A heterogeneous cooperative intelligent simulation method and system for silicon carbide chips

Abstract

This invention provides a heterogeneous collaborative intelligent simulation method and system for silicon carbide (SiC) chips, belonging to the field of collaborative simulation technology. The electrical and thermal behaviors of SiC chips are abstracted into independent simulation tasks, and standard solver templates are configured to specify data dependencies. Solvers are assigned to different types of computing cores to form active solver sets. Two types of verification criteria are set: power consumption propagation error and junction temperature limit, forming a unified verification process. Multiple active solver sets are collected to generate a simulation state record set, which is then split to obtain subsets of each verification item. Coupling efficiency is evaluated by recording latency and invalid deployment schemes are eliminated. The optimal solver deployment scheme is selected based on the verification coverage. This invention can adaptively schedule heterogeneous computing resources and dynamically filter simulation deployment schemes, which helps improve the execution efficiency and result reliability of electrothermal coupling simulation of SiC chips, and reduces the cost of manual parameter tuning.

Description

Technical Field

[0001] This invention relates to the field of collaborative simulation technology, and more specifically, to a heterogeneous collaborative intelligent simulation method and system for silicon carbide chips. Background Technology

[0002] Currently, common electro-thermal coupling simulation methods mainly employ the finite element method or finite volume method, sequentially calling the electrical and thermal solvers. That is, electrical calculations are performed first to obtain the power consumption distribution, and then this distribution is used as a heat source input to the thermal solver. However, this approach suffers from the following prominent problems:

[0003] In traditional serial simulation, electrical solvers and thermal solvers usually run on the same computing platform (such as a single CPU), which cannot take advantage of the parallelism of heterogeneous computing environments (CPU+GPU+FPGA). At the same time, the data transfer and synchronization between electrical and thermal calculations are entirely controlled manually, lacking dynamic scheduling capabilities, resulting in a large number of computing cores being idle.

[0004] In practical engineering, due to the different affinity of electrical and thermal tasks for the computing core, existing simulation software usually adopts a fixed solver-core binding strategy. Once started, it cannot be adjusted according to runtime load or convergence. When a solver is delayed due to resource contention, the entire simulation process is blocked, but the system cannot automatically identify and eliminate inefficient deployment schemes.

[0005] Most existing simulations only check whether the residuals of a single physical field converge, lacking a dedicated check for the consistency of electrothermal coupling; for example, when the power consumption value output by the electrical simulation is mapped to the thermal simulation, it may produce a non-physical offset due to mesh interpolation error, resulting in distortion of the temperature field calculation.

[0006] In heterogeneous computing environments, multiple solver allocation strategies can be used for the same simulation task. However, in current technologies, only one fixed configuration is run for each simulation, making it impossible to automatically select the optimal solution. Engineers often rely on experience to make repeated manual adjustments, resulting in long development cycles and unstable results. Summary of the Invention

[0007] The purpose of this invention is to provide a heterogeneous collaborative intelligent simulation method and system for silicon carbide chips, in order to solve the technical problems in existing silicon carbide chip electro-thermal coupling simulation, such as low utilization of heterogeneous computing resources, rigid solver deployment schemes that cannot be automatically selected, lack of electro-thermal coupling consistency verification criteria, and verification rules that cannot be adaptively adjusted according to simulation results.

[0008] In view of the above-mentioned technical problems, the present invention provides a heterogeneous collaborative intelligent simulation method and system for silicon carbide chips.

[0009] In a first aspect, the present invention provides a heterogeneous collaborative intelligent simulation method for silicon carbide chips, the method comprising:

[0010] Step S1: Construct a physics task layer, abstract the electrical and thermal behaviors of the silicon carbide chip during operation into independent simulation tasks, generate electrical simulation tasks and thermal simulation tasks, and specify the data dependency between electrical simulation tasks and thermal simulation tasks by configuring the standard solver template before simulation starts, wherein the output of the electrical simulation task is used as the input of the thermal simulation task.

[0011] Step S2: Construct a heterogeneous scheduling layer, set up a distribution interface for simulation tasks and a reusable solver library, and distribute electrical simulation tasks and thermal simulation tasks to different types of computing cores in the heterogeneous collaborative platform for execution through the distribution interface, and convert the standard solver template before simulation starts into a set of active solvers during runtime.

[0012] Step S3: Construct a result verification layer, set verification criteria based on the electrothermal coupling constraints of silicon carbide chips, and form a unified verification process for heterogeneous collaborative platforms by each verification criterion;

[0013] Step S4: Construct a simulation coordinator, collect a set of several active solvers dynamically generated from a standard solver template, form a simulation state record set, and obtain a subset of verification items corresponding to each verification criterion during the execution of a unified verification process on the heterogeneous collaborative platform.

