering right-angled polygon soft modules. Background Art

[0002] With the rapid development of integrated circuit technology, the functions and complexities of chips are increasing day by day. To meet the growing performance requirements, layout planning in chip design has become a key link. Layout planning refers to the process of reasonably arranging each functional module in the two-dimensional area of a chip according to the functional requirements and physical constraints of the circuit during chip design, so as to achieve efficient circuit layout, wiring, and good electrical performance. Layout planning not only affects the area and power consumption of the chip, but also directly determines the performance, manufacturing cost, and scalability of the chip.

[0003] In the physical design of integrated circuits, the fixed-outline constraint means that the boundary of the design area is determined at the initial stage of design, and the chip layout needs to be completed within this fixed area. This constraint requires designers to efficiently arrange modules within the existing area and optimize space utilization. At the same time, with the increase in chip complexity, the pre-placement of hard modules (such as memories, input / output ports, etc.) has also become common. The positions of these modules are determined during the layout process, and designers need to arrange other deformable soft modules between these pre-placed modules to reduce the complexity and delay of wiring.

[0004] In traditional layout planning, common modules are usually rectangular modules. Such modules have simple shapes and can well adapt to design requirements in many scenarios. However, with the increase in design complexity, the flexibility and efficiency of rectangular modules in layout gradually become insufficient. In recent years, in order to optimize space utilization and wiring efficiency in chip design, soft modules (such as right-angled polygon modules) have gradually become important elements in layout planning. Soft modules have flexible shapes and can adjust their sizes and shapes according to needs, so as to more effectively adapt to the layout area, reduce waste of blank areas, and improve the space utilization of the chip.

[0005] Although traditional layout planning methods (such as heuristic algorithms, analytical methods, and learning-based methods) have been widely applied to module placement to a certain extent, most of these methods are designed for rectangular modules and are difficult to effectively handle modules with variable shapes such as right-angled polygon soft modules. Therefore, it is necessary to propose a new layout planning method to further optimize space utilization, reduce the wire length of wiring, and ensure the rationality and legality of the layout on the premise of meeting the fixed-outline constraint, pre-placement of hard modules, and module shape constraint. Summary of the Invention

[0006] To solve the above problems, the present invention discloses a fixed outline layout planning method considering right-angled polygonal soft modules. This method can effectively determine the position and shape of right-angled polygonal soft modules and effectively reduce line length while considering fixed outline constraints and the existence of pre-placed hard modules.

[0007] To achieve the above object, the technical solution adopted by the present invention is:

[0008] A fixed outline layout planning method considering right-angled polygonal soft modules comprises the following steps:

[0009] Step S1, according to the given circuit netlist, each module is initially set to a point shape. The initial position of all modules is determined by the secondary line length layout method;

[0010] Step S2, if the soft module area has not reached the minimum required area, gradually increase the soft module area;

[0011] Step S3, based on the currently set soft module area, using the global layout planning model based on the electrostatic field to optimize the overlap between modules until the density overflow value is reduced to a predetermined target value;

[0012] Step S4, repeat steps S2 and S3, and when all modules reach the minimum required area, legalize. Divide the layout area into several grids, and exclude the grids occupied by hard modules. Soft modules compete for the remaining grid resources;

[0013] Step S5, calculating the negotiation order of the soft modules.

[0014] Step S6, negotiate with each soft module in turn, select the lowest cost solution for occupying the grid, determine its polygonal shape and update the corresponding grid congestion penalty cost;

[0015] Step S7, after completing the grid occupation of all modules, update the historical cost of the grid. Repeat step S5 until all overlaps between modules are eliminated;

[0016] Step S8, repeating steps S5, S6 and S7 until all overlaps between modules are eliminated, and performing post-optimization. Execute soft module movement to optimize line length, and optimize soft module shape by envelope rectangle filling and module area redistribution until line length cannot be further reduced.

[0017] In step S1, the initial area of ​​all soft modules is set to a very small value, making them similar to the state of points. Then, the secondary line length layout algorithm is used to determine the initial position of the right-angled polygon soft module.

