Dynamic parallel balancing method, device and medium for large-scale grid pre-processing

By uniformly distributing and weighting the number of interpolation cells in the stitched mesh before CFD dynamic calculation, the problem of uneven search time between kernels in parallel computation of stitched mesh is solved, thus improving the efficiency of parallel computation.

CN118761359BActive Publication Date: 2024-11-22CALCULATION AERODYNAMICS INST CHINA AERODYNAMICS RES & DEV CENT
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411246212.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-11-22
Estimated Expiration
2044-09-06

AI Technical Summary

Technical Problem

In CFD dynamic computing, when performing parallel computation of stitched meshes, the uneven number of stitched elements to be searched on each computational core leads to large differences in search time, which seriously affects parallel efficiency.

Method used

By uniformly distributing the number of interpolation units among the cores before static computation, recording the search time and performing a weighted average, establishing a transfer list, and dynamically adjusting the allocation of interpolation units, the search time of each core is ensured to be consistent.

Benefits of technology

The parallel efficiency of the spliced ​​mesh dynamic simulation has been optimized, reducing the phenomenon of individual cores dragging down the overall calculation speed and improving the uniformity of calculation speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118761359B_ABST
    Figure CN118761359B_ABST
Patent Text Reader

Abstract

The present invention provides a large-scale splicing grid pre-processing dynamic parallel balancing method, device and medium. Based on the splicing grid algorithm and parallel solution technology widely used in CFD, in the iterative process of dynamically solving the flow field, no manual intervention of the user is required, and the splicing grid pre-search process of each computing core is automatically globally dynamically balanced. By updating the transmission list, the real-time communication of each core is realized and the number of interpolation boundary units is allocated to ensure that the search time of different computing cores is as similar as possible; and the search efficiency of the splicing grid pre-processing in parallel calculation is improved, and the solution time is shortened.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of fluid simulation technology, and in particular to a large-scale splicing grid pre-processing dynamic parallel balancing method, equipment and medium. Background Art

[0002] At present, the application of CFD (Computational Fluid Dynamics) numerical simulation in aerospace is increasing, and its role is gradually increasing. Due to the characteristics of short cycle and low cost, the industrial sector has increasingly prominent requirements for the development of different CFD simulation technologies. The mainstream CFD numerical calculation is divided into two categories: structured grid method and unstructured grid method. The unstructured grid method has strong adaptability to complex shapes, but the calculation efficiency is low; and the structured grid method has always occupied an important position in the field of CFD due to its good efficiency and high precision. However, the traditional structured grid method has poor adaptability to complex shapes. For this reason, the structured grid has successively developed overlapping grid and splicing grid technology. Among them, the splicing grid technology has an irreplaceable role in CFD because of its ability to ensure the conservation of interface flow field flux.

[0003] The splicing grid technology can greatly reduce the difficulty of grid generation by cutting off the continuous topology of the structural space grid, thereby greatly improving the adaptability of the structural grid to complex shapes. It has been widely used in CFD complex shape simulation and dynamic simulation, including simulation of structural deformation, rotating machinery, projectile discharge and other motion processes. Unlike the traditional docking grid technology, the splicing grid technology requires pre-processing before flow field simulation to obtain the interpolation relationship of its splicing boundary interpolation units. Its main time-consuming operation is to determine which corresponding units a splicing surface unit has a splicing relationship with, and to find the interpolation coefficient between the splicing unit and the interpolation unit with a splicing relationship. At present, a variety of relatively mature serial algorithms have been developed for splicing interpolation search, and have achieved relatively good application results.