[0014] Step S5: Construct a solver fit evaluation model, remove invalid records from the simulation state record set, and after the removal is completed, construct a self-adjusting model of the verification process to evaluate the verification coverage of the unified verification process, and determine the final set of active solvers to be adopted based on the verification coverage.

[0015] Preferably, the specific steps for constructing the physics task layer include:

[0016] The electrical and thermal behaviors of silicon carbide chips during operation are abstracted into independent simulation tasks, and task attributes are marked to obtain electrical simulation tasks and thermal simulation tasks. The electrical simulation task is used to generate the electric field distribution inside the silicon carbide chip, and the thermal simulation task is used to generate the temperature field distribution inside the silicon carbide chip.

[0017] Based on the electrothermal coupling constraint requirements of silicon carbide chips, a standard solver template is configured before the simulation starts. The standard solver template records the calling order of the electrical solver and the thermal solver, as well as the mapping relationship between the output results of the electrical simulation task and the input of the thermal simulation task. The simulation task is received and executed through the simulation task distribution interface of the standard solver template.

[0018] Preferably, the specific steps for constructing a heterogeneous scheduling layer include:

[0019] The electrical and thermal solvers are assigned to available computing cores in the heterogeneous collaborative platform, and the simulation tasks are sent to the available computing cores in the heterogeneous collaborative platform through the distribution interface according to the calling order, and the running results of the computing cores are received.

[0020] A reusable solver library is built into the heterogeneous collaborative platform. The reusable solver library stores the executable code of electrical and thermal solvers. When a standard solver template is used for simulation, the executable code is obtained from the reusable solver library through the distribution interface. The standard solver template is then loaded onto the corresponding computing core for execution, forming an active solver set. Each element in the active solver set is a solver process, and each solver process records the identifier of the assigned computing core.

[0021] Preferably, the specific steps for constructing the result verification layer include:

[0022] Based on the electrothermal coupling constraint requirements of silicon carbide chips, verification criteria are set, including:

[0023] First verification criterion: According to the mapping relationship, the relative error between the power consumption value output by the electrical simulation task and the corresponding power consumption value input to the thermal simulation task does not exceed the data error threshold.

[0024] Second verification criterion: The peak value of the temperature field output by the thermal simulation task shall not exceed the junction temperature limit of the silicon carbide chip;

[0025] A unified verification process is established, which includes verification criteria. Each verification criterion corresponds to a verification item. A matching label is set for each verification item. The matching label has an associated mapping relationship with the attribute label of the simulation task. The matching label identifies the verification item to which the output result of each solver process in the active solver set belongs. The matching label is associated with the computational core identifier of the solver process and is indirectly mapped to the attribute label of the simulation task.

[0026] Preferably, the specific steps for constructing the simulation coordinator include:

[0027] During the execution of the unified verification process, a set of several active solvers dynamically generated corresponding to a standard solver template is collected to form a simulation state record set. The simulation state record set records an allocation relationship table associated with each active solver set. The allocation relationship table records each solver process in its associated active solver set and the computational core identifier assigned to each solver process.

[0028] Once the matching label of the verification item matches the attribute tags carried by the output results of each solver process in the active solver set, a subset of verification items is generated for each assignment table. The subset of verification items records the set of all verification items belonging to the same active solver set.

[0029] Preferably, the specific methods for constructing a solver fit evaluation model and removing invalid records from the simulation state record set include:

[0030] When a set of active solvers is dynamically generated, its corresponding allocation table is retrieved, and the electrical simulation task and the thermal simulation task are started in sequence according to the calling order of the electrical solver and the thermal solver recorded in the standard solver template. The delay time between the completion time of generating the temperature field distribution in the thermal simulation task and the completion time of generating the electric field distribution in the electrical simulation task is recorded.

[0031] The coupling efficiency of a particular set of active solvers is evaluated based on the delay time, and the coupling efficiency is negatively correlated with the delay time.

[0032] A preset efficiency threshold is set. If the coupling efficiency is less than or equal to the efficiency threshold, the scheme of allocating the solvers of a certain active solver set to the computational core is determined to be invalid, and the certain active solver set is marked as invalid and removed from the simulation state record set. If the coupling efficiency is greater than the efficiency threshold, the certain active solver set is retained in the simulation state record set.

[0033] Preferably, the specific methods for constructing a self-adjusting model for the verification process and determining the final set of active solvers to be adopted include:

[0034] Under the simulation state record set after the elimination or retention operation, evaluate the verification coverage of the unified verification process. The verification coverage reflects the proportion of the number of verification items commonly included among the subsets of verification items to the total number of verification items.

[0035] A preset coverage threshold is set. If the verification coverage is less than or equal to the coverage threshold, the simulation engineer is prompted to supplement or adjust the verification criteria and re-evaluate the verification coverage until the verification coverage is greater than the coverage threshold. Then, the set of active solvers with the largest number of verification items in the subset of verification items is selected as the optimal solver deployment scheme for the simulation runtime.