[0018] In step S2, the global layout planning process is divided into K gfphases, and the area of the soft module gradually increases in each phase. In the k gf -th phase, the area of module m i is:

[0019]

[0020] where Area min (m i ) is the minimum required area of module m i . Calculate the grid overlap rates at the left, bottom, right, and top sides of each current module respectively, and extend the four sides of the module proportionally to the required area of the current phase according to the current overlap situation, so that the sides with less overlap are extended by a greater distance.

[0021] In step S3, regarding each soft module as a rectangular soft module with a fixed current set value and a variable aspect ratio for global layout planning, this process can be transformed into a constrained minimization problem:

[0022]

[0023] where V = V f ∪V s represents the set of all modules, including the hard modules V f with fixed positions and shapes and the soft modules V s with variable positions and shapes. WL(V) is the total semi-perimeter line length between all modules, overflow is the density overflow value, which approximates the overlap between modules and is expressed as the ratio of the module overlap area to the total module area, and t of is the user-defined target overflow value.

[0024] Divide the layout area (W, H) into K×K equal-sized grids, and b ξ,η represents the grid in the ξ-th row and η-th column. Use a global layout planning model based on the electrostatic field to calculate the global position (x i ∈V s , y i ) and size (w i , i , ) of each soft module m i , and limit its aspect ratio ar within a given range. This model can be expressed as:

[0025]

[0026] where LSE(V) is the Log-Sum-Exp (LSE) line length between all modules, β is the penalty parameter, and the electrostatic potential energy N(m i ) of module mi ) It is calculated as:

[0027] N(m i ) = ∑ Area(b ξ，η ∩m i ) ψ(ξ, η) (4)

[0028] Where, Area(b ξ,η ∩m i ) is the intersection area of module m i and grid b ξ,η , and ψ(ξ, η) is the electric potential of grid b ξ,η :

[0029]

[0030] Where, are the discrete cosine transform coefficients. Subsequently, the multi-variable Nesterov method is used to optimize the position and size of the module until the overflow value overflow drops to the user-defined target value t of .

[0031] Then, steps S2 and S3 are repeated until all soft module areas reach the minimum required area and the overflow value overflow drops to the target value t of or below.

[0032] In step S4, the layout area is divided into several grids, and the grids occupied by hard modules are excluded according to the results of the global layout planning. The remaining grid resources are competed for by right-angled polygon soft modules. Let g μ,v represent the grid in the μ-th row and v-th column, and the side length of each grid is l g , then its congestion penalty cost p(g μ,v ) is

[0033] p(g μ，v ) = max(0, n(M μ,v ) - 1) (6)

[0034] Where, M μ,v is the set of soft modules currently occupying grid g μ,v , and n(M μ,v ) represents the number of soft modules currently occupying grid g μ,v . When the value of grid n(M μ,v ) is 0 or 1, it means that the grid is not occupied or only one soft module occupies it. Therefore, there is no resource conflict, and the congestion penalty cost of the grid is 0.

[0035] In addition, each grid also has its historical cost. At the beginning, the historical cost of each grid is initialized to 0.

[0036] In step S5, the negotiation order of the soft modules is calculated according to the minimum required area and the current congestion value of each soft module. The module order score is defined as follows,

[0037]

[0038] where G i is the set of grids occupied by module m i . Calculate the sum of the congestion penalty costs of the grids occupied by module m i as the congestion value of m i . The value of a f is between 0 and 1, representing the weight of the area term, and gradually decreases as the iteration progresses. The negotiation order is determined by the order score. Among the modules that have not been negotiated in the current iteration, the soft module with the highest score will be selected as the module for the next operation.

