A grid traversal method, device and equipment based on dual gradient enhanced regressor

Abstract

The present application relates to the technical field of spatial arrangement design of electronic devices, and in particular to a grid traversal method, device and equipment based on a dual gradient enhanced regressor. The method comprises: obtaining a target spatial design set; generating a target area according to the target spatial design set; selecting a sub-area to be traversed, and initializing a traversal of each spatial design grid point to obtain at least one reference grid point; determining the next sub-area to be traversed according to each reference grid point, and traversing the spatial design grid points in the next sub-area to be traversed, and predicting the performance parameters of each candidate grid point; obtaining the actual performance parameters of the target candidate grid point; updating one or more reference grid points; repeating the above steps until all spatial design grid points are traversed to find the reference grid point with the best performance parameters. The present application can reduce the number of comprehensive runs, reduce the computational cost of spatial design exploration, and at the same time, the reference grid point finally updated has the best performance parameters.

Description

Technical Field

[0001] This application relates to the field of spatial layout design technology for electronic devices, and in particular to a grid traversal method, apparatus and equipment based on a dual gradient enhancement regressor. Background Technology

[0002] Design space exploration (DSE) aims to identify the implementation scheme with the best quality characteristics, i.e., the best design parameters, while satisfying all imposed design constraints. In the field of electronic devices, such as integrated circuits and chips, design space exploration is used to find the most suitable design parameters. Existing methods use high-level synthesis to comprehensively traverse all design spaces to identify Pareto optimal designs. However, this approach requires running high-level synthesis tools many times to identify the performance parameters of different space designs. Due to the multi-dimensional nature of design and the high computational cost during synthesis, existing space design exploration becomes a difficult and arduous task. Therefore, existing technologies suffer from low efficiency and high computational and exploration costs. Summary of the Invention

[0003] This application provides a grid traversal method, apparatus, and device based on a dual-gradient enhanced regressor to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions.

[0004] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0005] According to one aspect of the embodiments of this application, a grid traversal method based on a dual-gradient enhanced regressor is provided, the method comprising:

[0006] S1, Obtain the target space design set, which consists of several space design instruction sets;

[0007] S2, Generate a target region according to the target space design set, wherein the target region includes space design grid points that correspond one-to-one with each of the space design instruction sets;

[0008] S3, select a sub-region to be traversed within the target area, and perform initial traversal on each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point, wherein the performance parameters of the reference grid point meet the preset performance conditions.

[0009] S4. Determine the next sub-region to be traversed based on each of the reference grid points, and traverse each spatial design grid point in the next sub-region to be traversed to obtain candidate grid points adjacent to each of the reference grid points.

[0010] S5, the performance parameters of each candidate grid point are predicted by a dual gradient enhanced regressor to obtain the performance prediction parameters of each candidate grid point;

[0011] S6, compare the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point. If a target candidate grid point is found whose performance prediction parameters are better than or partially better than any of the performance parameters, then obtain the actual performance parameters of each target candidate grid point.

[0012] S7. Based on a comprehensive comparison of the actual performance parameters of each of the target candidate grid points and the performance parameters of each of the reference grid points, if a target grid point is found whose actual performance parameters are better than or partially better than any of the performance parameters, then one or more of the reference grid points are updated based on the target grid point.

[0013] S8. Repeat steps S4-S7 until all spatial design grid points within the target area have been traversed to find the reference grid point with the optimal performance parameters.

[0014] In one embodiment of this application, based on the foregoing scheme, the initial traversal of each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point includes:

[0015] The performance parameters of each spatial design grid point in the sub-region to be traversed are obtained by using high-level synthesis tools. The performance parameters include the time delay and area for completing the spatial design.

[0016] By comprehensively comparing the various performance parameters, one or more of the reference grid points with the lowest latency and / or smallest area are determined;

[0017] The preset performance condition is that, among the traversed spatial design grid points, there is a preset non-dominant relationship between any two reference grid points.

[0018] In one embodiment of this application, based on the foregoing scheme, determining the next sub-region to be traversed according to each of the reference grid points includes:

[0019] If the number of reference grid points is one, then the next sub-region to be traversed is constructed with the reference grid point as the starting point, having a first preset shape and a first preset size.

[0020] If there are multiple reference grid points, a starting region is constructed based on the location of each reference grid point, and the next sub-region to be traversed is constructed based on the starting region with a second preset shape and a second preset size.

[0021] In one embodiment of this application, based on the foregoing scheme, the dual-gradient enhancement regressor includes a first group of regressors related to time delay and a second group of regressors related to area. The step of predicting performance parameters for each candidate grid point using the dual-gradient enhancement regressor to obtain the performance prediction parameters for each candidate grid point includes:

[0022] The first group of regressors is used to predict the time delay of each candidate grid point to obtain the predicted time delay of each candidate grid point.

[0023] The area of ​​each candidate grid point is predicted by the second group of regressors.

[0024] The performance prediction parameters include the prediction latency and the prediction area.

[0025] In one embodiment of this application, based on the foregoing scheme, if a target candidate grid point with a performance prediction parameter that is better than or partially better than any of the performance parameters is obtained, then obtaining the actual performance parameters of each of the target candidate grid points includes:

[0026] If a target candidate grid point is obtained whose predicted latency is lower than that of any of the reference grid points, the actual latency of each target candidate grid point is obtained through the high-level synthesis tool.

[0027] If a target candidate grid point is obtained whose predicted area is smaller than the area of ​​any of the reference grid points, then the actual area of ​​each target candidate grid point is obtained through the high-level synthesis tool.

[0028] If a target candidate grid point is obtained where the predicted latency is lower than the latency of any of the reference grid points and the predicted area is smaller than the area of ​​any of the reference grid points, then the actual latency and actual area of ​​each target candidate grid point are obtained through the high-level synthesis tool.

