Layered layout wiring method and system based on quantum cellular automaton circuit
By employing a hierarchical placement and routing method and dynamic clock synchronization constraints, the problems of low placement and routing efficiency and clock synchronization failure in quantum cellular automata circuits are solved, achieving an efficient and reliable routing process that meets the requirements of low latency and high integration.
Patent Information
- Application Number
- CN202510808066.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2025-03-21
- Filing Date
- 2025-06-17
- Publication Date
- 2025-11-07
AI Technical Summary
Existing placement and routing algorithms in quantum cellular automata circuits suffer from a significant decrease in efficiency as the solution space increases with the number of logic gates. They also struggle to eliminate long lines and cross lines between layers, leading to clock synchronization failures. Furthermore, they cannot handle clock signal distribution for cross-layer connections. Static routing strategies cannot dynamically adapt to complex constraints, resulting in a low success rate.
A hierarchical layout and routing method based on quantum cellular automata circuits is proposed. By preprocessing the hierarchical logic netlist, long inter-layer lines that do not conform to global clock synchronization and optimized cross lines are eliminated. Combined with dynamic clock synchronization constraints and node reordering, dynamic routing and clock signal allocation are performed.
It significantly improves the quality and reliability of layout and routing, reduces clock synchronization risks, increases routing efficiency and success rate, optimizes resource usage, and ensures that the layout scheme meets the requirements of low latency, low power consumption and high integration.
Smart Images

Figure CN120911387A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of integrated circuits, and in particular to a layered layout and routing method and system based on a quantum cellular automata circuit. BACKGROUND
[0002] Traditional CMOS technology faces problems such as power consumption, integration and delay due to physical limits. Quantum cellular automata (QCA) as the next generation of nanometer devices has advantages such as high integration and low power consumption, but its layout and routing needs to meet strict global clock synchronization constraints. Existing layout and routing algorithms have the following problems when solving large-scale circuits: the solution space increases dramatically with the number of logic gates, and the efficiency decreases significantly; it is difficult to eliminate long lines and cross lines between layers, resulting in clock synchronization failure and inability to handle clock signal distribution for cross-layer connections; static routing strategies cannot dynamically adapt to complex constraints, resulting in low success rate. SUMMARY
[0003] To solve the technical problems in the background art, the present application proposes a layered layout and routing method and system based on a quantum cellular automata circuit.
[0004] The layered layout and routing method based on a quantum cellular automata circuit proposed by the present application comprises:
[0005] Obtaining an original logic netlist, the original logic netlist containing a first logic order between a plurality of nodes, first fan-in and fan-out information of each node, and a first level;
[0006] Pretreating the original logic netlist to obtain a first logic netlist layered according to logic depth;
[0007] Eliminating long lines between layers that do not conform to global clock synchronization and optimizing cross lines according to a preset layout strategy for the first logic netlist to obtain a second logic netlist, the second logic netlist containing a second logic order between a plurality of nodes, second fan-in and fan-out information of each node, and a second level;
[0008] Dynamically routing the plurality of nodes in the second logic netlist according to a preset routing strategy and assigning clock signals.
[0009] Preferably, the pretreatment specifically comprises:
[0010] Dividing the plurality of nodes in the original logic netlist into n layers according to the logic depth of the nodes to generate a level set {L0, L1, …, Ln}, and numbering the plurality of nodes so that one node corresponds to one number; n
[0011] Eliminating long lines between layers that do not conform to global clock synchronization and optimizing cross lines according to a preset layout strategy for the first logic netlist to obtain a second logic netlist, the second logic netlist containing a second logic order between a plurality of nodes, second fan-in and fan-out information of each node, and a second level; n If the number of fan-out nodes of any node in the set is greater than 3, a virtual node is added to the set, and part of the fan-out is distributed to the virtual node, so that the number of fan-out of each node is less than or equal to 3, and the set of node fan-out adjustment levels {L0', L1',..., L n ′} as the first logical netlist.