[0039] In step S6, the soft modules negotiate in sequence according to the order determined in S5. The selected module is used as the negotiation object. For each module m i , determining its polygon shape is divided into three steps:

[0040] (1) Overlapping grid removal: Remove the occupancy of the overlapping grids in the outer boundary of module m i ;

[0041] (2) Candidate edge search: Identify the candidate grids adjacent to the remaining grids, and merge the adjacent candidate grids with the same congestion penalty cost into continuous candidate edges in the horizontal and vertical directions respectively. A candidate grid may appear in both the horizontal candidate edge and the vertical candidate edge;

[0042] (3) Candidate edge selection and expansion: For all the candidate edges obtained in step (2), calculate the cost after module m i occupies each candidate edge one by one, and select the scheme with the lowest cost among them for occupancy. The required cost Cost(m i ,G i ) for module m i to occupy grid G i is calculated as follows:

[0043]

[0044] where o f is a user-defined parameter, and b(m i ) is the wire length cost of module m i , which is defined as:

[0045]

[0046] where N(mi ) represents the set of all nets connected to module m i , WL(net) represents the semi-perimeter wire length of a certain net, and w(net) is the weight of the corresponding net. If occupying a certain edge causes the shape of the soft module to violate the shape constraint, the cost is set to infinity. Repeat (2) and (3) until the module reaches the minimum required area. Subsequently, according to the new occupation situation of the module, update the congestion penalty cost of the relevant grid according to formula (6).

[0047] In the step S7, when all soft modules have completed one negotiation, one iteration ends, and the historical cost of the relevant grid is updated. Specifically: when the congestion penalty cost of a certain grid is not 0, it means that there are multiple soft modules competing for the grid resources, and its historical cost h(g μ,v ) will increase:

[0048] h(g μ,v ) = h(g μ,v ) + h f ·p(g μ,v ) (10)

[0049] where h f is a user-defined parameter. At the same time, as the number of iterations increases, o f in formula (8) gradually increases, thereby increasing the weight of the grid congestion term in the entire cost function. Then repeat steps S5, S6, and S7 until all overlaps between modules are eliminated.

[0050] In the step S8,

[0051] In the step S8, calculate the gradients of the wire length reduction of each soft module in the x and y directions respectively, and determine its moving speed in the corresponding direction; then, let the soft module move along the x or y axis direction; when the right-angled polygon soft module moves, its bounding rectangle will be displaced and occupy the available blank area therein; if the right-angled polygon soft module remains legal after moving, the shapes of other modules do not need to be modified; otherwise, the soft modules newly contacted on the moving path will be compressed, causing them to deform; only when both the moving module and the deformed module remain legal and the wire length decreases after this move, the move is accepted; otherwise, the move is rejected. When all modules have attempted to move, one iteration is completed; perform envelope rectangle filling and module area reallocation to make the module shapes more regular; repeat the above process until the wire length cannot be further reduced.

[0052] Advantages of the present invention:

[0053] 1. First, regard the soft module as a rectangle, and use the global layout planning model based on the electrostatic field to determine the global position of the soft module;

[0054] 2. Subsequently, a negotiation-based legalization strategy is adopted to determine the polygonal shape and position of the soft module while minimizing the wiring line length and ensuring that the soft module shape constraints are not violated.

[0055] 3. Finally, the polygonal shape of the soft module is dynamically adjusted during its movement to further optimize the shape and line length of the soft module. This method fully considers the influence of fixed contour constraints and pre-placed hard modules, can flexibly adjust the polygonal shape of the soft module, improve the layout space utilization, and significantly reduce the wiring line length, improving the overall layout quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 The present invention is a flow chart of a fixed-contour layout planning method considering polygonal soft modules. DETAILED DESCRIPTION

[0057] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0058] The present invention provides a fixed outline layout planning method for right-angle polygon soft modules. The method first determines the initial positions of all modules according to a given circuit netlist using a secondary line length layout method, then uses a global layout planning model based on an electrostatic field in combination with a module area growth strategy to determine the global position and size of the soft module, then uses a negotiation-based legalization method to eliminate overlaps between modules, and finally optimizes line length and module shape.