[0036] Secondly, the present invention also provides a heterogeneous collaborative intelligent simulation system for silicon carbide chips, comprising:

[0037] A computer-readable storage medium for carrying a computer program that can be executed by a processor to implement the steps of the heterogeneous collaborative intelligent simulation method for silicon carbide chips described above.

[0038] One or more processors;

[0039] A heterogeneous collaborative platform is used to schedule one or more computer programs. When the one or more computer programs are executed by the one or more processors, the heterogeneous collaborative platform enables the implementation of the steps of the heterogeneous collaborative intelligent simulation method for silicon carbide chips described above.

[0040] Preferably, the heterogeneous collaborative platform further includes:

[0041] The physics task layer construction module is used to abstract the electrical and thermal behaviors of silicon carbide chips during operation into independent simulation tasks, generate electrical simulation tasks and thermal simulation tasks, and specify the data dependency relationship between electrical simulation tasks and thermal simulation tasks by configuring the standard solver template before simulation starts.

[0042] The heterogeneous scheduling layer construction module is used to set up the distribution interface for simulation tasks and the reusable solver library. Through the distribution interface, electrical simulation tasks and thermal simulation tasks are assigned to different types of computing cores in the heterogeneous collaborative platform for execution, and the standard solver template before the simulation starts is transformed into a set of active solvers during runtime.

[0043] The result verification layer construction module sets verification criteria based on the electrothermal coupling constraint conditions of silicon carbide chips, and the verification criteria constitute a unified verification process for heterogeneous collaborative platforms.

[0044] The simulation coordinator construction module is used to collect several sets of active solvers dynamically generated from a standard solver template, forming a simulation state record set, and obtaining a subset of verification items corresponding to each verification criterion during the execution of a unified verification process on the heterogeneous collaborative platform.

[0045] The solver fit evaluation model building module is used to remove invalid records from the simulation state record set. After the removal is completed, a self-adjusting model of the verification process is built to evaluate the verification coverage of the unified verification process, and the final set of active solvers to be adopted is determined based on the verification coverage.

[0046] One or more technical solutions provided in this invention have at least the following technical effects or advantages:

[0047] 1. This invention constructs a heterogeneous scheduling layer, which allocates the electrical and thermal solvers in the standard solver template to different types of computing cores according to a preset calling order, and records the computing core identifier of each solver process to form an "active solver set". This can dynamically generate multiple active solver sets (i.e., multiple deployment schemes) within a single simulation cycle, enabling the simulation system to explore different resource allocation strategies in parallel without the need for manual restarting and stopping of the simulation, thus helping to improve the utilization efficiency of heterogeneous computing resources.

[0048] 2. In the solver fit evaluation model, this invention records the delay between the completion time of the thermal simulation task and the completion time of the electrical simulation task, and defines that coupling efficiency is negatively correlated with the delay. By setting a preset efficiency threshold, it automatically eliminates active solver sets that have excessive delays due to unreasonable scheduling, thereby transforming task synchronization overhead into a quantifiable efficiency index. This solves the problem that "whether the scheduling is efficient" cannot be automatically judged in heterogeneous simulation, and realizes fine-grained screening of individual deployment schemes, eliminating low-quality candidate schemes for subsequent verification coverage optimization.

[0049] 3. This invention establishes two physically specific verification criteria. The first criterion requires that the relative error of the power consumption value transmitted according to the mapping relationship does not exceed a preset threshold, and the second criterion requires that the peak value of the temperature field does not exceed the junction temperature limit. Each verification criterion corresponds to a verification item, and the solver process output results are assigned to the corresponding verification item through matching tags, ultimately forming a "verification item subset" of each active solver set. This invention fills the gap in consistency verification in electrothermal coupling simulation by introducing "power consumption transmission error" verification. Through the splitting of the "verification item subset," the performance of different deployment schemes under the same verification criteria can be quantitatively compared.

[0050] 4. After eliminating invalid deployments, this invention evaluates the verification coverage of the unified verification process (i.e., the proportion of verification items commonly included in each subset of verification items to the total number of verification items). When the coverage is below a threshold, it automatically prompts simulation engineers to supplement or adjust the verification criteria. When the coverage meets the standard, it selects the set of active solvers with the largest number of verification items in the subset of verification items as the optimal deployment scheme, thereby realizing intelligent iteration of "simulation-verification-adjustment-optimization". Compared with the existing technology that relies on repeated trial and error by humans, this invention can automatically identify the weak links of the verification criteria (such as prompting to supplement criteria when the coverage is insufficient) and ensure that the final selected scheme passes the most verification items while meeting the coupling efficiency, thereby maximizing the reliability and physical consistency of the simulation results.