[0012] Preferably, the preset layout strategy specifically includes:
[0013] The Manhattan distance of the physical coordinates of the node pairs of adjacent layers in the first logical netlist is calculated, and the Manhattan distance related to each node is compared with the preset threshold value one by one;
[0014] When any Manhattan distance is greater than the preset threshold value, the logical line between the related nodes is regarded as an interlayer long line;
[0015] The node ordering in the same level layer is adjusted to eliminate the interlayer long line;
[0016] In the process of adjusting the node ordering, the number of cross lines is calculated synchronously, and the node ordering in each layer is optimized layer by layer by balancing the target through a weighted evaluation function, so as to generate a second logical netlist after optimization of the layout.
[0017] Preferably, the weighted evaluation function specifically includes:
[0018]
[0019] Wherein, N cross represents the number of cross lines between the i-th layer and the j-th layer; P ij represents the penalty value of layer sequence constraint, which ensures that the i-th layer appears before the j-th layer; and q is the evaluation function value.
[0020] Preferably, the weighted evaluation function specifically includes:
[0021] The nodes in the same level layer are ordered according to the evaluation function value q, and each level is traversed, so as to generate a second logical netlist after optimization of the layout, under the premise of ensuring the correct layer sequence, with the least number of cross lines and the optimal ordering when the evaluation function value q is the least.
[0022] Preferably, the preset layout strategy specifically includes:
[0023] The nodes included in the second logical netlist are arranged in levels according to the specific number of levels of the second level from small to large, so as to obtain a level arrangement set;
[0024] classifying logical connections between nodes in each second level in the hierarchical arrangement set and nodes in other second levels into short line sets and long line sets to obtain the short line sets and the long line sets;
[0025] dynamically adjusting an inter-level distance and a clock delay between the second levels in the hierarchical arrangement set based on the short line sets;
[0026] distributing clock signals based on the inter-level distance and the clock delay between the second levels in the hierarchical arrangement set.
[0027] Preferably, the classifying logical connections between nodes in each second level in the hierarchical arrangement set and nodes in other second levels into short line sets and long line sets to obtain the short line sets and the long line sets specifically comprises:
[0028] finding fan-in nodes of each node in each second level in the hierarchical arrangement set, and if a number of levels corresponding to any fan-in node and the second level of the node differ by 1, adding a logical connection between the node and the fan-in node to a short line set, otherwise adding a logical connection between the node and the fan-in node to a long line set;
[0029] traversing each second level in the hierarchical arrangement set according to the above process to obtain the short line sets and the long line sets corresponding to the hierarchical arrangement set.
[0030] Preferably, the dynamically adjusting an inter-level distance and a clock delay between the second levels in the hierarchical arrangement set based on the short line sets specifically comprises:
[0031] for any two adjacent second levels in the hierarchical arrangement set;
[0032] initializing the inter-level distance as 1, determining a Manhattan distance longest line and a Manhattan distance shortest line between the two levels, constantly increasing the inter-level distance, iteratively calculating a maximum clock delay of the Manhattan distance shortest line and a minimum clock delay of the Manhattan distance longest line, and taking an inter-level distance and a clock delay corresponding to a case where the maximum clock delay is equal to the minimum clock delay as the inter-level distance and the clock delay corresponding to the two adjacent second levels;
[0033] traversing all adjacent second levels in the hierarchical arrangement set according to the above process to obtain the inter-level distance and the clock delay corresponding to each second level and an adjacent second level.
[0034] Preferably, the distributing clock signals based on the inter-level distance and the clock delay between the second levels in the hierarchical arrangement set specifically comprises:
[0035] The horizontal coordinates are constructed according to the ordering of the nodes in each second level, and the vertical coordinates of the nodes in each second level are determined based on the layer spacing of each second level in the hierarchical arrangement set;
[0036] The horizontal coordinates, the vertical coordinates of each node and the corresponding clock delay configuration logic gate device are based on the corresponding clock signal.
