Domestic chip and CAE (Computer Aided Engineering) software collaborative optimization method and system combined with AI (Artificial Intelligence) technology

Abstract

The invention discloses a domestic chip and CAE software collaborative optimization method and system combined with an AI technology, and belongs to the technical field of collaborative optimization, and the method comprises the steps: obtaining, preprocessing and marking original data, and obtaining training data; the initialized AI model is trained, and a trained AI model is obtained; acquiring real-time state data and real-time load characteristics, inputting the real-time state data and the real-time load characteristics into the trained AI model to obtain an optimization decision, performing instruction translation on the optimization decision to obtain a specific control instruction, sending the specific control instruction to chip hardware, automatically adjusting hardware behaviors and / or operating an interface through software. And automatically adjusting software behaviors. According to the method, dynamic and systematic software and hardware collaborative optimization is realized through an AI-driven optimization closed loop, and meanwhile, relevant parameter indexes can be adjusted and optimized.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of collaborative optimization, and particularly relates to a method and system for collaborative optimization of a domestic chip combined with AI technology and CAE software. BACKGROUND

[0002] At present, when the domestic CAE software runs on the emerging domestic chip platform, it still faces many challenges, mainly in the aspects of adaptability and performance optimization. Specifically, first, the current software layer does not fully understand and utilize the architecture characteristics of the domestic chip, which may result in the software not being able to fully utilize the computing potential of the chip. Secondly, the traditional software optimization and adaptation process often relies on human experience, which is time-consuming and costly, and is difficult to dynamically adapt to complex and variable application loads and hardware environments.

[0003] In view of the above problems, although there are optimization schemes for specific hardware in the prior art, there is no systematic and intelligent software and hardware collaborative optimization scheme, and the current technology can only determine whether the software has failed, but cannot optimize the system. Therefore, how to provide an effective technical solution to solve the problem of no systematic and intelligent software and hardware collaborative optimization scheme in the prior art, and the system cannot be adjusted and optimized, has become a difficult problem to be solved in the prior art. SUMMARY

[0004] The purpose of the present application is to provide a method and system for collaborative optimization of a domestic chip combined with AI technology and CAE software, to solve the above problems in the prior art.

[0005] In order to achieve the above purpose, the present application adopts the following technical scheme: In a first aspect, the present application provides a method for collaborative optimization of a domestic chip combined with AI technology and CAE software, comprising: obtaining original data, the original data including historical performance data and / or benchmark test data of the CAE software running on the domestic chip platform, the benchmark test data including software and hardware states, software operation sequences and corresponding performance indicators; preprocessing the original data to obtain original data features, labeling part of the original data features according to a preset label value to obtain labeled part of the original data features, and taking the original data features and the labeled part of the original data features as training data; inputting the training data into an initialized AI model for training to obtain a trained AI model; obtaining real-time state data of the domestic chip platform and real-time load features of the CAE software; The real-time state data and the real-time load feature are input into the trained AI model to obtain an optimization decision, the optimization decision includes dynamically adjusting parameters of software runtime, adjusting chip resource configuration, and selecting or recommending to use a specific algorithm, the optimization decision is translated into an instruction to obtain a specific control instruction, and the specific control instruction is sent to the chip hardware to automatically adjust hardware behavior and / or through a software running interface to automatically adjust software behavior.

[0006] In a possible design, the original data is preprocessed to obtain original data features, including: The original data is subjected to data cleaning, and the data cleaning includes removing outliers and filling missing values to obtain cleaned original data. The cleaned original data is subjected to conversion processing to obtain converted original data. The converted original data is subjected to feature engineering processing to obtain original data features.

[0007] In a possible design, the preset label value includes a performance indicator and a performance bottleneck type; and the training data is input into the initialized AI model for training to obtain the trained AI model, including: The training data is input into the initialized AI model for training, the training data is used as an input value of the initialized AI model, and the predicted performance indicator and the performance bottleneck type are used as output values of the initialized AI model. A loss value is calculated based on a loss function, the predicted performance indicator, the performance bottleneck type, and the performance indicator and the performance bottleneck type in the preset label value, and parameters of the initialized AI model are updated based on the loss value. The above updating step is repeated until a preset iteration termination condition is reached to obtain the trained AI model.

[0008] In a possible design, the AI model is an AI-driven software and hardware interaction model, including a perception module, a prediction module, and a decision module, the perception module is configured to extract features from real-time state data and real-time load features to obtain real-time feature vectors, the prediction module is configured to predict the real-time feature vectors to obtain prediction results, and the decision module is configured to generate an optimization decision based on the prediction results.