[0004] With the rapid improvement of computer capabilities, the scale of CFD calculations has developed from tens of thousands of grid cells in the early days to tens of millions of grid cells in the mainstream in recent years, and the calculation mode has also completely changed from serial calculation to parallel calculation. However, the current parallel algorithm research is mainly focused on the parallel calculation of flow fields. There are not many mature parallel calculation methods for pre-processing processes such as splicing grid interpolation relationship search, and most of them are not well adapted to large-scale calculations. When the number of parallel cores exceeds a thousand cores, the unevenness of the parallel calculation amount of different computing cores will cause the search process of individual computing cores to be very slow, and the search time of different cores may differ by more than ten times. For example, if there are no cells to be spliced ​​on some computing cores, the computing cores will be completely vacant; some computing cores are distributed with many cells to be interpolated, resulting in a long search time; and the parallel calculation time depends on the time of the slowest computing core, so this unevenness of each computing core will seriously affect the global parallel efficiency. Especially when performing dynamic simulations, since each time step in the dynamic process must be pre-processed by splicing the grid, and it often takes hundreds of time steps to complete the simulation of a dynamic process, the problem of large differences in efficiency among cores due to uneven parallel search speeds will be more prominent in a large number of search processes. It is necessary to develop an efficient and highly uniform parallel method to curb the pathological time distribution of different computing cores. Summary of the invention

[0005] In view of this, the present invention provides a large-scale splicing grid pre-processing dynamic parallel balancing method, device and medium to solve the above problems.

[0006] In order to solve the above technical problems, the present invention provides a large-scale splicing grid pre-processing dynamic parallel balancing method, comprising:

[0007] S1: Parallel slicing of the original large-scale computing grid, evenly distributing all its spatial units to all computing cores;

[0008] S2: Perform constant static calculation on the original large-scale computing grid and establish a transmission list; evenly distribute the computing cores according to the total number of spliced ​​interpolation units, so that the number of interpolation units to be searched by different computing cores is consistent;

[0009] S3: Perform a global search and record the search time of each grid cell; use the total search time of all interpolation cells for weighted average;

[0010] S4: Automatically update the displaced object surface and slice it according to the original parallel slicing scheme, and perform unsteady dynamic iteration on the original large-scale computational grid;

[0011] S5: according to the transmission list, send and receive the splicing units for all computing cores;

[0012] S6: In subsequent iteration steps, the above S4 and S5 are repeated; wherein any calculation core always uses the weight coefficient obtained in the previous step as the allocation basis until all flow field iteration steps are completed.

[0013] As an optional method, in the above S3, after fully weighted averaging the total search time, an array is created to store the time weight coefficient of each interpolation unit.

[0014] As an optional method, in the above S4, when re-searching through unsteady dynamic iteration, the interpolation units allocated to any computing core are no longer allocated according to the number of splicing units, but are allocated according to the unit time weight after weighting the previous step behavior.

[0015] As an optional method, in the above S4, after the unsteady dynamic iteration is completed, the sending and receiving transmission lists of each computing core are obtained, including:

[0016] All computing cores are sorted according to the time weight coefficient, with the cores with high weight coefficients placed at the head of the table and the cores with low weight coefficients placed at the end of the table; after the sorting is completed, the transmission list is divided into three sections according to the preset threshold, and the transmission is allocated according to the rule of starting from the frontmost sending core to the rearmost receiving core; this allocation process is repeated until the transmission list is formed.

[0017] As an optional method, if the average time weight coefficient of the computing core exceeds a preset threshold during the process of the frontmost sending core sending to the rearmost receiving core in sequence, the sending core is sent to the next adjacent tail receiving core.

[0018] As an optional method, in the above S5, the information sent is the xyz coordinates of the four corner points of each splicing unit; after the sending is completed, the interpolation unit search is performed on the core splicing unit to solve the interpolation coefficients of different interpolation units.

[0019] As an optional manner, the receiving core is used to return the interpolation coefficients of the received interpolation unit to the sending core.

[0020] On the other hand, an electronic device is also provided, comprising: a memory for storing a computer program; and a processor for implementing the steps of the large-scale splicing grid pre-processing dynamic parallel equalization method as described above when executing the computer program.