[0059] like Figure 1 As shown, the method comprises the following steps:

[0060] 101. According to the given circuit netlist, each module is initially set to a point shape. The initial positions of all modules are determined by the secondary line length layout method;

[0061] 102, if the area of ​​the soft module has not reached the minimum required area, gradually increase the area of ​​the soft module;

[0062] 103, based on the currently set soft module area, using the global layout planning model based on the electrostatic field to optimize the overlap between modules until the density overflow value is reduced to a predetermined target value;

[0063] 104, repeat steps S2 and S3, and when all modules reach the minimum required area, legalize. Divide the layout area into several grids, and exclude the grids occupied by hard modules. Soft modules compete for the remaining grid resources;

[0064] 105, calculate the negotiation order of the soft module.

[0065] 106, negotiate with each soft module in turn, select the lowest cost solution for occupying the grid, determine its polygonal shape and then update the corresponding grid congestion penalty cost;

[0066] 107, after completing the grid occupation of all modules, update the historical cost of the grid. Repeat step S5 until all overlaps between modules are eliminated;

[0067] 108, repeating steps S5, S6 and S7 until all overlaps between modules are eliminated, and performing post-optimization. Execute soft module movement to optimize line length, and optimize soft module shape by envelope rectangle filling and module area redistribution until line length cannot be further reduced.

[0068] In 101, the initial area of ​​all soft modules is set to a very small value to make them similar to the state of points. Then, the secondary line length layout algorithm is used to determine the initial position of the right-angled polygon soft modules. This initial position does not consider the overlap between modules and focuses on minimizing the line length between modules.

[0069] In 102, the global layout planning process is divided into K gf The area of ​​the soft module increases gradually in each stage. gf In the stage, module m i The area is:

[0070]

[0071] Among them, Area min (m i ) is module m i The minimum required area. Calculate the square overlap rate of the left, bottom, right, and top sides of each module respectively, and expand the four sides of the module to the required area of ​​the current stage in proportion to the current overlap situation, so that the sides with less overlap can expand a greater distance.

[0072] In 103, each soft module is regarded as a rectangular soft module with a fixed area of ​​the current setting value and a variable aspect ratio for global layout planning. The process can be transformed into a constrained minimization problem:

[0073]

[0074] Where V = V f ∪Vs Denotes the set of all modules, including the hard module V with fixed position and shape f and the soft module V with variable position and shape s , WL(V) is the total sum of the half-perimeter line lengths between all modules, overflow is the density overflow value, to approximate the overlap between modules, expressed as the ratio of the module overlap area to the total module area, t of is the user-defined target overflow value.

[0075] Divide the layout area (W, H) into K×K equal-sized squares, b ξ,η denotes the square in the ξ-th row and η-th column. Use the global layout planning model based on the electrostatic field to calculate the global position (x i ∈V s , y i ) and size i of each soft module m and limit its aspect ratio ar i within a given range , and this model can be expressed as:

[0076]

[0077] where, LSE(V) is the Log-Sum-Exp (LSE) line length between all modules, β is the penalty parameter, and the electrostatic potential energy N(m i ) of module m i is calculated as:

[0078] N(m i ) = ∑Area(b ξ，n ∩m i )ψ(ξ, η) (4)

[0079] where, Area(b ξ,η ∩m i ) is the intersection area of module m i and square b ξ,η , and ψ(ξ, η) is the electric potential of square b ξ,η :

[0080]

[0081] where, are the discrete cosine transform coefficients. Subsequently, use the multi-variable Nesterov method to optimize the position and size of the modules until the overflow value overflow drops to the user-defined target value t of , according to experience, usually t ofThe value is set to about 10%. Then, steps S2 and S3 are repeated until the area of all soft modules reaches the minimum required area and the overflow value overflow drops to the target value t of as follows.