[0051] The above description is merely an overview of the technical solution of the present invention. To better understand the technical means of the present invention and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of the present invention more apparent, specific embodiments of the present invention are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of the present invention, nor is it intended to limit the scope of the present invention. Other features of the present invention will become readily apparent from the following description. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0053] Figure 1 This is a schematic diagram illustrating the steps of a heterogeneous collaborative intelligent simulation method for silicon carbide chips according to the present invention.

[0054] Figure 2 This is a schematic diagram of the structure of a heterogeneous collaborative intelligent simulation system for silicon carbide chips according to the present invention. Detailed Implementation

[0055] This invention provides a heterogeneous collaborative intelligent simulation method and system for silicon carbide (SiC) chips, addressing the technical problems of low utilization of heterogeneous computing resources, rigid solver deployment schemes with no automatic selection, lack of electro-thermal coupling consistency verification criteria, and inability to adaptively adjust verification rules in existing SiC chip electro-thermal coupling simulations. By constructing a physics task layer, electrical and thermal behaviors are abstracted into independent tasks and configured with standard solver templates. By constructing a heterogeneous scheduling layer, solvers are dynamically allocated to different computing cores, forming multiple active solver sets. By constructing a result verification layer, two types of verification criteria—power consumption propagation error and junction temperature limit—are set and broken down into subsets of verification items. By constructing a solver suitability evaluation model, latency is recorded to assess coupling efficiency and invalid deployment schemes are eliminated. Finally, a verification process self-adjusting model is constructed to calculate verification coverage and iteratively optimize until the optimal solver deployment scheme is selected. This achieves the technical effects of fully utilizing heterogeneous computing resources, adaptively selecting the optimal deployment scheme, quantifying electro-thermal coupling consistency, automatically adjusting verification rules, and significantly improving simulation efficiency and result reliability.

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. It should be understood that the present invention is not limited to the exemplary embodiments described herein. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. It should also be noted that, for ease of description, only the parts related to the present invention are shown in the accompanying drawings, and not all of them.

[0057] Example 1, please refer to Figure 1 This paper provides a heterogeneous collaborative intelligent simulation method for silicon carbide chips, which includes:

[0058] Step S1: Construct a physics task layer, abstract the electrical and thermal behaviors of the silicon carbide chip during operation into independent simulation tasks, generate electrical simulation tasks and thermal simulation tasks, and specify the data dependency between electrical simulation tasks and thermal simulation tasks by configuring the standard solver template before simulation starts, wherein the output of the electrical simulation task is used as the input of the thermal simulation task.

[0059] For example, the electrical and thermal behaviors of the silicon carbide chip during operation are abstracted into independent simulation tasks, and the task attributes are marked to obtain electrical simulation tasks and thermal simulation tasks. The electrical simulation task is used to generate the electric field distribution inside the silicon carbide chip, and the thermal simulation task is used to generate the temperature field distribution inside the silicon carbide chip.

[0060] Based on the electrothermal coupling constraint requirements of silicon carbide chips, a standard solver template is configured before the simulation starts. The standard solver template records the calling order of the electrical solver and the thermal solver, as well as the mapping relationship between the output of the electrical simulation task and the input of the thermal simulation task. The simulation task is received and executed through the simulation task distribution interface of the standard solver template.

[0061] It's important to note that the core task of the physics task layer is to decompose the actual physical processes (electro-thermal coupling) of the silicon carbide chip into two simulation tasks that can be independently scheduled on heterogeneous computing platforms, but with clear data dependencies. Under high voltage and high current conditions, the electric field distribution of the silicon carbide chip generates Joule heating, and the increased temperature, in turn, affects the material mobility and breakdown voltage. To accurately simulate this process, the electric field must first be solved to obtain the power consumption distribution, and then the power consumption must be used as a heat source input to solve the temperature field. Therefore, by configuring a standard solver template, the calling order (electrical first, then thermal) of the two solvers (electrical then thermal) and the mapping relationship (how the power consumption value output by the electrical solver corresponds to the corresponding spatial location of the thermal input) are predefined. Furthermore, the standard solver template provides a simulation task distribution interface, facilitating subsequent scheduling by starting execution according to this specification. It is worth mentioning that the electrical simulation task is a numerical calculation task for the electric field and current distribution inside the silicon carbide chip, which can usually solve the Poisson equation and the carrier continuity equation, and output the power density distribution. The thermal simulation task is a numerical calculation task for the temperature field distribution inside the silicon carbide chip, which solves the heat conduction equation and requires the power consumption output by the electrical simulation to be used as the heat source term as input. The standard solver template is a predefined simulation configuration file that records the solver type, calling order, data transfer mapping rules and task distribution interface.

[0062] Step S2: Construct a heterogeneous scheduling layer, set up a distribution interface for simulation tasks and a reusable solver library, distribute electrical simulation tasks and thermal simulation tasks to different types of computing cores in the heterogeneous collaborative platform through the distribution interface, and convert the standard solver template before simulation starts into a set of active solvers during runtime.