[0021] On the other hand, a computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is executed by a processor, the steps of the large-scale splicing grid pre-processing dynamic parallel balancing method as described above are implemented.

[0022] The beneficial effects of the present invention are:

[0023] The present invention proposes a pre-processing parallel technology for structured grid splicing grid technology. During the CFD dynamic calculation process, this technology performs global unified management by weighting the search time of interpolation boundary units to obtain the send / receive transmission list of each core splicing interpolation unit; the cores communicate with each other in real time based on the transmission list, and dynamically allocate interpolation units between the computing cores by sending or receiving splicing units, and ensure that the search time of all computing cores is basically the same, thereby greatly reducing the phenomenon that individual inefficient computing cores drag down the global calculation speed, and providing an effective new algorithm for improving the parallel efficiency of splicing grid dynamic simulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic diagram of a grid splicing simulation provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of searching for different interpolation units provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of a parallel processing flow provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with specific implementation methods.

[0028] At present, CFD dynamic calculations often require the use of splicing grid technology to meet the needs of grid deformation, and the iterative process of dynamic calculations requires multi-time step calculations of the moving surface. At the same time, each time step of the surface movement requires pre-processing of the splicing grid to determine the interpolation relationship of the splicing boundary units to facilitate information exchange during flow field calculations. However, in the parallel process of splicing grid pre-processing, due to the different number of splicing interpolation units to be searched on each computing core, the search speed of each unit is also inconsistent, resulting in a large difference in the search time of each core and a serious reduction in parallel efficiency. The specific description of these two influencing factors is as follows:

[0029] 1) When parallel partitioning, it can only ensure that the number of space units allocated to each computing core is basically equal, but it cannot ensure that the number of stitching surface units on each core is basically the same. It is likely to cause uneven distribution, where some cores have no stitching units and some cores have very concentrated stitching units.

[0030] 2) For a certain splicing unit to be searched, if the area of ​​its interpolation unit is close to or greatly exceeds the area of ​​the splicing unit to be searched, the number of interpolation units is small, only 1~2, and the search speed is very fast; if the area of ​​its interpolation unit is much smaller than the area of ​​the splicing unit to be searched, the number of interpolation units increases significantly, possibly hundreds or thousands, and the search speed will be greatly slowed down. Figure 2 For the two sets of spliced ​​grids, the small grid has only one interpolation unit (the large grid unit that overlaps with its area), while the large grid has dozens of interpolation units (the small grid units that overlap with its area). Therefore, the search speeds of the interpolation units of the two sets of grids will be very different, and the search time of the large grid will greatly exceed the search time of the small grid.

[0031] Under the combined effect of the above two differences, the search speeds between different cores often differ by more than an order of magnitude. To address this problem, this embodiment proposes the following technical solutions:

[0032] Before the start of the splicing dynamic calculation, a constant static calculation must first be performed at the initial position to obtain the initial flow field for the dynamic calculation. Static calculation also requires pre-processing of the splicing unit search. In the static calculation, that is, the first splicing grid pre-processing process, for the inconsistent number of splicing boundary units on different computing cores, global allocation is performed, and each core is evenly allocated according to the total number of interpolation units. Before the search starts, the splicing unit information is sent from the core with more splicing units to the core with fewer splicing units to ensure that the number of interpolation units that need to be searched on different computing cores is basically the same. This approach can solve the first problem of the uneven distribution of the number of splicing units on each core caused by parallel block division; for the second problem, since the actual search time required for each unit is not the same, the search time of each unit is recorded after the search is completed, and the total search time of all splicing interpolation units is used for weighted average after the global search is completed. In the search of subsequent time steps, the number of units is no longer allocated according to the number of interpolation units, but according to the weighted interpolation unit time weight. This weight coefficient is dynamically adjusted as the position of the object plane changes during the dynamic calculation process, and the search time of each core is always kept basically the same.