[0029] In one embodiment of this application, based on the foregoing scheme, if a target grid point is obtained where the actual performance parameter is better than or partially better than any of the performance parameters, then updating one or more of the reference grid points according to the target grid point includes:

[0030] If a target grid point is found whose actual delay is lower than that of any of the reference grid points, then a target reference grid point whose delay is higher than the actual delay is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

[0031] If a target grid point with an actual area smaller than the area of ​​any of the reference grid points is found, then a target reference grid point with an area greater than the actual time delay is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

[0032] If a target grid point is found whose actual delay and actual area are both less than the delay and area of ​​any of the reference grid points, then a target reference grid point with a delay higher than the actual delay and an area greater than the actual area is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

[0033] In one embodiment of this application, based on the foregoing scheme, after comprehensively comparing the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point, the method further includes:

[0034] If the prediction delay of each candidate grid point is higher than the delay of each reference grid point and the prediction area of ​​each candidate grid point is greater than the area of ​​each reference grid point, then the candidate grid point is marked as a poor grid point.

[0035] In one embodiment of this application, based on the foregoing scheme, after all spatial design grid points within the target area have been traversed, the method further includes:

[0036] Obtain the abandoned grid points that were not predicted for performance parameters by the dual-gradient augmented regressor.

[0037] The time delay and area of ​​each of the inferior grid points and the abandoned grid points are obtained through the high-level synthesis tool.

[0038] If the delay of any of the inferior grid points or the abandoned grid points is lower than the delay of the reference grid point and / or the area of ​​any of the inferior grid points or the abandoned grid points is smaller than the area of ​​the reference grid point, then the inferior grid point or the abandoned grid point whose delay is lower than the delay of the reference grid point and / or whose area is smaller than the area of ​​the reference grid point is marked as the reference grid point.

[0039] According to one aspect of the embodiments of this application, a grid traversal device based on a dual-gradient enhanced regressor is provided, the device comprising:

[0040] The acquisition unit is used to acquire a target space design set, which consists of several space design instruction sets;

[0041] A generation unit is used to generate a target region based on the target spatial design set, wherein the target region includes spatial design grid points that correspond one-to-one with each of the spatial design instruction sets;

[0042] The selection unit is used to select a sub-region to be traversed within the target area and perform initial traversal on each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point.

[0043] The determining unit is used to determine the next sub-region to be traversed based on each of the reference grid points, and to traverse each spatial design grid point in the next sub-region to be traversed to obtain candidate grid points adjacent to each of the reference grid points.

[0044] The prediction unit is used to predict the performance parameters of each candidate grid point through a dual gradient enhanced regressor, and obtain the performance prediction parameters of each candidate grid point.

[0045] The first comprehensive comparison unit is used to comprehensively compare the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point. If a target candidate grid point is found to have a performance prediction parameter that is better than or partially better than any of the performance parameters, then the actual performance parameters of each target candidate grid point are obtained.

[0046] The second comprehensive comparison unit is used to comprehensively compare the actual performance parameters of each of the target candidate grid points with the performance parameters of each of the reference grid points. If a target grid point is found whose actual performance parameters are better than or partially better than any of the performance parameters, then one or more of the reference grid points are updated based on the target grid point.

[0047] An iterative unit is used to repeat the above steps until all spatial design grid points within the target area have been traversed to find the reference grid point with the optimal performance parameters.

[0048] According to one aspect of the embodiments of this application, an electronic device is provided, including: one or more processors; and a memory for storing executable instructions of the processors, which, when executed by the one or more processors, cause the one or more processors to perform the method as described in the above embodiments.

[0049] The beneficial effect of this application is that by obtaining a pre-set target space design set, which consists of several space design instruction sets, and each space design instruction set corresponds to a space design, the application needs to find the space design with the best performance among the various space design instruction sets in the target space design set, and then manufacture electronic devices based on the space design with the best quality characteristics found.

[0050] The difference between this application and existing spatial design exploration methods lies in the fact that it obtains the performance parameters of a partial region through initial traversal, that is, by performing the first high-level synthesis. This is the performance parameter of all spatial design grid points in the sub-region to be traversed, as described in this application. This allows for the acquisition of one or more reference grid points with optimal performance. It should be noted that the performance parameters can be composed of multiple evaluation parameters. The reference grid point with optimal performance may represent that a certain evaluation parameter of the reference grid point is optimal, while other evaluation parameters are not better than the evaluation parameters of other reference grid points. Therefore, there may be one or more reference grid points with optimal performance.

[0051] Furthermore, after the initial traversal of the sub-region to be traversed is completed, the next sub-region to be traversed is determined. By traversing each spatial design grid point in the next traversal sub-region, that is, finding candidate grid points adjacent to each of the reference grid points, since the performance parameters of adjacent candidate grid points are similar to those of the reference grid points, the performance parameters of the candidate grid points are continuously compared with those of the reference grid points to find spatial design grid points with better performance parameters, so as to update the reference grid points in the future.

[0052] Furthermore, performance parameters of each candidate grid point are predicted using a dual-gradient augmented regressor without requiring high-level synthesis tools to actually obtain the performance parameters. This reduces computational costs during spatial design exploration. The dual-gradient augmented regressor method also effectively improves traversal efficiency, quickly predicting the performance parameters of candidate grid points. These predicted parameters are then compared with those of reference grid points. If a candidate grid point has performance parameters superior to one or more reference grid points, it is marked as a target candidate grid point. The actual performance parameters of this target candidate grid point are then obtained using high-level synthesis tools and compared with those of the reference grid points. If the target candidate grid point outperforms or partially outperforms the reference grid point, the reference grid point is updated, meaning it is now marked as a reference grid point.