[0063] For example, the electrical solver and the thermal solver are assigned to the available computing cores in the heterogeneous collaborative platform, and the simulation task is sent to the available computing cores in the heterogeneous collaborative platform through the distribution interface according to the calling order, and the running results of the computing cores are received.

[0064] The heterogeneous collaborative platform has a built-in reusable solver library, which stores the executable code of electrical and thermal solvers. When a standard solver template is used for simulation, the executable code is obtained from the reusable solver library through the distribution interface. The standard solver template is then loaded onto the corresponding computing core for execution and forms an active solver set. Each element in the active solver set is a solver process, and each solver process records the identifier of the assigned computing core.

[0065] It's important to note that different types of solvers have varying affinity for different computational cores: for example, electrical solvers involve numerous matrix operations and are better suited for parallel execution on GPUs; thermal solvers involve unstructured mesh iteration and are better suited for CPUs. The heterogeneous scheduling layer utilizes a distribution interface to allocate electrical and thermal solvers to currently available computational cores for execution, according to the calling order specified in the standard solver template. Each solver, after having its executable code loaded, forms a solver process, which records the computational core identifier (e.g., GPU0, CPU1). The set of all solver processes is called the active solver set; thus, multiple different active solver sets can be dynamically generated during simulation, resulting in various deployment schemes for subsequent evaluation.

[0066] Step S3: Construct a result verification layer, set verification criteria based on the electrothermal coupling constraints of silicon carbide chips, and form a unified verification process for heterogeneous collaborative platforms by each verification criterion;

[0067] For example, based on the electrothermal coupling constraint requirements of silicon carbide chips, verification criteria are set, including:

[0068] First verification criterion: According to the mapping relationship, the relative error between the power consumption value output by the electrical simulation task and the corresponding power consumption value input to the thermal simulation task does not exceed the data error threshold.

[0069] Second verification criterion: The peak value of the temperature field output by the thermal simulation task shall not exceed the junction temperature limit of the silicon carbide chip;

[0070] Establish a unified verification process that includes verification criteria. Each verification criterion corresponds to a verification item. Set a matching label for each verification item. The matching label has an associated mapping relationship with the attribute label of the simulation task. Identify the verification item to which the output result of each solver process in the active solver set belongs by using the matching label. The matching label is associated with the computational core identifier of the solver process and is indirectly mapped to the attribute label of the simulation task.

[0071] It should be noted that the purpose of setting the verification criterion of "relative error not exceeding the data error threshold" is to ensure that the deviation between the power distribution received by the thermal task and the power distribution actually calculated by the electrical task is within the acceptable range in engineering, thereby avoiding errors in temperature field calculation due to data transmission distortion. The data error threshold is a preset engineering parameter that can be set based on historical data from typical simulation cases of silicon carbide chips. For example, when the relative error exceeds 5%, the deviation between the subsequent temperature field and the measured infrared thermogram increases significantly. Therefore, the threshold is usually set to 5% or a more conservative 3%.

[0072] Step S4: Construct a simulation coordinator, collect a set of several active solvers dynamically generated from a standard solver template, form a simulation state record set, and obtain a subset of verification items corresponding to each verification criterion during the execution of a unified verification process on the heterogeneous collaborative platform.

[0073] For example, during the execution of the unified verification process, several sets of active solvers dynamically generated corresponding to a standard solver template are collected to form a simulation state record set. The simulation state record set records the allocation relationship table associated with each set of active solvers. The allocation relationship table records each solver process in its associated active solver set and the computational core identifier assigned to each solver process.

[0074] Once the matching label of the verification item matches the attribute tags carried by the output results of each solver process in the active solver set, a subset of verification items is generated for each assignment table. The subset of verification items records the set of all verification items belonging to the same active solver set.

[0075] Step S5: Construct a solver fit evaluation model, remove invalid records from the simulation state record set, and after the removal is completed, construct a self-adjusting model of the verification process to evaluate the verification coverage of the unified verification process, and determine the final set of active solvers to be adopted based on the verification coverage.

[0076] For example, the specific methods for constructing a solver fit evaluation model and removing invalid records from the simulation state record set include:

[0077] When a set of active solvers is dynamically generated, its corresponding allocation table is retrieved, and the electrical simulation task and the thermal simulation task are started in sequence according to the calling order of the electrical solver and the thermal solver recorded in the standard solver template. The delay time between the completion time of generating the temperature field distribution in the thermal simulation task and the completion time of generating the electric field distribution in the electrical simulation task is recorded.

[0078] The coupling efficiency of a particular set of active solvers is evaluated based on the delay time, and the coupling efficiency is negatively correlated with the delay time.