[0033] In this regard, this embodiment proposes a real-time balancing method for parallel solution of spliced ​​grid dynamic process calculation based on the spliced ​​grid algorithm and parallel solution technology widely used in CFD, with the purpose of: in the iterative process of dynamically solving the flow field, without any manual intervention by the user, automatically performing global dynamic balancing on the spliced ​​grid pre-search process of each computing core, and by updating the transmission list, achieving real-time communication between each core and allocating the number of interpolation boundary units to ensure that the search time of different computing cores is as similar as possible; improving the search efficiency of spliced ​​grid pre-processing in parallel calculations and shortening the solution time.

[0034] See also Figure 3 , this embodiment is implemented as follows:

[0035] S1: Complete the parallel slicing of the original large-scale computing grid and distribute all its spatial units to all computing cores as evenly as possible. In order to simplify the parallel slicing and ensure the uniformity of the number of spatial units, this process does not consider the distribution of the splicing units on each core;

[0036] S2: Perform steady-state calculations on the original grid. During the pre-processing of the splicing units before the flow field calculation, a transmission list is established to evenly distribute the number of splicing interpolation units to each core, ensuring that the number of interpolation units to be searched by different computing cores is basically the same.

[0037] S3: Record the search time of each grid unit at the same time. After the global search is completed, use the total search time of all interpolation units for weighted average and create an array to save the time weight coefficient of each unit. If the search time of a certain interpolation unit is twice the average search time of all global units, its weight coefficient is 2; if the search time of a certain spatial unit is 1 / 2 of the global average time, its weight coefficient is 0.5.

[0038] S4: When starting the dynamic calculation, the global calculation grid is automatically updated for the object surface after displacement, and the blocks are cut according to the original parallel block cutting scheme, so that the number of spatial grid units on each computing core at different positions is completely consistent with that in static calculation, the spatial position of each grid unit is relatively close before and after displacement, and the distance between each splicing unit and the corresponding interpolation unit is not too far, and the mutual interpolation coefficient will not change significantly. Then, the unsteady dynamic iteration is carried out. When the splicing grid unit is re-searched, the interpolation units are allocated to each computing core no longer according to the number of splicing units, but according to the unit time weight after weighting in the previous step, and the sum of the interpolation unit weight coefficients obtained on each computing core is guaranteed to be as equal as possible. For example, if the area of ​​the splicing unit to be searched on computing core A is similar to that of the interpolation unit, the search time is short, and the average weight is 0.5; while the area of ​​the splicing unit to be searched on computing core B is much larger than that of the interpolation unit, the search time is long, and the average weight is 2.5, then computing core B will send the interpolation unit to computing core A until the average weight of the remaining units on computing core B is 1 or the average weight of the accumulated units on computing core A is 1, and then no more units will be sent.

[0039] According to the above rules, each core is iterated to obtain the transmission list of sending and receiving. The specific method is: sort all computing cores according to time weight, put the computing core with the highest weight, that is, the slowest computing core, at the head of the table, and the computing core with the lowest weight, that is, the fastest computing core, at the end of the table. After the sorting is completed, the transmission list is divided into three sections: the head of the table is divided into sending cores with an average weight coefficient greater than 1.02, the middle part of the table is the regular core with an average weight coefficient less than 1.02 and greater than 0.98, and the end of the table is the receiving core with an average weight coefficient less than 0.98.

[0040] Then, according to the rule of sending units from the frontmost sending core to the rearmost receiving core in sequence, ensure that the average time weight of each core does not exceed 1.02; if it exceeds, send it to the next adjacent tail receiving core. Repeat this process to obtain the transmission list;

[0041] S5: According to the transmission list, all computing cores send and receive splicing units. The information sent is the xyz coordinates of the four corner points of each splicing unit. After the sending is completed, the interpolation unit search is carried out for the splicing unit of this core to solve the interpolation coefficients of different interpolation units. For the receiving core, the interpolation coefficients of the receiving interpolation unit need to be returned to the sending core;