[0037] The application provides a hierarchical layout and wiring system based on a quantum cellular automaton circuit, which comprises:
[0038] The data acquisition module is used for acquiring an original logic netlist, and the original logic netlist comprises a first logic ordering between a plurality of nodes, first fan-in and fan-out information of each node and a first level;
[0039] The data processing module is used for preprocessing the original logic netlist to obtain a first logic netlist layered according to a logic depth;
[0040] The data layout module is used for eliminating long interlayer lines that do not conform to global clock synchronization and optimizing cross lines according to a preset layout strategy, so as to obtain a second logic netlist, and the second logic netlist comprises a second logic ordering between a plurality of nodes, second fan-in and fan-out information of each node and a second level;
[0041] The data wiring module is used for dynamically wiring a plurality of nodes in the second logic netlist according to a preset wiring strategy and allocating a clock signal.
[0042] In the application, the method and system for hierarchical layout and wiring based on a quantum cellular automaton circuit are proposed, which significantly improves the layout and wiring quality and reliability of the quantum cellular automaton (QCA) circuit by using the Manhattan distance-based long interlayer line elimination strategy, combining dynamic clock synchronization constraints and node reordering optimization, reducing the clock synchronization risk, limiting the physical distance of the cross-layer connection and the clock region span, ensuring that the signal is transmitted within a single clock cycle, reducing the QCA cell excited state error caused by long lines, improving the wiring efficiency and success rate, avoiding path conflicts and deadlocks through dynamic obstacle pre-generation and improved A* algorithm, optimizing resource occupation, releasing more space for other logic units after eliminating long interlayer lines and reducing the number of cross lines, enhancing design robustness, and ensuring that the layout scheme meets the requirements of low delay, low power consumption and high integration by deeply integrating the clock region constraints and the cross line minimization strategy. BRIEF DESCRIPTION OF DRAWINGS
[0043] Figure 1 The application provides a working flowchart of a method for hierarchical layout and wiring based on a quantum cellular automaton circuit;
[0044] Figure 2This is a schematic diagram of a hierarchical layout comparison structure of one embodiment of the hierarchical layout and wiring method based on quantum cellular automata circuits proposed in this invention.
[0045] Figure 3 This is a schematic representation of the original logic net of an embodiment of a hierarchical layout and routing method based on quantum cellular automata circuits proposed in this invention.
[0046] Figure 4 This invention proposes a hierarchical layout and routing method based on quantum cellular automata circuits. Figure 3 The corresponding first logical network represents the intent;
[0047] Figure 5 This invention proposes a hierarchical layout and routing method based on quantum cellular automata circuits. Figure 3 A schematic diagram of the structure after layout and routing processing;
[0048] Figure 6 This is a schematic diagram of the system architecture of a hierarchical layout and wiring system based on quantum cellular automata circuits proposed in this invention. Detailed Implementation
[0049] Reference Figures 1-6 The present invention proposes a hierarchical layout and routing method based on quantum cellular automata circuits, comprising the following steps:
[0050] S1. Obtain the original logical netlist, which contains the first logical ordering between multiple nodes, the first fan-in and fan-out information of each node, and the first level.
[0051] S2. Preprocess the original logical netlist to obtain the first logical netlist layered according to logical depth.
[0052] In this embodiment, the preprocessing specifically includes:
[0053] Based on the logical depth of the nodes, the multiple nodes in the original logical netlist are divided into n layers to generate a hierarchical set {L0, L1, ..., L...}. n}, and number multiple nodes so that each node corresponds to a number;
[0054] For the hierarchical set {L0,L1,…,L…} n For any node in the set {L0′, L1′, ..., L1′}, if its fan-out number is greater than 3, then add a virtual node to it, distributing some of the fan-out to the virtual node, so that the fan-out number of each node is less than or equal to 3. Adjust the fan-out of the nodes to form the resulting hierarchical set {L0′, L1′, ..., L1′}. n ′} serves as the first logical netlist.