[0079] A preset efficiency threshold is set. If the coupling efficiency is less than or equal to the efficiency threshold, the scheme of allocating the solvers of a certain active solver set to the computational core is determined to be invalid, and the certain active solver set is marked as invalid and removed from the simulation status record set. If the coupling efficiency is greater than the efficiency threshold, the certain active solver set is retained in the simulation status record set.

[0080] The specific methods for constructing a self-adjusting model for the verification process and determining the final set of active solvers to be adopted include:

[0081] Under the simulation state record set after the removal or retention operation, evaluate the verification coverage of the unified verification process. The verification coverage reflects the proportion of the number of verification items commonly included among the subsets of verification items to the total number of verification items.

[0082] A preset coverage threshold is set. If the verification coverage is less than or equal to the coverage threshold, the simulation engineer is prompted to supplement or adjust the verification criteria and re-evaluate the verification coverage until the verification coverage is greater than the coverage threshold. Then, the set of active solvers with the largest number of verification items in the subset of verification items is selected as the optimal solver deployment scheme for the simulation runtime.

[0083] It should be noted that coupling efficiency decreases with increasing delay time. It is used to measure the degree of synchronization between electrical and thermal tasks within an active solver set (i.e., a solver-core allocation scheme). This can be achieved through... To obtain it using a formula, To delay the duration, For the preset maximum delay time, an exponential function can more smoothly distinguish between medium and long delays; a linear formula can also be used. The efficiency threshold can be set to 0.8 times the average coupling efficiency of multiple active solver sets. When the coupling efficiency is less than or equal to the efficiency threshold, the scheme is deemed invalid and removed, indicating that the latency of the active solver set is too high and the resource scheduling is unreasonable. Considering the fault tolerance mechanism (allowing some verification items to fail temporarily, and relaxing the junction temperature limit by 20℃), the coverage threshold can be set to 0.8. If the verification coverage is less than or equal to the coverage threshold, it means that even among multiple deployment schemes with qualified scheduling efficiency, some verification items are still generally missing (e.g., all schemes fail the temperature peak verification). This indicates that the current verification criteria themselves may be improperly set (e.g., the junction temperature limit is too low, or the data error threshold is too strict). At this time, the system prompts the simulation engineer to supplement or adjust the verification criteria (e.g., relax the junction temperature limit based on measured data, or add new verification items such as thermal stress), and then rerun S3. Continue until the coverage rate reaches the target at S5. If the verification coverage rate is greater than the coverage threshold, it indicates that the current verification criteria are reasonable and the simulation conditions are sufficient. At this point, the optimal solution should be selected from the candidate solutions. The selection criteria are: under the condition that the pass rate is the same (i.e., all verification items in the intersection are satisfied), additionally examine the other verification items that each solution passes independently. The solution with the most verification items in the subset of verification items means that it has passed more additional checks on the basis of meeting the common basic requirements. Therefore, the overall credibility of the simulation results is the highest. This solution is used as the optimal solver deployment solution, which can ensure both qualified efficiency and the most reliable results.

[0084] Example 2, please refer to Figure 2 This paper provides a heterogeneous collaborative intelligent simulation system for silicon carbide chips, the system comprising:

[0085] A computer-readable storage medium for carrying a computer program that can be executed by a processor to implement the steps of a heterogeneous collaborative intelligent simulation method for silicon carbide chips according to Embodiment 1 above.

[0086] One or more processors;

[0087] A heterogeneous collaborative platform is used to schedule one or more computer programs. When one or more computer programs are executed by one or more processors, the heterogeneous collaborative platform enables the implementation of the steps of the heterogeneous collaborative intelligent simulation method for silicon carbide chips described in Embodiment 1 above.

[0088] The heterogeneous collaboration platform also includes:

[0089] The physics task layer construction module is used to abstract the electrical and thermal behaviors of silicon carbide chips during operation into independent simulation tasks, generate electrical simulation tasks and thermal simulation tasks, and specify the data dependency relationship between electrical simulation tasks and thermal simulation tasks by configuring the standard solver template before simulation starts.

[0090] The heterogeneous scheduling layer construction module is used to set up the distribution interface for simulation tasks and the reusable solver library. Through the distribution interface, electrical simulation tasks and thermal simulation tasks are assigned to different types of computing cores in the heterogeneous collaborative platform for execution, and the standard solver template before the simulation starts is transformed into a set of active solvers at runtime.

[0091] The result verification layer construction module sets verification criteria based on the electrothermal coupling constraint conditions of silicon carbide chips, and the verification criteria constitute a unified verification process for heterogeneous collaborative platforms.

[0092] The simulation coordinator construction module is used to collect several sets of active solvers dynamically generated from a standard solver template, forming a simulation state record set, and obtaining a subset of verification items corresponding to each verification criterion during the execution of a unified verification process on the heterogeneous collaborative platform.

[0093] The solver fit evaluation model building module is used to remove invalid records from the simulation state record set. After the removal is completed, a self-adjusting model of the verification process is built to evaluate the verification coverage of the unified verification process, and the final set of active solvers to be adopted is determined based on the verification coverage.