[0042] S6: For all subsequent iterative time steps, steps 3 to 5 are repeated. Each core always uses the weight coefficient obtained in the previous step as the basis for allocation, and ensures that the average time weight coefficient on all cores is basically the same until all flow field iteration steps are completed. Since the time and displacement differences between each two time steps of the spliced ​​grid technology calculation motion process are very small, the relative displacement of the spliced ​​grid unit and its interpolation unit in two adjacent time steps is not large, and the interpolation relationship is also similar. Under this premise, it is reasonable and feasible to use the weight coefficient of the previous time step to replace the current time step. This method of allocating interpolation units by searching time weights instead of unit numbers has better uniformity because it avoids the inefficiency caused by some interpolation units with large area differences and slow search.

[0043] Based on the above scheme, the traditional method is to determine the number of boundary units of the spliced ​​grid on each core in the pre-process of splicing the grid according to the initial allocation result during parallel slicing, which is very random, resulting in serious uneven search time of different computing cores, greatly affecting the parallel efficiency; the method proposed in this embodiment establishes a transmission list between each computing core through global management, and uniformly allocates interpolation units to each computing core. At the same time, the search time of each unit in the previous iteration step in the dynamic process is introduced as a weight coefficient for weighting, which can ensure that the time for each core to complete the pre-search is basically equal, optimizes the uneven problem caused by the serious pathological distribution of parallel time on different computing cores, and greatly improves the parallel search efficiency.

[0044] On the other hand, this embodiment also provides an electronic device, which can be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.) or other types of mobile devices, or a stationary computing device such as a desktop computer or a PC. The electronic device can also be a mobile or stationary server.

[0045] As an optional method, it includes: a memory for storing a computer program; a processor for implementing a large-scale grid pre-processing dynamic parallel balancing method when executing the computer program, including:

[0046] S1: Parallel slicing of the original large-scale computing grid, evenly distributing all its spatial units to all computing cores;

[0047] S2: Perform constant static calculation on the original large-scale computing grid and establish a transmission list; evenly distribute the computing cores according to the total number of spliced ​​interpolation units, so that the number of interpolation units to be searched by different computing cores is consistent;

[0048] S3: Perform a global search and record the search time of each grid cell; use the total search time of all interpolation cells for weighted average;

[0049] S4: Automatically update the displaced object surface and slice it according to the original parallel slicing scheme, and perform unsteady dynamic iteration on the original large-scale computational grid;

[0050] S5: according to the transmission list, send and receive the splicing units for all computing cores;

[0051] S6: In subsequent iteration steps, repeat S4 and S5 above; wherein any computing core always uses the weight coefficient obtained in the previous step as the basis for allocation until all flow field iteration steps are completed. The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned large-scale splicing grid pre-processing dynamic parallel balancing method belong to the same concept. For details not described in detail in the technical scheme of the computing device, please refer to the description of the technical scheme of the above-mentioned large-scale splicing grid pre-processing dynamic parallel balancing method.

[0052] On the other hand, this embodiment further provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, a large-scale grid pre-processing dynamic parallel balancing method is implemented, including:

[0053] S1: Parallel slicing of the original large-scale computing grid, evenly distributing all its spatial units to all computing cores;

[0054] S2: Perform constant static calculation on the original large-scale computing grid and establish a transmission list; evenly distribute the computing cores according to the total number of spliced ​​interpolation units, so that the number of interpolation units to be searched by different computing cores is consistent;

[0055] S3: Perform a global search and record the search time of each grid cell; use the total search time of all interpolation cells for weighted average;

[0056] S4: Automatically update the displaced object surface and slice it according to the original parallel slicing scheme, and perform unsteady dynamic iteration on the original large-scale computational grid;

[0057] S5: according to the transmission list, send and receive the splicing units for all computing cores;

[0058] S6: In subsequent iteration steps, the above S4 and S5 are repeated; wherein any calculation core always uses the weight coefficient obtained in the previous step as the allocation basis until all flow field iteration steps are completed.