[0053] Furthermore, by continuously repeating steps S4-S7, the next sub-region to be traversed can be continuously determined, that is, the spatial design grid points within the target region can be continuously traversed. If a spatial design grid point with better performance parameters is found during the traversal process, it is marked as a reference grid point and updated until all spatial design grid points within the target region have been traversed. In this way, the marked or updated reference grid point is the reference grid point with the best performance parameters. Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly explained below. Obviously, the described drawings are only a part of the embodiments of this application, and not all of them. Those skilled in the art can obtain other design schemes and drawings based on these drawings without creative effort.

[0055] Figure 1 This is a flowchart illustrating a grid traversal method based on a dual-gradient enhancement regressor according to an embodiment of this application;

[0056] Figure 2 This is a schematic diagram illustrating the logic of traversing a target region according to an embodiment of this application;

[0057] Figure 3 The following is a flowchart illustrating the grid traversal method based on a dual-gradient enhanced regressor according to an embodiment of this application;

[0058] Figure 4 This is a block diagram of a grid traversal device based on a dual-gradient enhancement regressor, according to an embodiment of this application.

[0059] Figure 5 This is a schematic diagram of the system structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0060] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that this application will be more complete and comprehensive, and will fully convey the concept of the exemplary embodiments to those skilled in the art.

[0061] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0062] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller node devices.

[0063] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0064] It should be noted that "multiple" in this article refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0065] The background technology of this application is described in detail below:

[0066] Design space exploration (DSE) methods can be broadly categorized into two types: model-based methods and synthesis-based methods. Model-based methods include supervised learning-based and graph analysis-based methods. Supervised learning-based methods rely on knowledge of previous synthesized designs (the relationship between design configurations and characteristics) for offline supervised model training. The trained model possesses rich pre-screening capabilities, approximating the characteristics of new designs and thus reducing the synthesis burden. Graph analysis-based methods use prediction techniques on the Control Data Flow Diagram (CDFG) to simulate the behavior of High-Level Synthesis (HLS) tools without prior knowledge of the synthesized design. Then, techniques such as graph neural networks are used to infer the characteristics of the input design based on the CDFG. With the assistance of some intelligent sampling strategies, this method can quickly generate design space exploration results, but it suffers from deficiencies in maintenance and accuracy.

[0067] Synthesis-based methods rely on High-level Synthesis (HLS) tools to perform the synthesis process, acquiring design characteristics and using heuristic algorithms (specific heuristics or metaheuristics) or prior knowledge to identify important regions. New candidate designs are continuously generated within these regions, and the most promising design is selected for synthesis. As the number of synthesis iterations increases, the quality of the Design Space Evaluation (DSE) results is iteratively optimized until the exit criteria are met. Synthesis-based methods, due to their reliance on the synthesis process, offer better accuracy and flexibility. However, their limited learning capacity leads to a higher synthesis burden, which can be further exacerbated if candidate selection is inappropriate. Furthermore, synthesis-based methods only utilize the characteristic information of the synthesized design to identify important regions, neglecting the synthesis knowledge generated during the design space exploration process.

[0068] Therefore, based on the aforementioned shortcomings, this application proposes a grid traversal method based on dual gradient boosting regression (DGBR-Lattice), which combines learning and optimization capabilities, effectively solving the design space exploration (DSE) problem in high-level synthesis. DGBR-Lattice utilizes the synthesis knowledge learned by two gradient boosting regressions (GBRs) based on ensemble learning techniques—corresponding to the delay-related first regression group and the area-related second regression group described in this application—to assist the grid traversal method in identifying the most promising designs during the optimization process (traversal process). DGBR-Lattice can achieve near-true Pareto fronts while reducing the number of synthesis tool runs by at least 1.42 times and up to 9.62 times, significantly reducing the computational cost of space design exploration.

[0069] The implementation details of the technical solutions in the embodiments of this application are described in detail below:

[0070] According to one aspect of the embodiments of this application, a grid traversal method based on a dual-gradient enhanced regressor is provided. Figure 1 The flowchart below shows a grid traversal method based on a dual-gradient enhancement regressor according to an embodiment of this application. The method includes at least steps S1 to S8, which are described in detail below:

[0071] In step S1, a target space design set is obtained, which consists of several space design instruction sets.

[0072] Specifically, in HLS (High-Level Synthesis), space design instructions (i.e., #pragma instructions) and specific behavioral descriptions are mapped to RTL (Register Transfer Level) microarchitecture by invoking the high-level synthesis tools to execute the synthesis process (synthesis run). Given a specific behavioral description, by selecting different combinations of #pragma instruction values ​​(i.e., space design), RTL outputs with different performance parameters (latency and area) can be obtained.

[0073] In other words, the target space design set contains multiple space design instruction sets, and a single space design instruction set is specifically defined by the value of the space design instruction (i.e., the #pragma instruction). That is, a space design instruction set is composed of instructions under multiple different dimensions, which can be represented by the following formula:

[0074]

[0075] Wherein, DS stands for Space Design Instruction Set. , ... Each of these is a space design directive, and a space design directive is composed of multiple #pragma directive values. A single space design directive can be represented by the following formula:

[0076]

[0077] in, For specific instruction values, since the instruction values ​​of space design instructions grow exponentially, the time required to call synthesis tools to execute the synthesis process, i.e., the number of synthesis runs, ranges from several minutes to several hours. Therefore, it is very necessary to propose an intelligent and efficient design space exploration (DSE) method in order to find a high-performance, low-cost Pareto design (the best / ideal space design) while reducing the number of synthesis runs.

[0078] In step S2, a target region is generated based on the target spatial design set. The target region includes spatial design grid points that correspond one-to-one with each of the spatial design instruction sets.

[0079] Specifically, it can be like Figure 2 As shown in a, Figure 2 a, b, c, or d in the example are target regions, where the target regions have multiple spatial design grid points ( Figure 2 The design group in the diagram (i.e., the blue grid points and the red reference grid points) represents the design. The area on the horizontal axis and the time delay on the vertical axis are the performance parameters, or evaluation parameters, for each spatial design grid point. Each spatial design grid point corresponds to a spatial design, which is DS in the formula above.