[0055] Specifically, first, the logical netlist G=(V, E) is layered according to the logical depth of the nodes, and let L={L0, L1, … L n} represent the hierarchy set, and the hierarchy of a node v∈V is defined as the numerical value of the layer where the node is located L x The hierarchy of the input node is set to 0. For any node v i ∈V, if the number of fan-out nodes is greater than 3, a virtual node is added, and part of the fan-out is distributed to the virtual node, reducing the number of node fan-out to 3 (QCA circuit characteristics). After preprocessing, the first logical netlist G'=(V', E') is obtained according to the logical depth layering, and the fan-out number of all nodes is less than or equal to 3. QCA circuit data transmission is achieved through the electrostatic interaction between adjacent cells to transfer information. If the fan-out number exceeds 3, the cell cannot provide enough energy to reliably drive the downstream cell to the correct and stable polarization state (logic 0 or 1).
[0056] S3, eliminate the interlayer long line that does not conform to the global clock synchronization and optimize the cross line according to the preset layout strategy to obtain a second logical netlist, the second logical netlist comprising a second logical order between a plurality of nodes, second fan-in and fan-out information of each node, and a second hierarchy.
[0057] In this embodiment, the short line refers to the connection between nodes between adjacent two layers, the length of which does not exceed the preset threshold; the interlayer long line refers to the connection between nodes between adjacent two layers, the length of which exceeds the preset threshold; and the cross-layer long line refers to the connection between nodes between non-adjacent two layers.
[0058] In this embodiment, the preset layout strategy specifically includes:
[0059] The Manhattan distance of the physical coordinates of the node pairs of the adjacent layers in the first logical netlist is calculated, and the Manhattan distance related to each node is compared with the preset threshold one by one;
[0060] When any Manhattan distance is greater than the preset threshold, the logical line between the related nodes is regarded as an interlayer long line;
[0061] The interlayer long line is eliminated by adjusting the node order in the same hierarchy layer;
[0062] In the process of adjusting the node order, the number of cross lines is calculated synchronously, and the second logical netlist after optimization of the layout is generated by balancing the target through a weighted evaluation function and optimizing the node order in each layer layer by layer.
[0063] Specifically, the Manhattan distance formula between two points is: D=|x1-x2|+|y1-y2|;
[0064] In this embodiment, the weighted evaluation function is specifically:
[0065]
[0066] wherein, N cross represents the number of cross lines between the i-th layer and the j-th layer level; P ij represents the penalty value of layer sequence constraint, ensuring that the i-th layer level appears before the j-th layer level; q is the evaluation function value.
[0067] In this embodiment, the node ordering in each layer is optimized layer by layer by balancing the target through the weighted evaluation function, to generate the second logical netlist after the optimized layout, specifically including:
[0068] With the goal of reducing the number of interlayer cross lines while ensuring the correct layer sequence, the nodes in the same layer level are sorted according to the evaluation function value q, and each layer level is traversed to generate the second logical netlist after the optimized layout, wherein the number of cross lines is the least and the sorting is the optimal when the evaluation function value q is the least.
[0069] Specifically, all nodes are traversed and numbered through the Depth_control() function, first-sorting is the initial interlayer sorting, and the maximum number of iterations is set. In each iteration process, firstly, it is judged whether there is an interlayer long line in the sorting by using the L_man(v,w) function; then the weighted median function weightmedian() is introduced to improve the sorting based on the k-layer cross minimization heuristic function; finally, the Transposing() function is called to check each adjacent node pair, and if the number of cross lines is reduced, the original sorting will be replaced by the new sorting. Figure 2 As shown in the figure, the long lines between the nodes are eliminated and the number of cross lines is reduced after the first four layers of nodes of the benchmark test circuit parity are subjected to the hierarchical layout algorithm.
[0070] S4, dynamically routing the plurality of nodes in the second logical netlist according to the preset routing strategy, and assigning a clock signal.
[0071] In this embodiment, the preset routing strategy specifically includes:
[0072] The plurality of nodes contained in the second logical netlist are arranged in levels according to the order from small to large of the specific number of levels of the second levels, to obtain a level arrangement set;
[0073] The logical connections between the nodes in each second level in the level arrangement set and the nodes in other second levels are classified into a short line set and a long line set, to obtain the short line set and the long line set;
[0074] Based on the short line set, the interlayer spacing between each second level in the level arrangement set and the clock delay are dynamically adjusted;
[0075] The layer distance between each second level in the hierarchical arrangement set and the clock delay are determined.