[0009] In a possible design, after obtaining the trained AI model, a cross-validation method is further used to evaluate the trained AI model, and hyperparameters of the trained AI model are adjusted according to an evaluation result.

[0010] In a possible design, the optimization decision includes dynamically adjusting parameters of software runtime, adjusting chip resource configuration, and selecting or recommending to use a specific algorithm; after obtaining the optimization decision, the following steps are further included: Obtaining field expert experience, updating the pre-constructed knowledge base based on the trained AI model, benchmark test data and field expert experience, the knowledge base including optimization rules; The optimization rules, optimization decisions and trained AI model are sent to each node of the pre-constructed cloud edge end collaborative architecture through a cloud edge end collaborative mechanism to form a data optimization closed loop mechanism.

[0011] In a possible design, the cloud edge end collaborative architecture includes a cloud data platform, and the method further includes: Obtaining running data, optimization records and effect evaluation data within a preset time interval, and uploading the running data, optimization records and effect evaluation data to the cloud data platform, the running data being running data of the CAE software on the domestic chip platform, the optimization records being data recorded after the optimization decisions are executed, and the effect evaluation data being data for evaluating the CAE software after the optimization decisions are executed; Periodically training the trained AI model based on the running data, optimization records and effect evaluation data stored in the cloud data platform to obtain a periodically trained AI model; Using the periodically trained AI model again to process the real-time state data of the chip and the real-time load characteristics of the CAE software to obtain new optimization decisions.

[0012] In a second aspect, the present application provides a domestic chip and CAE software collaborative optimization system combining AI technology, which is used to implement the method in the first aspect, and includes: A first obtaining module is configured to obtain original data, the original data including historical performance data of the CAE software running on the domestic chip platform and / or benchmark test data, the benchmark test data including software and hardware states, software operation sequences and corresponding performance indicators; A data preprocessing module is configured to preprocess the original data to obtain original data features, label part of the original data features according to preset label values to obtain labeled part of the original data features, and use the original data features and the labeled part of the original data features as training data; A model training module is configured to input the training data into an initialized AI model for training to obtain a trained AI model; A second obtaining module is configured to obtain real-time state data of the domestic chip platform and real-time load characteristics of the CAE software; An optimization decision module is configured to input real-time state data and real-time load characteristics into the trained AI model to obtain an optimization decision, the optimization decision including dynamically adjusting parameters of a software runtime, adjusting chip resource configuration, and selecting or recommending use of a specific algorithm, performing instruction translation on the optimization decision to obtain specific control instructions, and sending the specific control instructions to chip hardware to automatically adjust hardware behavior and / or automatically adjust software behavior through a software runtime interface.

[0013] In a third aspect, the present application provides a computer device comprising a memory, a processor and a transceiver connected in sequence for communication, wherein the memory is configured to store a computer program, the transceiver is configured to transceive messages, and the processor is configured to read the computer program and execute the method for collaborative optimization of a domestic chip and CAE software in combination with AI technology according to any one of the preceding aspects.

[0014] In a fourth aspect, the present application provides a computer program product comprising instructions for causing a computer to execute the method for collaborative optimization of a domestic chip and CAE software in combination with AI technology according to any one of the preceding aspects when the instructions are run on the computer.

[0015] The beneficial effects of the present application are as follows: The present application discloses a method and system for collaborative optimization of a domestic chip and CAE software in combination with AI technology, comprising obtaining original data, the original data including historical performance data and / or benchmark test data of the CAE software running on the domestic chip platform; preprocessing the original data to obtain original data characteristics, labeling part of the original data characteristics according to a preset label value to obtain labeled part of the original data characteristics, and taking the original data characteristics and the labeled part of the original data characteristics as training data; inputting the training data into an initialized AI model for training to obtain a trained AI model; obtaining real-time state data of the domestic chip platform and real-time load characteristics of the CAE software; inputting the real-time state data and the real-time load characteristics into the trained AI model to obtain an optimization decision, the optimization decision including dynamically adjusting parameters of a software runtime, adjusting chip resource configuration, and selecting or recommending use of a specific algorithm, performing instruction translation on the optimization decision to obtain specific control instructions, and sending the specific control instructions to chip hardware to automatically adjust hardware behavior and / or automatically adjust software behavior through a software runtime interface. The present application realizes dynamic, systematic and intelligent software and hardware collaborative optimization through an AI-driven optimization closed-loop mechanism, can adjust and optimize the system, improves the competitiveness of domestic industrial software, and is convenient for application and promotion. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flowchart of the method for collaborative optimization of a domestic chip and CAE software in combination with AI technology according to the first aspect of the embodiments is provided. Figure 2 This is a block diagram of a domestically produced chip and CAE software collaborative optimization system that combines AI technology, as provided in the second aspect of the embodiment. Detailed Implementation