[0080] First, after initializing the traversal, we get... Figure 2 The diagram in Figure a shows four reference grid points. There is a non-dominant relationship between each reference grid point. It should be noted that in this embodiment of the application, a non-dominant relationship is defined between each reference grid point. That is to say, the non-dominant relationship is one of the conditions of the preset performance conditions.

[0081] In step S3, a sub-region to be traversed is selected within the target area, and each spatial design grid point within the sub-region to be traversed is initialized and traversed to obtain at least one reference grid point. The performance parameters of the reference grid point meet the preset performance conditions.

[0082] Specifically, the preset performance conditions can be non-dominant relationships as mentioned above. As long as a spatial design grid point has a non-dominant relationship with any reference grid point, it means that the spatial design grid point meets the preset performance conditions and can be marked as a reference grid point.

[0083] The sub-region to be traversed can be any region within the target region. The area and shape of the sub-region can be set as needed. High-level synthesis tools can be used to obtain the performance parameters (latency and area) of each spatial design grid point within the sub-region, and then the reference grid points—one or more of the best-performing grid points within the sub-region—can be derived. Figure 2 As shown in 'a', there are four reference grid points, and there are non-dominant relationships between each reference grid point.

[0084] In one embodiment of this application, the initial traversal of each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point includes:

[0085] The performance parameters of each spatial design grid point in the sub-region to be traversed are obtained by using high-level synthesis tools. The performance parameters include the time delay and area for completing the spatial design.

[0086] By comprehensively comparing the various performance parameters, one or more of the reference grid points with the lowest latency and / or smallest area are determined;

[0087] The preset performance condition is that, among the traversed spatial design grid points, there is a preset non-dominant relationship between any two reference grid points.

[0088] Specifically, the non-dominant relation can be defined as follows: select any reference grid point as the first reference grid point, and select any reference grid point among all the reference grid points other than the first reference grid point as the second reference grid point.

[0089] If the latency of the first reference grid point is higher than that of the second reference grid point and the area of ​​the first reference grid point is smaller than that of the second reference grid point, or if the latency of the first reference grid point is lower than that of the second reference grid point and the area of ​​the first reference grid point is greater than that of the second reference grid point, this is called a non-dominant relationship.

[0090] Conversely, the dominance relationship can be specifically defined as follows: if the delay and area of ​​the first reference grid point are both smaller than those of the second reference grid point, then the first reference grid point can be considered to dominate the second reference grid point. In this case, all performance parameters of the first reference grid point are superior to those of the second reference grid point.

[0091] Alternatively, if the delay and area of ​​the first reference grid point are both greater than those of the second reference grid point, it can be considered that the second reference grid point can dominate the first reference grid point. In this case, all performance parameters of the second reference grid point are superior to those of the first reference grid point.

[0092] In step S4, the next sub-region to be traversed is determined based on each of the reference grid points, and each spatial design grid point in the next sub-region to be traversed is traversed to obtain candidate grid points adjacent to each of the reference grid points.

[0093] Specifically, the next sub-region to be traversed can also be set according to actual needs. After traversing the current next sub-region to be traversed, the next sub-region to be traversed in the next round will be determined. By continuously exploring the untraversed areas in the target area (finding reference grid points / updating reference grid points), the spatial design grid point with the best performance parameters is found during the exploration process, and iterative updates are performed. Finally, after the target area is traversed, the spatial design grid point with the best performance design can be obtained, which is the reference grid point after continuous updates and iterations.

[0094] In one embodiment of this application, determining the next sub-region to be traversed based on each of the reference grid points includes:

[0095] If the number of reference grid points is one, then the next sub-region to be traversed is constructed with the reference grid point as the starting point, having a first preset shape and a first preset size.

[0096] If there are multiple reference grid points, a starting region is constructed based on the location of each reference grid point, and the next sub-region to be traversed is constructed based on the starting region with a second preset shape and a second preset size.

[0097] Specifically, the next sub-region to be traversed can be referred to Figure 2 The circular area corresponding to the search range shown in b is just an example. It should be noted that the next sub-region to be traversed can also be set as a rectangle, polygon, or irregular area.

[0098] Specifically, if only one reference grid point is found during the current traversal, it means that the latency and area of ​​this reference grid point are both less than those of the other spatial design grid points that have been traversed. If multiple reference grid points appear, it means that there is a non-dominant relationship between these reference grid points, that is, the performance parameters of each reference grid point are better than those of other reference grid points. For example, one reference grid point may have a lower latency but a larger area, while another reference grid point may have a higher latency but a smaller area.

[0099] It should be noted that latency, as an evaluation parameter or a performance parameter, corresponds to the time required to complete the spatial design of electronic devices. Area, on the other hand, is easier to understand; it represents the area required to complete the spatial design.

[0100] The example of the starting region is illustrated below. It involves determining the starting region based on at least two reference grid points, and then using the midpoint of this starting region as the starting point, setting a traversal region of arbitrary step size as the next sub-region to be traversed. The first and second preset shapes can be arbitrarily set. Figure 2 The shape shown is circular, but the first and second preset dimensions can be set arbitrarily. Figure 2 The area shown is a circular region with a radius of r.

[0101] In step S5, the performance parameters of each candidate grid point are predicted by a dual gradient boosting regressor to obtain the performance prediction parameters of each candidate grid point.

[0102] Specifically, it can be like Figure 3 As shown, Figure 3 The following is a flowchart of the logic of this application: after finding candidate grid points, they are archived, and then each candidate grid point is input into regressor group 1 and regressor group 2.

[0103] In one embodiment of this application, the dual-gradient augmented regressor includes a first group of regressors related to time delay and a second group of regressors related to area. The step of predicting performance parameters for each candidate grid point using the dual-gradient augmented regressor to obtain performance prediction parameters for each candidate grid point includes:

[0104] The first group of regressors is used to predict the time delay of each candidate grid point to obtain the predicted time delay of each candidate grid point.