[0082] In 104, the layout area is divided into a number of grids, and the grids occupied by hard modules are excluded according to the results of the global layout planning. The remaining grid resources are competed for by right-angled polygon soft modules. Let g μ,ν represent the grid in the μ-th row and ν-th column, and the side length of each grid is l g , then its congestion penalty cost p(g μ,ν ) is

[0083] p(g μ,v ) = max(0, n(M μ，v ) - 1) (6)

[0084] where M μ,ν is the set of soft modules currently occupying grid g μ,ν , and n(M μ,ν ) represents the number of soft modules currently occupying grid g μ,ν . When the value of grid n(M μ,ν ) is 0 or 1, it means that the grid is not occupied or only one soft module occupies it. Therefore, there is no resource conflict, and the congestion penalty cost of the grid is 0.

[0085] In addition, each grid also has its historical cost. At the beginning, the historical cost of each grid is initialized to 0.

[0086] In 105, the negotiation order of soft modules is calculated according to the minimum required area and the current congestion value of each soft module. The module order score is defined as follows,

[0087]

[0088] where G i is the set of grids occupied by module m i . Calculate the sum of the congestion penalty costs of the grids occupied by module m i as the congestion value of m i . The value of a f ranges from 0 to 1, representing the weight of the area term, and gradually decreases as the iteration progresses. The negotiation order is determined by the order score. Among the modules that have not been negotiated in the current iteration, the soft module with the highest score will be selected as the module for the next operation.

[0089] In 106, the soft modules negotiate in turn according to the order determined in S5. The selected module is used as the negotiation object. For each module m i , determining its polygon shape is divided into three steps:

[0090] (1) Overlapping grid removal: Remove module m i Occupation of overlapping grids in the outer boundary;

[0091] (2) Candidate edge search: Identify candidate grids adjacent to the remaining grids, and merge adjacent candidate grids with the same congestion penalty cost into continuous candidate edges in the horizontal and vertical directions respectively. A candidate grid may appear in both the horizontal candidate edge and the vertical candidate edge;

[0092] (3) Candidate edge selection and expansion: For all candidate edges obtained in step (2), calculate the cost of module m i occupying each candidate edge one by one, and select the option with the lowest cost for occupation. The cost required for module m i to occupy grid G i Cost(m i , G i ) is calculated as follows:

[0093]

[0094] where o f is a user-defined parameter, and b(m i ) is the wire length cost of module m i , which is defined as:

[0095]

[0096] where N(m i ) represents the set of all nets connected to module m i , WL(net) represents the semi-perimeter wire length of a certain net, and w(net) is the weight of the corresponding net. If occupying a certain edge causes the shape of the soft module to violate the shape constraint, the cost is set to infinity. Repeat (2) and (3) until the module reaches the minimum required area. Subsequently, according to the new occupation situation of the module, update the congestion penalty cost of the relevant grids according to formula (6).

[0097] In 107, when all soft modules have completed one negotiation, one round of iteration ends, and the historical cost of the relevant grids is updated. Specifically: when the congestion penalty cost of a certain grid is not 0, it means that there are multiple soft modules competing for the grid resources, and its historical cost h(g μ,ν ) will increase:

[0098] h(g μ,v ) = h(g μ，v ) + h f ·p(g μ,v ) (10)

[0099] Among them, h f is a user-defined parameter. At the same time, as the number of iterations increases, o in formula (8) f Gradually increase, thereby increasing the weight of the grid congestion term in the entire cost function. Then repeat steps S5, S6 and S7 until all overlaps between modules are eliminated.

[0100] In 108, the gradient of the line length reduction of each soft module in the x and y directions is calculated respectively, and its moving speed in the corresponding direction is determined. Then, the soft module is moved along the x or y axis. When the module moves, its enclosing rectangular frame will be displaced and occupy the available blank area therein. If the module remains legal after moving, the shapes of other modules do not need to be modified. Otherwise, the newly contacted soft modules on the moving path will be compressed, causing them to deform. The movement is accepted only when both the moving module and the deformed module remain legal and the line length is reduced after this movement; otherwise, the movement will be rejected. When all modules have tried to move, a round of iteration is completed. Envelope rectangle filling and module area redistribution are performed to make the module shape more regular. Repeat the above process until the line length cannot be further reduced and the optimization process terminates. At this point, the shape and position information of all right-angled polygonal soft modules are represented by the corner coordinates of their polygonal outlines and output in clockwise order as the final layout result.