[0094] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. The heterogeneous collaborative intelligent simulation method and specific examples for silicon carbide chips described in Embodiment 1 above are also applicable to the heterogeneous collaborative intelligent simulation system for silicon carbide chips in this embodiment. Through the foregoing detailed description of the heterogeneous collaborative intelligent simulation method for silicon carbide chips, those skilled in the art can clearly understand the heterogeneous collaborative intelligent simulation system for silicon carbide chips in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here. As for the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple, and relevant parts can be referred to in the method section.

[0095] In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0096] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0097] In addition, in the various embodiments of the present invention, each functional unit can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0098] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory, random access memory, magnetic disks, or optical disks.

[0099] Alternatively, if the integrated units of this invention are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, RAM, magnetic disks, or optical disks.

[0100] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0101] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. If such modifications and variations fall within the scope of this invention and its equivalents, then this invention is also intended to include such modifications and variations.

Claims

1. A heterogeneous collaborative intelligent simulation method for silicon carbide chips, characterized in that, The method includes the following steps: Step S1: Construct a physics task layer, abstract the electrical and thermal behaviors of the silicon carbide chip during operation into independent simulation tasks, generate electrical simulation tasks and thermal simulation tasks, and specify the data dependency between electrical simulation tasks and thermal simulation tasks by configuring the standard solver template before simulation starts, wherein the output of the electrical simulation task is used as the input of the thermal simulation task. Step S2: Construct a heterogeneous scheduling layer, set up a distribution interface for simulation tasks and a reusable solver library, and distribute electrical simulation tasks and thermal simulation tasks to different types of computing cores in the heterogeneous collaborative platform for execution through the distribution interface, and convert the standard solver template before simulation starts into a set of active solvers during runtime. Step S3: Construct a result verification layer, set verification criteria based on the electrothermal coupling constraints of silicon carbide chips, and form a unified verification process for heterogeneous collaborative platforms by each verification criterion; Step S4: Construct a simulation coordinator, collect a set of several active solvers dynamically generated from a standard solver template, form a simulation state record set, and obtain a subset of verification items corresponding to each verification criterion during the execution of a unified verification process on the heterogeneous collaborative platform. Step S5: Construct a solver fit evaluation model, remove invalid records from the simulation state record set, and after the removal is completed, construct a self-adjusting model of the verification process to evaluate the verification coverage of the unified verification process, and determine the final set of active solvers to be adopted based on the verification coverage.

2. The heterogeneous collaborative intelligent simulation method for silicon carbide chips according to claim 1, characterized in that, The specific steps for constructing the physics task layer include: The electrical and thermal behaviors of silicon carbide chips during operation are abstracted into independent simulation tasks, and task attributes are marked to obtain electrical simulation tasks and thermal simulation tasks. The electrical simulation task is used to generate the electric field distribution inside the silicon carbide chip, and the thermal simulation task is used to generate the temperature field distribution inside the silicon carbide chip. Based on the electrothermal coupling constraint requirements of silicon carbide chips, a standard solver template is configured before the simulation starts. The standard solver template records the calling order of the electrical solver and the thermal solver, as well as the mapping relationship between the output results of the electrical simulation task and the input of the thermal simulation task. The simulation task is received and executed through the simulation task distribution interface of the standard solver template.

3. The heterogeneous collaborative intelligent simulation method for silicon carbide chips according to claim 2, characterized in that, The specific steps for constructing a heterogeneous scheduling layer include: The electrical and thermal solvers are assigned to available computing cores in the heterogeneous collaborative platform, and the simulation tasks are sent to the available computing cores in the heterogeneous collaborative platform through the distribution interface according to the calling order, and the running results of the computing cores are received. A reusable solver library is built into the heterogeneous collaborative platform. The reusable solver library stores the executable code of electrical and thermal solvers. When a standard solver template is used for simulation, the executable code is obtained from the reusable solver library through the distribution interface. The standard solver template is then loaded onto the corresponding computing core for execution, forming an active solver set. Each element in the active solver set is a solver process, and each solver process records the identifier of the assigned computing core.

4. The heterogeneous collaborative intelligent simulation method for silicon carbide chips according to claim 3, characterized in that, The specific steps for constructing the result validation layer include: Based on the electrothermal coupling constraint requirements of silicon carbide chips, verification criteria are set, including: First verification criterion: According to the mapping relationship, the relative error between the power consumption value output by the electrical simulation task and the corresponding power consumption value input to the thermal simulation task does not exceed the data error threshold. Second verification criterion: The peak value of the temperature field output by the thermal simulation task shall not exceed the junction temperature limit of the silicon carbide chip; A unified verification process is established, which includes verification criteria. Each verification criterion corresponds to a verification item. A matching label is set for each verification item. The matching label has an associated mapping relationship with the attribute label of the simulation task. The matching label identifies the verification item to which the output result of each solver process in the active solver set belongs. The matching label is associated with the computational core identifier of the solver process and is indirectly mapped to the attribute label of the simulation task.