[0105] The area of ​​each candidate grid point is predicted by the second group of regressors.

[0106] The performance prediction parameters include the prediction latency and the prediction area.

[0107] Specifically, Figure 3 In the diagram, regressor group 1 is the first regressor group related to time delay, and regressor group 2 is the second regressor group related to area. Figure 3 The selected space design instruction values ​​d1, d2...d n Corresponding to DS in the above formula , This involves using a dual-gradient augmented regressor to predict the time delay and area of ​​each instruction value, resulting in the predicted time delay and area for each candidate grid point. (See reference...) Figure 2 As shown in c, the nearest design is found as the target candidate grid point. Figure 2 In this context, d indicates that the comprehensive design (i.e., the target grid point obtained in the current iteration) has been found, and the comprehensive design (target grid point) is then updated to the reference grid point.

[0108] In step S6, the performance prediction parameters of each candidate grid point are compared with the performance parameters of each reference grid point. If the performance prediction parameters of a target candidate grid point are better than or partially better than any of the performance parameters, the actual performance parameters of each target candidate grid point are obtained.

[0109] Specifically, the step of comprehensively comparing the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point includes:

[0110] The predicted delay of each candidate grid point is compared with the delay of each reference grid point, and the predicted area of ​​each candidate grid point is compared with the area of ​​each reference grid point.

[0111] In one embodiment of this application, if a target candidate grid point is found to have a performance prediction parameter that is better than or partially better than any of the performance parameters, then obtaining the actual performance parameters of each of the target candidate grid points includes:

[0112] If a target candidate grid point is obtained whose predicted latency is lower than that of any of the reference grid points, the actual latency of each target candidate grid point is obtained through the high-level synthesis tool.

[0113] If a target candidate grid point is obtained whose predicted area is smaller than the area of ​​any of the reference grid points, then the actual area of ​​each target candidate grid point is obtained through the high-level synthesis tool.

[0114] If a target candidate grid point is obtained where the predicted latency is lower than the latency of any of the reference grid points and the predicted area is smaller than the area of ​​any of the reference grid points, then the actual latency and actual area of ​​each target candidate grid point are obtained through the high-level synthesis tool.

[0115] Specifically, if a candidate grid point has a predicted latency or area greater than that of any reference grid point, it is marked as a target candidate grid point. Subsequently, a high-level synthesis tool is used to obtain the actual area and latency of the target candidate grid point, which are then compared with the latency and area of ​​the reference grid points. This approach reduces the number of runs required by the high-level synthesis tool. By using a dual-gradient augmented regressor for prediction, it effectively pre-screens potential target candidate grid points before making a final comparison, improving efficiency, reducing computational costs, and ensuring accuracy.

[0116] In step S7, the actual performance parameters of each target candidate grid point are compared with the performance parameters of each reference grid point. If a target grid point is found whose actual performance parameters are better than or partially better than any of the performance parameters, then one or more reference grid points are updated based on the target grid point.

[0117] Specifically, the step of comprehensively comparing the actual performance parameters of each of the target candidate grid points with the performance parameters of each of the reference grid points includes:

[0118] The actual time delay of each of the target candidate grid points is compared with the time delay of each of the reference grid points, and the actual area of ​​each of the target candidate grid points is compared with the area of ​​each of the reference grid points.

[0119] In one embodiment of this application, updating one or more reference grid points based on the target grid point if the actual performance parameter is better than or partially better than any of the performance parameters includes:

[0120] If a target grid point is found whose actual delay is lower than that of any of the reference grid points, then a target reference grid point whose delay is higher than the actual delay is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

[0121] If a target grid point with an actual area smaller than the area of ​​any of the reference grid points is found, then a target reference grid point with an area greater than the actual time delay is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

[0122] If a target grid point is found whose actual delay and actual area are both less than the delay and area of ​​any of the reference grid points, then a target reference grid point with a delay higher than the actual delay and an area greater than the actual area is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

[0123] Specifically, if the delay and area of ​​a target grid point are both smaller than a certain target reference grid point, then the target reference grid point will be marked as a failed grid point, and the target grid point will be marked as a reference grid point. The target grid point is obtained from the target candidate grid points.

[0124] In one embodiment of this application, after comprehensively comparing the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point, the method further includes:

[0125] If the prediction delay of each candidate grid point is higher than the delay of each reference grid point and the prediction area of ​​each candidate grid point is greater than the area of ​​each reference grid point, then the candidate grid point is marked as a poor grid point.

[0126] Specifically, if the prediction delay and prediction area of ​​a candidate grid point are both greater than the areas of each of the reference grid points, then the candidate grid point is marked as a poor-quality grid point.

[0127] In step S8, steps S4-S7 are repeated until all spatial design grid points within the target area have been traversed to find the reference grid point with the optimal performance parameters.

[0128] Specifically, by repeatedly performing steps S4-S7, the next sub-region to be traversed can be continuously determined, which means continuously traversing the spatial design grid points within the target region. If a spatial design grid point with better performance parameters is found during the traversal, it is marked as a reference grid point and updated until all spatial design grid points within the target region have been traversed. In this way, the marked or updated reference grid point is the reference grid point with the best performance parameters.

[0129] In one embodiment of this application, after all spatial design grid points within the target area have been traversed, the method further includes:

[0130] The delay and area of ​​each of the inferior grid points are obtained using the high-level synthesis tool.

[0131] If the delay of any of the inferior grid points is lower than the delay of the reference grid point and / or the area of ​​any of the inferior grid points is smaller than the area of ​​the reference grid point, then the inferior grid point with the delay lower than the delay of the reference grid point and / or the area smaller than the area of ​​the reference grid point is marked as the reference grid point.