[0101] The technical means disclosed in the scheme of the present invention are not limited to the technical means disclosed in the above-mentioned implementation mode, but also include technical schemes composed of any combination of the above technical features.

Claims

1. A fixed outline layout planning method considering right-angled polygonal soft modules, characterized by: The following steps are involved: Step S1, according to a given circuit netlist, each soft module is initially set to be a point; the initial position of all soft modules is determined by a secondary line length layout method; Step S2, each soft module is first treated as a rectangle, and if the area of ​​the soft module has not reached the minimum required area, the area of ​​the soft module is gradually increased; Step S3, based on the currently set soft module area, using the global layout planning model based on the electrostatic field to optimize the overlap between all modules until the density overflow value is reduced to a predetermined target value; Step S4, repeating steps S2 and S3, and when all soft modules reach the minimum required area, legalization is performed; the layout area is divided into a number of grids, and the grids occupied by hard modules are excluded; The soft modules compete for the remaining grid resources; Step S5, calculating the negotiation order of the soft modules; Step S6, negotiate with each soft module in turn, select the lowest cost solution for occupying the grid, determine its polygonal shape and update the corresponding grid congestion penalty cost; Step S7, after completing the grid occupation of all modules, update the historical cost of the grid; repeat step S5 until the overlap between all modules is eliminated; Step S8, repeating steps S5, S6 and S7 until all overlaps between modules are eliminated, and performing post-optimization; performing soft module movement to optimize line length, and optimizing soft module shape by envelope rectangle filling and module area redistribution, until the line length cannot be further reduced.

2. A method for fixed-contour layout planning considering right-angled polygonal soft modules according to claim 1, characterized in that: In step S2, the global layout planning process is divided into K gf stages, the area of ​​the soft module gradually increases in each stage; gf In the stage, module m i The area is: Among them, Area min (m i ) is module m i The minimum required area; calculate the square overlap rate at the left, bottom, right and top four sides of each module respectively, and expand the four sides of the module to the required area of ​​the current stage in proportion according to the current overlap situation, so that the sides with less overlap can expand a greater distance.

3. A method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 1, characterized in that: In step S3, each soft module is regarded as a rectangular soft module with a fixed area of ​​a current setting value and a variable aspect ratio for global layout planning, and the process is converted into a constrained minimization problem: Where V = V f ∪V s represents the set of all modules, including hard modules Vf with fixed positions and shapes and soft modules V with variable positions and shapes s , WL(V) is the sum of the semi-circumference lengths between all modules, overflow is the density overflow value, which is expressed as the ratio of the module overlap area to the total module area to approximate the overlap between modules, t of A user-defined target overflow value.

4. A method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 3, characterized in that: Divide the layout area (W, H) into K×K squares of equal size, b ξ,η represents the square of the ξth row and ηth column; a global layout planning model based on electrostatic field is used to calculate each soft module m i ∈V s The global position (x i ,y i )、Size(w i ,h i ), and its aspect ratio ar i Restricted to a given range The model can be expressed as: Where LSE(V) is the Log-Sum-Exp line length between all modules, β is the penalty parameter, and module m i Electrostatic potential energy N(m i ) is calculated as: N(m i )=∑Area(b ξ,η ∩m i )ψ(ξ,η) (4) Among them, Area(b ξ,η ∩m i ) is module m i and square b ξ,η The intersection area of ​​ψ(ξ,η) is the square b ξ,η The electric potential: in, is the discrete cosine transform coefficient; then, the multivariate Nesterov method is used to optimize the position and size of the module until the overflow value overflow drops to the user-defined target value t of Then, repeat steps S2 and S3 until the area of ​​all soft modules reaches the minimum required area and the overflow value overflow drops to the target value t of the following.