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in conjunction with the accompanying drawings and descriptions of the embodiments or the prior art. Obviously, the following description of the structure of the accompanying drawings is only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. It should be noted that the description of these embodiments is for the purpose of helping to understand the present invention, but does not constitute a limitation of the present invention.

[0018] It should be understood that although the terms first, second, etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another. For example, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit, without departing from the scope of the exemplary embodiments of the invention.

[0019] It should be understood that the term "and / or" that may appear in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist simultaneously. The term " / and" that may appear in this document describes another relationship between related objects, indicating that two relationships can exist. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this document generally indicates that the related objects before and after it are in an "or" relationship.

[0020] Example: like Figure 1 As shown, the first aspect of this embodiment provides a method for collaborative optimization of domestically produced chips and CAE software combining AI technology. This method can be executed, but is not limited to, by a computer device or virtual machine with certain computing resources, such as a personal computer or smartphone, or by a virtual machine. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology includes, but is not limited to, the following steps: S1. Obtain raw data, which includes historical performance data and / or benchmark test data of CAE software running on a domestic chip platform. The benchmark test data includes hardware and software status, software operation sequence and corresponding performance indicators. Specifically, the benchmark test data is generated by running a series of carefully designed benchmark test programs, which cover different computing modes, data sizes, and concurrency levels; the hardware and software states in the benchmark test data include but are not limited to CPU (Central Processing Unit) / DCU (Data Center Unit) utilization, cache behavior, memory bandwidth, and power; and the corresponding performance indicators include but are not limited to execution time, throughput, and frame rate.

[0021] S2. Preprocessing the original data to obtain original data features, labeling part of the original data features according to a preset label value to obtain labeled part of the original data features, and taking the original data features and the labeled part of the original data features as training data; Specifically, in step S2, the original data is preprocessed to obtain original data, including: S21. Data cleaning is performed on the original data, and the data cleaning includes removing outliers and filling missing values to obtain cleaned original data; S22. The cleaned original data is converted to obtain converted original data; S23. The converted original data is subjected to feature engineering processing to obtain original data features.

[0022] Further, part of the original data features is labeled according to a preset label value, and the preset label value includes performance indicators and performance bottleneck types.

[0023] In a possible design, the cleaned original data includes time series data, the time series data is converted to obtain converted original data, and the converted original data includes frequency domain features.

[0024] S3. Training the training data in the initialized AI model to obtain a trained AI model; Specifically, in step S3, the training data is input into the initialized AI model to obtain a trained AI model, including: S31. Training the training data in the initialized AI model, taking the training data as the input value of the initialized AI model, and taking the predicted performance indicators and performance bottleneck types as the output value of the initialized AI model; S32. Based on the loss function, the predicted performance indicators, the performance bottleneck types, and the performance indicators and performance bottleneck types in the preset label value, a loss value is calculated, and the parameters of the initialized AI model are updated based on the loss value; S33. Repeat the above update step until a preset iteration termination condition is reached to obtain a trained AI model.

[0025] In a possible design, the AI model is an AI-driven software-hardware interaction model, and the AI model includes a perception module, a prediction module, and a decision module. The perception module is configured to perform feature extraction on real-time state data and real-time load features to obtain a real-time feature vector. The prediction module is configured to perform prediction on the real-time feature vector to obtain a prediction result. The decision module is configured to generate an optimization decision based on the prediction result.

[0026] Further, the AI model can be trained by using a supervised learning method or an unsupervised learning method. For example, the supervised learning method can be used to train the AI model by using a regression method and historical running data (including performance of a specific software operation under different chip configurations) to predict a performance indicator (such as a response time or a throughput) of the CAE software under a current chip state and a given load, or the supervised learning method can be used to train the AI model by using a classification method to identify a main performance bottleneck type (such as a CPU computing bottleneck, a memory bandwidth bottleneck, or an I / O bottleneck) in a current running state. For example, the unsupervised learning method can be used to automatically find a typical behavior mode of the CAE software in a running process and a corresponding chip resource consumption feature by using clustering analysis, to provide a basis for a subsequent adaptation decision.