[0132] Abandoned grid points are spatial design grid points that are discarded during the traversal process. Inferior grid points are those selected by comprehensively comparing the performance parameters of candidate grid points and reference grid points. To improve the guarantee rate of optimal performance of reference grid points, the actual performance parameters of all inferior and abandoned grid points are obtained by running the synthesis tool once at the end. That is, the latency and area of ​​each inferior and abandoned grid point are obtained by the high-level synthesis tool and then compared. If the latency of any inferior or abandoned grid point is lower than that of the reference grid point and / or the area of ​​any inferior or abandoned grid point is smaller than that of the reference grid point, then the inferior or abandoned grid point can be updated as the reference grid point. This can prevent some reference grid points with better performance parameters from being discarded due to misjudgment based on the dual gradient enhancement regressor.

[0133] The following is a detailed description of the integration operation proposed in this application, that is, the integration process performed by the integration tool:

[0134] The overall framework of the embodiments of this application is as follows: Figure 3 As shown, it mainly consists of two parts: optimization and its learning. Compared with the original grid traversal method, the main innovation of this application lies in training a learning machine (bigradient boosting regressor) online to predict whether the selected design can update the reference frontier (reference grid), rather than performing a cumbersome high-level synthesis tool synthesis operation.

[0135] The optimization part utilizes a grid traversal algorithm (i.e., the grid traversal method based on a dual-gradient augmented regressor proposed in this application) to explore the design space, while the learning part employs a dual-gradient augmented regressor (regressor group 1 and regressor group 2), based on ensemble learning techniques to learn comprehensive knowledge to predict the area and latency of each spatial design instruction set. For example... Figure 3 As shown, each regressor group constructs N regression trees (RT) in a progressive manner. At each stage, a regression tree is fitted to optimize the loss function on the negative gradient. Therefore, the prediction accuracy of both regressor groups 1 and 2 can be improved through loss function optimization. In this application, we use mean squared error (MSE) as the loss function and set N to the size of the spatial design.

[0136] In the overall architecture of this application, all integrated designs are stored in a single file, which serves as a training tool. The comprehensive knowledge. Given a comprehensive design (i.e., the space design instruction set described in this application), the corresponding comprehensive knowledge can be represented as ( , , ),in [ , ..., , ..., ], and Define the configuration, area, and latency of the integrated design separately. During the learning phase, two independent regressors ( ) respectively learned from training samples ( , ) and , Training is performed. As the exploration progresses, regressor group 1 and regressor group 2 will undergo multiple training cycles. In each training cycle, we incorporate the latest comprehensive design (i.e., the target candidate grid points in the next sub-region to be traversed in each round of this application) into the training data.

[0137] A final update will be performed when grid traversal stalls. The necessity of this final update stems from two aspects. First, there may be promising regions that have not yet been explored by the grid traversal algorithm, especially when the search radius is small. Second, true non-dominated relations may be incorrectly identified as dominant relations by double gradient augmentation regression, leading to poor-quality reference fronts. Therefore, once the grid traversal algorithm exits, we will perform a final update (i.e., the aforementioned content regarding obtaining the delay and area of ​​each poor-quality grid point using the high-level synthesis tool) to further improve the quality of the DSE results.

[0138] In summary, the embodiments of this application can perform grid traversal of the target region by continuously traversing the spatial grid points of the next sub-region to be traversed. Through continuous traversal, the reference grid points, i.e., the reference frontier, can be continuously updated. After the target region has been traversed, the obtained reference grid points / reference frontier possess optimal performance parameters. Applied to the manufacturing of existing electronic devices, this can significantly improve the quality and performance characteristics of electronic devices. By using a dual-gradient enhanced regressor for continuous prediction and learning, the prediction accuracy of performance parameters such as time delay and area is high. As an auxiliary reference method, this can greatly reduce the computational cost of space design exploration while improving exploration efficiency. More importantly, subsequent comprehensive comparison based on the predicted performance parameters further improves realism and accuracy. Therefore, the grid traversal method based on a dual-gradient enhanced regressor proposed in this application has the technical effects of reducing computational costs, alleviating overall operational load, improving space exploration efficiency, and improving space exploration accuracy.

[0139] Figure 4The diagram shows a grid traversal device 300 based on a dual gradient enhanced regressor according to an embodiment of this application. The grid traversal device 300 based on a dual gradient enhanced regressor according to an embodiment of this application includes: an acquisition unit 301, a generation unit 302, a selection unit 303, a determination unit 304, a prediction unit 305, a first comprehensive comparison unit 306, a second comprehensive comparison unit 307, and an iteration unit 308.

[0140] The acquisition unit 301 is used to acquire a target space design set, which consists of several space design instruction sets;

[0141] The generation unit 302 is used to generate a target region according to the target spatial design set, wherein the target region includes spatial design grid points that correspond one-to-one with each of the spatial design instruction sets;

[0142] The selection unit 303 is used to select a sub-region to be traversed within the target area and perform initial traversal on each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point.

[0143] The determining unit 304 is used to determine the next sub-region to be traversed based on each of the reference grid points, and to traverse each spatial design grid point in the next sub-region to be traversed to obtain candidate grid points adjacent to each of the reference grid points.

[0144] Prediction unit 305 is used to predict the performance parameters of each candidate grid point through a dual gradient enhanced regressor, and obtain the performance prediction parameters of each candidate grid point.

[0145] The first comprehensive comparison unit 306 is used to comprehensively compare the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point. If a target candidate grid point is found to have a performance prediction parameter that is better than or partially better than any of the performance parameters, then the actual performance parameters of each target candidate grid point are obtained.

[0146] The second comprehensive comparison unit 307 is used to comprehensively compare the actual performance parameters of each of the target candidate grid points with the performance parameters of each of the reference grid points. If a target grid point is found whose actual performance parameters are better than or partially better than any of the performance parameters, then one or more of the reference grid points are updated based on the target grid point.