[0059] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the large-scale splicing grid pre-processing dynamic parallel balancing method described above belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the large-scale splicing grid pre-processing dynamic parallel balancing method described above.

[0060] An embodiment of the present application further provides a chip storing a computer program, which, when executed by the chip, implements the steps of the large-scale splicing grid pre-processing dynamic parallel balancing method.

[0061] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0062] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0063] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described action sequence, because according to the present application, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0064] The above are only preferred embodiments of the present invention. It should be noted that the above preferred embodiments should not be regarded as limiting the present invention, and the protection scope of the present invention should be based on the scope defined by the claims. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the spirit and scope of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention.

Claims

1. A dynamic parallel balancing method suitable for large-scale splicing grid pre-processing, characterized in that: include: S1: Parallel slicing of the original large-scale computing grid, evenly distributing all its spatial units to all computing cores; S2: Perform constant static calculation on the original large-scale computing grid and establish a transmission list; evenly distribute the computing cores according to the total number of spliced ​​interpolation units, so that the number of interpolation units to be searched by different computing cores is consistent; S3: Perform a global search and record the search time of each grid cell; use the total search time of all interpolation cells for weighted average; S4: Automatically update the displaced object surface and slice it according to the original parallel slicing scheme, and perform unsteady dynamic iteration on the original large-scale computational grid; S5: sending and receiving splicing units for all computing cores according to the transmission list; S6: In the subsequent iteration steps, the above S4 and S5 are repeated; wherein any calculation core always uses the weight coefficient obtained in the previous step as the allocation basis until all flow field iteration steps are completed; In the above S4, when re-searching through unsteady dynamic iteration, the interpolation units allocated to any computing core are no longer allocated according to the number of splicing units, but are allocated according to the unit time weight after the weighting of the previous step behavior; and after completing the unsteady dynamic iteration, the sending and receiving transmission list of each computing core is obtained, including: All computing cores are sorted according to the time weight coefficient, with the cores with high weight coefficients placed at the head of the table and the cores with low weight coefficients placed at the end of the table; after the sorting is completed, the transmission list is divided into three sections according to the preset threshold, and the transmission is allocated according to the rule of starting from the frontmost sending core to the rearmost receiving core; this allocation process is repeated until the transmission list is formed.

2. A dynamic parallel equalization method suitable for large-scale splicing grid pre-processing according to claim 1, characterized in that: In the above S3, after fully weighted averaging the total search time, an array is created to store the time weight coefficient of each interpolation unit.

3. A dynamic parallel balancing method suitable for large-scale splicing grid pre-processing according to claim 1, characterized in that: If the average time weight coefficient of the computing core exceeds a preset threshold value during the process of the frontmost sending core sending to the rearmost receiving core in sequence, the sending core is sent to the next adjacent tail receiving core.

4. A dynamic parallel equalization method suitable for large-scale splicing grid pre-processing according to claim 3, characterized in that: In the above S5, the information sent is the xyz coordinates of the four corner points of each splicing unit; after the sending is completed, the interpolation unit search is performed on the core splicing unit to solve the interpolation coefficients of different interpolation units.

5. A dynamic parallel balancing method suitable for large-scale splicing grid pre-processing according to claim 4, characterized in that: The receiving core is used to return the interpolation coefficients of the received interpolation unit to the sending core.

6. A dynamic parallel balancing method suitable for large-scale splicing grid pre-processing according to claim 2, characterized in that: The value determination basis of the weight coefficient is the average search time of each interpolation unit.

7. An electronic device, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of the dynamic parallel balancing method applicable to large-scale splicing grid pre-processing as claimed in any one of claims 1 to 6 when executing the computer program.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the dynamic parallel balancing method applicable to large-scale splicing grid pre-processing are implemented as claimed in any one of claims 1 to 6.