[0076] In the embodiment, the logical connections between the nodes in each second level in the hierarchical arrangement set and the nodes in other second levels are classified into the short line set and the long line set to obtain the short line set and the long line set, which specifically includes:
[0077] Fan-in nodes of each node in each second level in the hierarchical arrangement set are found, and if the number of levels corresponding to any fan-in node and the second level of the node differ by 1, the logical connection formed by the node and the fan-in node is added to the short line set, otherwise the logical connection formed by the node and the fan-in node is added to the long line set.
[0078] Each second level in the hierarchical arrangement set is traversed according to the above process to obtain the short line set and the long line set corresponding to the hierarchical arrangement set.
[0079] In the embodiment, the layer distance between each second level in the hierarchical arrangement set and the clock delay are determined based on the short line set, which specifically includes:
[0080] Any two adjacent second levels in the hierarchical arrangement set are selected.
[0081] The layer distance is initialized to 1, all lines between the two layers are traversed, the Manhattan distance of the longest line and the shortest line are determined, the layer distance is increased, the maximum clock delay of the shortest line and the minimum clock delay of the longest line are iteratively calculated, and the layer distance and the clock delay corresponding to the maximum clock delay equal to the minimum clock delay are taken as the layer distance and the clock delay corresponding to the current two adjacent second levels.
[0082] All adjacent second levels in the hierarchical arrangement set are traversed according to the above process to obtain the layer distance and the clock delay corresponding to each second level and the adjacent second level.
[0083] Specifically, the maximum clock delay is L-1, and the minimum clock delay is floor((L-1) / M), where floor() represents rounding down, L is the line length, and M is the maximum continuous clock length, which refers to the maximum number of continuous cells that can reliably transmit signals within a single clock region of a QCA circuit.
[0084] In the embodiment, the layer distance between each second level in the hierarchical arrangement set and the clock delay are determined based on the layer distance and the clock delay, which specifically includes:
[0085] The horizontal coordinates are constructed according to the ordering of the nodes in each second level, and the vertical coordinates of the nodes in each second level are determined based on the layer distance between each second level in the hierarchical arrangement set.
[0086] Based on the abscissa, ordinate and corresponding clock delay configuration logic gate device and the generation of the corresponding clock signal of each node.
[0087] Specifically, after the hierarchical layout, the second logic netlist G * =(V, E) and the ordering information S(v i )=(x i , y i ), to meet the global clock synchronization constraints, the path between nodes must contain explicit clock information. Therefore, the routing algorithm needs to solve the routing problem between logic gates and design the clock-synchronized information flow topology under the requirements of delay and area.
[0088] The preprocessed file and the layout file are read by the read_graph() function, and the fan-in, fan-out, level and serial number of each node are obtained by the former and stored in the node structure; the coordinates of each node are obtained by the latter and also stored in the node structure.
[0089] In this embodiment, the second logic netlist laid out in step S4 is implemented for routing, and the routing process is as follows:
[0090] A. First, short lines are routed: the improved A* pathfinding algorithm findpath() is used to find the path, and the clock delay obtained by the auto_set() function is used to assign the clock signal to the path by the generate_sequence() function, so that the line after assigning the clock signal is obtained.
[0091] It should be noted that a 200*200 grid is defined first, a total of 40000 squares, which can be modified according to the actual size of the network, to construct a two-dimensional coordinate system suitable for the second logic netlist, and map the second logic netlist into the two-dimensional coordinate system according to the abscissa and ordinate of each node, so that a node corresponds to a unique coordinate position.
[0092] The improved A* pathfinding algorithm compares f(n)=g(n)+h(n) of the non-obstacle squares around the current square and selects the square with the shortest path length.