5. The heterogeneous collaborative intelligent simulation method for silicon carbide chips according to claim 4, characterized in that, The specific steps for building a simulation coordinator include: During the execution of the unified verification process, a set of several active solvers dynamically generated corresponding to a standard solver template is collected to form a simulation state record set. The simulation state record set records an allocation relationship table associated with each active solver set. The allocation relationship table records each solver process in its associated active solver set and the computational core identifier assigned to each solver process. Once the matching label of the verification item matches the attribute tags carried by the output results of each solver process in the active solver set, a subset of verification items is generated for each assignment table. The subset of verification items records the set of all verification items belonging to the same active solver set.

6. The heterogeneous collaborative intelligent simulation method for silicon carbide chips according to claim 5, characterized in that, The specific methods for constructing a solver fit evaluation model and removing invalid records from the simulation state record set include: When a set of active solvers is dynamically generated, its corresponding allocation table is retrieved, and the electrical simulation task and the thermal simulation task are started in sequence according to the calling order of the electrical solver and the thermal solver recorded in the standard solver template. The delay time between the completion time of generating the temperature field distribution in the thermal simulation task and the completion time of generating the electric field distribution in the electrical simulation task is recorded. The coupling efficiency of a particular set of active solvers is evaluated based on the delay time, and the coupling efficiency is negatively correlated with the delay time. A preset efficiency threshold is set. If the coupling efficiency is less than or equal to the efficiency threshold, the scheme of allocating the solvers of a certain active solver set to the computational core is determined to be invalid, and the certain active solver set is marked as invalid and removed from the simulation state record set. If the coupling efficiency is greater than the efficiency threshold, the certain active solver set is retained in the simulation state record set.

7. The heterogeneous collaborative intelligent simulation method for silicon carbide chips according to claim 6, characterized in that, The specific methods for constructing a self-adjusting model for the verification process and determining the final set of active solvers to be adopted include: Under the simulation state record set after the elimination or retention operation, evaluate the verification coverage of the unified verification process. The verification coverage reflects the proportion of the number of verification items commonly included among the subsets of verification items to the total number of verification items. A preset coverage threshold is set. If the verification coverage is less than or equal to the coverage threshold, the simulation engineer is prompted to supplement or adjust the verification criteria and re-evaluate the verification coverage until the verification coverage is greater than the coverage threshold. Then, the set of active solvers with the largest number of verification items in the subset of verification items is selected as the optimal solver deployment scheme for the simulation runtime.

8. A heterogeneous collaborative intelligent simulation system for silicon carbide chips, characterized in that, The invention includes a computer-readable storage medium for carrying a computer program that can be executed by a processor to implement the steps of a heterogeneous collaborative intelligent simulation method for silicon carbide chips as described in any one of claims 1 to 7.

9. A heterogeneous collaborative intelligent simulation system for silicon carbide chips according to claim 8, characterized in that, Also includes: One or more processors; A heterogeneous collaborative platform is used to schedule one or more computer programs, which, when executed by one or more processors, enable the heterogeneous collaborative platform to implement the steps of the heterogeneous collaborative intelligent simulation method for silicon carbide chips.

10. A heterogeneous collaborative intelligent simulation system for silicon carbide chips according to claim 9, characterized in that, The heterogeneous collaboration platform also includes: The physics task layer construction module is used to abstract the electrical and thermal behaviors of silicon carbide chips during operation into independent simulation tasks, generate electrical simulation tasks and thermal simulation tasks, and specify the data dependency relationship between electrical simulation tasks and thermal simulation tasks by configuring the standard solver template before simulation starts. The heterogeneous scheduling layer construction module is used to set up the distribution interface for simulation tasks and the reusable solver library. Through the distribution interface, electrical simulation tasks and thermal simulation tasks are assigned to different types of computing cores in the heterogeneous collaborative platform for execution, and the standard solver template before the simulation starts is transformed into a set of active solvers during runtime. The result verification layer construction module sets verification criteria based on the electrothermal coupling constraint conditions of silicon carbide chips, and the verification criteria constitute a unified verification process for heterogeneous collaborative platforms. The simulation coordinator construction module is used to collect several sets of active solvers dynamically generated from a standard solver template, forming a simulation state record set, and obtaining a subset of verification items corresponding to each verification criterion during the execution of a unified verification process on the heterogeneous collaborative platform. The solver fit evaluation model building module is used to remove invalid records from the simulation state record set. After the removal is completed, a self-adjusting model of the verification process is built to evaluate the verification coverage of the unified verification process, and the final set of active solvers to be adopted is determined based on the verification coverage.

Images