5. The method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 1, characterized in that: Step S4: μ,v Represents the grid of the μth row and the νth column, and the side length of each grid is l g , then its congestion penalty cost p(g μ,ν )for p(g μ,ν )=max(0,n(M μ,v )-1) (6) Among them, M μ,v is the currently occupied grid g μ,v The soft module set, n(M μ,v ) indicates the currently occupied grid g μ,ν The number of soft modules; when the grid n(M μ,ν ) is 0 or 1, indicating that the grid is not occupied or is occupied by only one soft module, so there is no resource conflict and the congestion penalty cost of the grid is 0; at the same time, each grid has its own historical cost. At the beginning, the historical cost of each grid is initialized to 0.

6. A method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 5, characterized in that: Step S5 calculates the negotiation order of the soft modules according to the minimum required area and current congestion value of each soft module; the module order score is defined as follows: Among them, G i For module m i The set of grids occupied by the computation module m i The sum of the grid congestion penalty costs is taken as m i The congestion value of f The value of ranges from 0 to 1, indicating the weight of the area term, and gradually decreases as the iteration proceeds; the negotiation order is determined by the order score, and the soft module with the highest score among the modules that have not been negotiated in the current iteration will be selected as the module for the next operation.

7. A method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 1, characterized in that: In step S6, the soft modules negotiate in sequence according to the order determined in step S5; the selected modules are used as negotiation objects, and for each module m i , determining its polygonal shape is divided into three steps: (1) Overlapping grid removal: Remove module m i Occupancy of overlapping grids in the outer boundary; (2) Candidate edge search: Identify candidate grids adjacent to the remaining grids and merge adjacent candidate grids with the same congestion penalty cost into continuous candidate edges in the horizontal and vertical directions. A candidate grid may appear in both horizontal and vertical candidate edges. (3) Candidate edge selection and expansion: For all candidate edges obtained in step (2), calculate the module m one by one i The cost after occupying the candidate edge, and choose the solution with the lowest cost to occupy it; Module m i Occupies grid G i Cost(m i ,G i ) is calculated as: Among them, f For user-defined parameters, b(m i ) is module m i The line length cost is defined as: Among them, N(m i ) indicates that the module m i The set of all connected nets, WL(net) represents the semi-circular length of a net, and w(net) is the weight of the corresponding net. If occupying an edge causes the shape of the soft module to violate the shape constraint, the cost is set to infinity. Repeat (2) and (3) until the module reaches the minimum required area. Then, according to the new occupation status of the module, the congestion penalty cost of the relevant grid is updated according to formula (6).

8. The method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 1, characterized in that: In step S7, when all soft modules complete a negotiation, a round of iteration ends and the historical cost of the relevant grid is updated; specifically, when the congestion penalty cost of a grid is not 0, it means that there are multiple soft modules competing for the grid resources, and their historical cost h(g μ,v ) will increase: h(g μ,ν )=h(g μ,ν )+h f ·p(g μ,ν ) (10) Among them, h f is a user-defined parameter; at the same time, as the number of iterations increases, o in formula (8) f Gradually increase, thereby increasing the weight of the grid congestion term in the entire cost function; then repeat steps S5, S6 and S7 until all overlaps between modules are eliminated.

9. The method for fixed outline layout planning considering right-angled polygonal soft modules according to claim 5, characterized in that: In step S8, the gradient of the line length reduction of each soft module in the x and y directions is calculated respectively to determine its moving speed in the corresponding direction; then, the soft module is moved along the x or y axis direction; when the right-angled polygon soft module moves, its enclosing rectangle will be displaced and occupy the available blank area therein; if the right-angled polygon soft module remains legal after the movement, there is no need to modify the shapes of other modules; otherwise, the new soft modules contacted on the moving path will be compressed, causing them to deform; the movement is accepted only when both the moving module and the deformed module remain legal and the line length is reduced after this movement; otherwise, the movement will be rejected; when all modules have tried to move, a round of iteration is completed; the envelope rectangle is filled and the module area is redistributed to make the module shape more regular; the above process is repeated until the line length cannot be further reduced.

Images