[0093] g(n): the actual cost from the starting point to node n (known path length);
[0094] h(n): heuristic estimated cost from node n to the target node (predicted path length).
[0095] Improvement: explore the volume of the surrounding grid, add the estimated cost of the grid with large volume, and make the pathfinding algorithm prefer the grid without placing wires and gates. Explore whether the surrounding grid meets the obstacle condition, if it meets, set it as an obstacle and call findpath() again;
[0096] Due to the nature of the A* algorithm, it will only find a fixed path without changing other conditions, and sometimes the path cannot be assigned a correct clock signal in the subsequent code. At this time, only the path can be changed, and the path can only be changed by adding obstacles. The following obstacles are added while exploring one side in the pathfinding algorithm findpath(). This avoids the time waste caused by the inability to correctly assign a clock signal after finding a path, which leads to re-searching.
[0097] Rules for adding obstacles (all added obstacles are only for the line being laid, and the obstacles are removed after the layout is completed):
[0098] a. Check the two determined clocks in the path (these two determined clocks are due to the path passing through them when laying other lines, and a determined clock is given). If it does not meet the clock rule, set the second grid as an obstacle;
[0099] b. Set the grid where all other logic gates (i.e. other than the starting and ending points of the line) are located as an obstacle;
[0100] c. Check if the surrounding of a grid is all obstacles, if so, set the grid as an obstacle. A* algorithm may fail to explore further due to the surrounding obstacles after exploring a certain node, so detecting such a grid in advance and adding it as an obstacle can avoid the situation of a dead end.
[0101] The generate_sequence() function uses recursion and memoization search methods, combined with various constraint conditions, to gradually try to generate a sequence that meets the target delay. Constraint conditions include but are not limited to the clock signal at the start of the path being abstracted as the first position of the sequence, the clock signal at the end of the path being abstracted as the last position of the sequence, the length of the path being abstracted as the length of the sequence, such as a sequence with a first position of 2, a last position of 1, and a length of 8. The target clock delay is 4, so the sequence can be 22330001 or 23330001, etc. If the target delay is 8, it can only be 23012301, and the clock signal can only be 0123. The function of this function is to get such a sequence that meets the requirements.
[0102] B. Re-route a long line: The clock delays of the several layers spanned by the long line are added together as the clock delay of the long line. Under normal circumstances, the routing is almost the same as that of the short line. However, there may be cases where the path length is less than the clock delay. In this case, the generate_sequence() function will not be able to obtain a feasible clock sequence for the path. Therefore, when the above situation is detected, the findpath_2() function will be used instead. The findpath_2() function can ensure that a path with a path length greater than the clock delay is obtained.
[0103] Once the path is found, it is the same as the short line. The clock delay is then passed to the generate_sequence() function to assign a clock signal to the path.
[0104] The findpath_2() function consists of three stages:
[0105] Phase 1: Downward Extension Phase. First, extend downwards a certain distance to avoid the second phase's horizontal extension causing excessive space occupation around the starting point, which could lead to wiring failure.
[0106] Phase Two: Lateral Extension Phase. The purpose of this phase is to extend the path length. It will automatically stop when the predicted distance reaches the input path length and then start Phase Three.
[0107] The third stage: Reaching the destination. The A* algorithm is used to extend the path to the destination, ultimately obtaining a path that meets the required length.
[0108] After all upward connections (including short and long lines) of the current level are completed, the level is incremented by 1. This process continues until all connections between levels are completed, ending the layout and routing. For example, if the current level is 4, first traverse to all nodes at level 3 to find the short upward lines (i.e., nodes at this level are the endpoints of short lines). Then traverse to nodes at levels 0, 1, and 2 to find all long upward lines. Place the short lines first, then the long lines. At this point, all connections between nodes in this level and nodes in the levels above are complete. Then set the level to 5 and repeat this process until the maximum level is reached, completing the routing for all nodes.