[0147] The iteration unit 308 is used to repeat the above steps until all spatial design grid points in the target area have been traversed in order to find the reference grid point with the optimal performance parameters.

[0148] In another aspect, this application also provides a computer-readable storage medium storing a program product capable of implementing the methods provided above in this specification. In some possible implementations, various aspects of this application may also be implemented as a program product comprising program code that, when run on a terminal device, causes the terminal device to perform the steps described in the "Embodiment Methods" section of this specification according to various exemplary embodiments of this application.

[0149] The program product for implementing the above-described method according to the embodiments of this application may employ a portable compact disc read-only memory (CD-ROM) and include program code, and may run on a terminal device, such as a personal computer. However, the program product of this application is not limited thereto. In this document, the readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0150] The program product may employ any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0151] Computer-readable signal media may include data signals propagated as part of a carrier wave in baseband, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium, capable of sending, propagating, or transmitting programs for use by or in connection with an instruction execution system, apparatus, or device.

[0152] The program code contained on the readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0153] Program code for performing the operations of this application can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0154] In another respect, this application also provides an electronic device capable of implementing the above-described method.

[0155] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0156] The following reference Figure 5 To describe an electronic device 400 according to this embodiment of the present application. Figure 5 The electronic device 400 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0157] like Figure 5 As shown, the electronic device 400 is manifested in the form of a general-purpose computing device. The components of the electronic device 400 may include, but are not limited to: at least one processing unit 410, at least one storage unit 420, and a bus 430 connecting different system components (including storage unit 420 and processing unit 410).

[0158] The storage unit stores program code that can be executed by the processing unit 410, causing the processing unit 410 to perform the steps described in the "Embodiment Methods" section above according to various exemplary embodiments of this application.

[0159] Storage unit 420 may include a readable medium in the form of a volatile storage unit, such as random access memory (RAM) 421 and / or cache memory 422, and may further include a read-only memory (ROM) 423.

[0160] Storage unit 420 may also include a program / utility 424 having a set (at least one) of program modules 425, such program modules 425 including but not limited to: an operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0161] Bus 430 can represent one or more of several types of bus structures, including a memory cell bus or memory cell control node, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the various bus structures.

[0162] Electronic device 400 can also communicate with one or more external devices 1200 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 400, and / or with any device that enables electronic device 400 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed via input / output (I / O) interface 450. Furthermore, electronic device 400 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 460. As shown, network adapter 460 communicates with other modules of electronic device 400 via bus 430. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 400, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0163] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including instructions to cause a computing device (such as a personal computer, server, terminal device, or network device, etc.) to execute the methods according to the embodiments of this application.

[0164] Furthermore, the above figures are merely illustrative of the processes included in the method according to exemplary embodiments of this application, and are not intended to be limiting. It is readily understood that the processes shown in the above figures do not indicate or limit the temporal order of these processes. Additionally, it is readily understood that these processes may be executed synchronously or asynchronously in multiple modules.

[0165] It should be understood that this application is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A grid traversal method based on a dual-gradient boosting regressor, characterized in that, The method includes: S1, Obtain the target space design set, which consists of several space design instruction sets; S2, Generate a target region according to the target space design set, wherein the target region includes space design grid points that correspond one-to-one with each of the space design instruction sets; S3, select a sub-region to be traversed within the target area, and perform initial traversal on each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point, wherein the performance parameters of the reference grid point meet the preset performance conditions. S4. Determine the next sub-region to be traversed based on each of the reference grid points, and traverse each spatial design grid point in the next sub-region to be traversed to obtain candidate grid points adjacent to each of the reference grid points. S5, the performance parameters of each candidate grid point are predicted by a dual gradient enhanced regressor to obtain the performance prediction parameters of each candidate grid point; S6, compare the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point. If a target candidate grid point is found whose performance prediction parameters are better than or partially better than any of the performance parameters, then obtain the actual performance parameters of each target candidate grid point. S7. Based on a comprehensive comparison of the actual performance parameters of each target candidate grid point and the performance parameters of each reference grid point, if a target grid point is found whose actual performance parameters are better than or partially better than any of the performance parameters, then one or more of the reference grid points are updated based on the target grid point. S8. Repeat steps S4-S7 until all spatial design grid points within the target area have been traversed to find the reference grid point with the optimal performance parameters. The performance parameters include the time delay and area required to complete the spatial design.

2. The grid traversal method based on a dual-gradient enhancement regressor according to claim 1, characterized in that, The initial traversal of each spatial design grid point within the sub-region to be traversed, to obtain at least one reference grid point, includes: The performance parameters of each spatial design grid point in the sub-region to be traversed are obtained by using high-level synthesis tools. The performance parameters include the time delay and area for completing the spatial design. By comprehensively comparing the various performance parameters, one or more of the reference grid points with the lowest latency and / or smallest area are determined; The preset performance condition is that, among the traversed spatial design grid points, there is a preset non-dominant relationship between any two reference grid points; The preset non-dominance relationship is determined through the following steps: Select any one of the reference grid points as the first reference grid point, and select any one of the reference grid points other than the first reference grid point as the second reference grid point; When the delay of the first reference grid point is higher than the delay of the second reference grid point and the area of ​​the first reference grid point is smaller than the area of ​​the second reference grid point, the non-dominant relationship between the first reference grid point and the second reference grid point is determined. When the delay of the first reference grid point is lower than the delay of the second reference grid point and the area of ​​the first reference grid point is greater than the area of ​​the second reference grid point, the non-dominant relationship between the first reference grid point and the second reference grid point is determined.