[0109] Specifically, such as Figure 3 As shown, taking a 4-to-1 multiplexer as an example, the layout and routing of the 4-to-1 multiplexer are performed according to steps S1-S4, as follows: Figure 4 and Figure 5 As shown, Figure 4 This reflects the fan-in and fan-out behavior between nodes, providing... Figure 3 The first logical netlist obtained after preprocessing, where 16 is a virtual node added during the preprocessing process. Figure 5 It is based on Figure 4 The layout is specifically configured with each line path and clock, inFigure 5 In the figure, green represents clock 0, pink represents clock 1, blue represents clock 2, white represents clock 3, and black dots represent virtual nodes.
[0110] Referring to Figures 1-6 The application provides a layered layout and wiring system based on a quantum cellular automaton circuit, which comprises:
[0111] A data acquisition module is configured to acquire an original logic netlist, wherein the original logic netlist comprises first logic ordering between multiple nodes, first fan-in and fan-out information of each node, and a first level.
[0112] A data processing module is configured to pre-process the original logic netlist to obtain a first logic netlist layered according to logic depth.
[0113] A data layout module is configured to eliminate long lines between layers that do not conform to global clock synchronization and optimize cross lines according to a preset layout strategy, so as to obtain a second logic netlist, wherein the second logic netlist comprises second logic ordering between multiple nodes, second fan-in and fan-out information of each node, and a second level.
[0114] A data wiring module is configured to dynamically wire the multiple nodes in the second logic netlist according to a preset wiring strategy and allocate clock signals.
[0115] The above description is only a preferred embodiment of the application, but the protection scope of the application is not limited to this. Any person skilled in the art can make equivalent replacements or changes to the technical solution and the inventive concept of the application within the technical range disclosed by the application, which should be covered by the protection scope of the application.
Claims
1. A method for hierarchical placement and routing based on quantum cellular automata circuit, characterized in that, The method comprises the following steps: obtaining an original logical netlist, the original logical netlist comprising a first logical order between a plurality of nodes, first fan-in and fan-out information of each node, and a first hierarchy; preprocessing the original logical netlist to obtain a first logical netlist layered according to logical depth; eliminating long lines between layers that do not conform to global clock synchronization and optimizing cross lines according to a preset layout strategy to obtain a second logical netlist, the second logical netlist comprising a second logical order between a plurality of nodes, second fan-in and fan-out information of each node, and a second hierarchy; dynamically routing the plurality of nodes in the second logical netlist according to a preset routing strategy and assigning clock signals.
2. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 1, wherein, The preprocessing specifically comprises: According to the logical depth of the nodes, the plurality of nodes in the original logical netlist is divided into n layers to generate a hierarchy set {L0, L1, …, L n} and the plurality of nodes is numbered so that one node corresponds to one number. For any node in the level set {L0, L1, …, L n}, if the number of fan-out nodes is greater than 3, add a virtual node to it, and distribute part of the fan-out to the virtual node, so that the fan-out of each node is less than or equal to 3. The level set {L0', L1', …, L n ′} after adjusting the node fan-out is taken as the first logical netlist.
3. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 1, wherein, The preset layout strategy specifically comprises: calculating the Manhattan distance of the physical coordinates of the node pairs of adjacent layers in the first logical netlist, and comparing the Manhattan distance of each node with a preset threshold one by one; when any Manhattan distance is greater than the preset threshold, the logical line between the related nodes is regarded as a long line between layers; adjusting the node order in the same hierarchy layer to eliminate the long line between layers; in the process of adjusting the node order, the number of cross lines is calculated synchronously, the target is balanced through a weighted evaluation function, and the node order in each layer is optimized layer by layer to generate the second logical netlist after optimization.
4. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 3, wherein, The weighted evaluation function specifically comprises: where N cross represents the number of cross lines between the i-th layer and the j-th layer hierarchy; P ij represents the penalty value of layer sequence constraint, ensuring that the i-th layer hierarchy appears before the j-th layer hierarchy; q is the evaluation function value.
5. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 4, wherein, The second logical netlist after optimization is generated by balancing the target through the weighted evaluation function and optimizing the node order in each layer, specifically comprising: under the premise of ensuring the correct layer sequence, the number of interlayer cross lines is reduced as the target, the nodes in the same hierarchy layer are sorted according to the evaluation function value q, and each hierarchy is traversed to generate the second logical netlist after optimization, wherein the number of cross lines is the least and the sorting is the optimal when the evaluation function value q is the least.
6. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 1, wherein, The preset routing strategy specifically comprises: arranging the plurality of nodes contained in the second logical netlist in a hierarchy arrangement set according to the order from small to large of the specific hierarchy number of the second hierarchy; classifying the logical connections between the nodes in each second hierarchy in the hierarchy arrangement set and the nodes in other second hierarchies into a short line set and a long line set to obtain the short line set and the long line set; dynamically adjusting the layer spacing and clock delay between each second hierarchy in the hierarchy arrangement set based on the short line set; allocating clock signals based on the layer spacing and clock delay of each second hierarchy in the hierarchy arrangement set.
7. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 6, wherein, The classification of the logical connections between the nodes in each second hierarchy in the hierarchy arrangement set and the nodes in other second hierarchies into a short line set and a long line set to obtain the short line set and the long line set specifically comprises: finding the fan-in nodes of each node in each second hierarchy in the hierarchy arrangement set, and if the hierarchy number corresponding to any fan-in node and the second hierarchy of the node differ by 1, the logical connection formed by the node and the fan-in node is added to the short line set, otherwise the logical connection formed by the node and the fan-in node is added to the long line set; traversing each second hierarchy in the hierarchy arrangement set according to the above process to obtain the short line set and the long line set corresponding to the hierarchy arrangement set.
8. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 6, wherein, The method dynamically adjusts the layer spacing and clock delay between each second hierarchical level in the hierarchical arrangement set based on the short line set, and specifically comprises the following steps. Any two adjacent second hierarchical levels in the hierarchical arrangement set are selected; The layer spacing is initialized as 1, and all lines between the two layers are traversed to determine the longest and shortest Manhattan distance lines. The layer spacing is increased, and the maximum clock delay of the shortest line and the minimum clock delay of the longest line are iteratively calculated. The layer spacing and clock delay corresponding to the case that the maximum clock delay is equal to the minimum clock delay are taken as the layer spacing and clock delay corresponding to the current two adjacent second hierarchical levels. The above process is used to traverse all adjacent second hierarchical levels in the hierarchical arrangement set to obtain the layer spacing and clock delay corresponding to each second hierarchical level and the adjacent second hierarchical level.
9. The method for hierarchical placement and routing based on quantum cellular automata circuit according to claim 6, wherein, The clock signal is distributed based on the layer spacing and clock delay of each second hierarchical level in the hierarchical arrangement set, and specifically comprises the following steps. The horizontal coordinates of the nodes in each second hierarchical level are constructed according to the ordering of the nodes, and the vertical coordinates of the nodes in each second hierarchical level are determined based on the layer spacing of each second hierarchical level in the hierarchical arrangement set. The logic gate device is configured and the corresponding clock signal is generated based on the horizontal coordinates, vertical coordinates and corresponding clock delay of each node.
10. A system for hierarchical placement and routing based on quantum cellular automata circuits, the system comprising: The method comprises the following steps: The data acquisition module is configured to acquire an original logic netlist, wherein the original logic netlist comprises a first logic ordering between multiple nodes, first fan-in and fan-out information of each node, and a first hierarchical level. The data processing module is configured to preprocess the original logic netlist to obtain a first logic netlist arranged according to a logic depth. The data layout module is configured to eliminate long lines between layers that do not conform to global clock synchronization and optimize cross lines according to a preset layout strategy, so as to obtain a second logic netlist, wherein the second logic netlist comprises a second logic ordering between multiple nodes, second fan-in and fan-out information of each node, and a second hierarchical level. The data routing module is configured to dynamically route the multiple nodes in the second logic netlist according to a preset routing strategy and distribute a clock signal.
Citation Information
Cited By
Internal module diagram-oriented hierarchical automatic layout method and device
CN121093650A