3. The grid traversal method based on a dual-gradient enhancement regressor according to claim 2, characterized in that, The step of determining the next sub-region to be traversed based on each of the reference grid points includes: If the number of reference grid points is one, then the next sub-region to be traversed is constructed with the reference grid point as the starting point, having a first preset shape and a first preset size. If there are multiple reference grid points, a starting region is constructed based on the location of each reference grid point, and the next sub-region to be traversed is constructed based on the starting region with a second preset shape and a second preset size.

4. The grid traversal method based on a dual-gradient enhancement regressor according to claim 3, characterized in that, The dual-gradient augmented regressor includes a first group of regressors related to time delay and a second group of regressors related to area. The process of predicting performance parameters for each candidate grid point using the dual-gradient augmented regressor to obtain the predicted performance parameters for each candidate grid point includes: The first group of regressors is used to predict the time delay of each candidate grid point, thereby obtaining the predicted time delay of each candidate grid point. The area of ​​each candidate grid point is predicted by the second group of regressors. The performance prediction parameters include the prediction latency and the prediction area.

5. The grid traversal method based on a dual-gradient enhancement regressor according to claim 4, characterized in that, If a target candidate grid point is found to have a performance prediction parameter that is better than or partially better than any of the performance parameters, then the actual performance parameters of each target candidate grid point are obtained, including: If a target candidate grid point is obtained whose predicted latency is lower than that of any of the reference grid points, the actual latency of each target candidate grid point is obtained through the high-level synthesis tool. If a target candidate grid point is obtained whose predicted area is smaller than the area of ​​any of the reference grid points, then the actual area of ​​each target candidate grid point is obtained through the high-level synthesis tool. If a target candidate grid point is obtained where the predicted latency is lower than the latency of any of the reference grid points and the predicted area is smaller than the area of ​​any of the reference grid points, then the actual latency and actual area of ​​each target candidate grid point are obtained through the high-level synthesis tool.

6. The grid traversal method based on a dual-gradient enhancement regressor according to claim 5, characterized in that, If a target grid point is found whose actual performance parameter is better than or partially better than any of the performance parameters, then updating one or more of the reference grid points based on the target grid point includes: If a target grid point is found whose actual delay is lower than that of any of the reference grid points, then a target reference grid point whose delay is higher than the actual delay is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point. If a target grid point with an actual area smaller than the area of ​​any of the reference grid points is found, then a target reference grid point with an area greater than the actual time delay is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point. If a target grid point is found whose actual delay and actual area are both less than the delay and area of ​​any of the reference grid points, then a target reference grid point with a delay higher than the actual delay and an area greater than the actual area is found, and each of the target reference grid points is marked as a failed grid point, and the target grid point is marked as a reference grid point.

7. The grid traversal method based on a dual-gradient enhancement regressor according to claim 6, characterized in that, After comprehensively comparing the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point, the method further includes: If the prediction delay of each candidate grid point is higher than the delay of each reference grid point and the prediction area of ​​each candidate grid point is greater than the area of ​​each reference grid point, then the candidate grid point is marked as a poor grid point.

8. The grid traversal method based on a dual-gradient enhancement regressor according to claim 7, characterized in that, After all spatial design grid points within the target area have been traversed, the method further includes: Obtain the abandoned grid points that were not predicted for performance parameters by the dual-gradient augmented regressor. The time delay and area of ​​each of the inferior grid points and the abandoned grid points are obtained through the high-level synthesis tool. If the delay of any of the inferior grid points or the abandoned grid points is lower than the delay of the reference grid point and / or the area of ​​any of the inferior grid points or the abandoned grid points is smaller than the area of ​​the reference grid point, then the inferior grid point or the abandoned grid point whose delay is lower than the delay of the reference grid point and / or whose area is smaller than the area of ​​the reference grid point is marked as the reference grid point.

9. A grid traversal device based on a dual-gradient enhanced regressor, characterized in that, The device includes: The acquisition unit is used to acquire a target space design set, which consists of several space design instruction sets; A generation unit is used to generate a target region based on the target spatial design set, wherein the target region includes spatial design grid points that correspond one-to-one with each of the spatial design instruction sets; The selection unit is used to select a sub-region to be traversed within the target area, and to perform initial traversal on each spatial design grid point within the sub-region to be traversed to obtain at least one reference grid point. The performance parameters of the reference grid point meet preset performance conditions, and the performance parameters include the time delay and area for completing the spatial design. The determining unit is used to determine the next sub-region to be traversed based on each of the reference grid points, and to traverse each spatial design grid point in the next sub-region to be traversed to obtain candidate grid points adjacent to each of the reference grid points. The prediction unit is used to predict the performance parameters of each candidate grid point through a dual gradient enhanced regressor, and obtain the performance prediction parameters of each candidate grid point. The first comprehensive comparison unit is used to comprehensively compare the performance prediction parameters of each candidate grid point with the performance parameters of each reference grid point. If a target candidate grid point is found to have a performance prediction parameter that is better than or partially better than any of the performance parameters, then the actual performance parameters of each target candidate grid point are obtained. The second comprehensive comparison unit is used to comprehensively compare the actual performance parameters of each of the target candidate grid points with the performance parameters of each of the reference grid points. If a target grid point is found whose actual performance parameters are better than or partially better than any of the performance parameters, then one or more of the reference grid points are updated based on the target grid point. An iterative unit is used to repeat the execution steps of the determining unit, the predicting unit, the first comprehensive comparison unit, and the second comprehensive comparison unit until all spatial design grid points in the target area have been traversed to find the reference grid point with the optimal performance parameters. The execution order of the determining unit, the predicting unit, the first comprehensive comparison unit, and the second comprehensive comparison unit is as follows: determining unit, predicting unit, first comprehensive comparison unit, and second comprehensive comparison unit.

10. An electronic device, characterized in that, The electronic device includes one or more processors and one or more memories, wherein at least one piece of program code is stored in the one or more memories, and the at least one piece of program code is loaded and executed by the one or more processors to perform the operation performed by the method as described in any one of claims 1 to 8.

Images