[0027] In a possible design, after obtaining the trained AI model, a method of cross-validation is further used to evaluate the trained AI model, and a hyperparameter of the trained AI model is adjusted according to an evaluation result.

[0028] Specifically, the cross-validation is a statistical method for evaluating model performance, and the generalization ability of the model is evaluated by dividing an original data set into multiple subsets, training and testing multiple times on different subsets. The hyperparameter is a parameter manually set before training of the model, and the hyperparameter includes, but is not limited to, a learning rate and a network layer number.

[0029] S4. Obtain real-time state data of a domestic chip platform and real-time load features of the CAE software. S5. Input the real-time state data and the real-time load features into the trained AI model to obtain an optimization decision, the optimization decision including dynamic adjustment of a parameter of a software running time, adjustment of a chip resource configuration, and selection or recommendation of using a specific algorithm, performing instruction translation on the optimization decision to obtain a specific control instruction, sending the specific control instruction to a chip hardware to automatically adjust a hardware behavior and / or automatically adjusting a software behavior through a software running interface.

[0030] The optimization decision includes dynamically adjusting parameters of a software runtime, such as a number of threads, core binding, and a memory allocation strategy, adjusting chip resource configuration, and selecting or recommending to use a specific algorithm, wherein the adjusting chip resource configuration includes CPU frequency, DCU power consumption mode, and HBM data migration; and the cloud-edge-end collaborative architecture combines terminal devices to improve the efficiency of data processing and enhance privacy protection.

[0031] Specifically, the specific control instruction acts on chip hardware through a hardware abstraction layer or adjusts software and hardware behaviors through an operating system API and / or a software runtime interface, wherein the action on the chip hardware includes, but is not limited to, adjusting CPU frequency through ACPI (Advanced Configuration and Power Interface) or a specific driver interface; and the software runtime interface includes, but is not limited to, JVM parameter adjustment and a configuration interface provided by CAD / CAE software.

[0032] Specifically, in step S5, after obtaining the optimization decision, the following steps are further included: S51. Obtain field expert experience, update the pre-constructed knowledge base based on the trained AI model, benchmark test data, and field expert experience, wherein the knowledge base includes optimization rules; S52. Send the optimization rules, optimization decisions, and trained AI model to each node of the pre-constructed cloud-edge-end collaborative architecture through a cloud-edge-end collaborative mechanism to form a data optimization closed-loop mechanism.

[0033] Further, the optimization rules are stored in the form of IF-THEN, and the knowledge base further includes efficient parameter configuration combinations and solutions to typical performance problems; based on the trained AI model, the accuracy of CAE software (computer aided engineering) runtime performance index prediction and the identification accuracy of different types of performance bottlenecks are obtained; and the pre-constructed knowledge base is updated based on the accuracy of CAE software runtime performance index prediction and the identification accuracy of different types of performance bottlenecks, benchmark test data, and field expert experience.

[0034] In a possible design, the cloud-edge-end collaborative architecture includes a cloud data platform, and the method further includes: obtaining running data, optimization records, and effect evaluation data within a preset time interval, and uploading the running data, optimization records, and effect evaluation data to the cloud data platform, wherein the running data is running data of the CAE software on a domestic chip platform, the optimization records are data recorded after the optimization decision is executed, and the effect evaluation data is data for evaluating the CAE software after the optimization decision is executed; Periodically train the trained AI model based on the running data, optimization records and effect evaluation data stored in the cloud data platform to obtain a periodically trained AI model; Again use the periodically trained AI model to process the real-time state data of the chip and the real-time load characteristics of the CAE software to obtain a new optimization decision.

[0035] The method provided by the embodiment of the application provides a chip and CAE software collaborative optimization method combining AI technology, and through AI-driven closed-loop optimization, dynamic fine management of software and hardware resources and deep optimization at the algorithm level are realized, the efficiency of domestic hardware is maximized, the system is endowed with the ability of perception, analysis, decision and learning, and the transformation from passive adaptation to active collaboration is realized.

[0036] As shown in Figure 2 The second aspect of the embodiment provides a chip and CAE software collaborative optimization system combining AI technology, which comprises: A first acquisition module configured to acquire original data, wherein the original data comprises historical performance data of CAE software running on a domestic chip platform and / or benchmark test data, and the benchmark test data comprises software and hardware states, software operation sequences and corresponding performance indicators; A data preprocessing module configured to preprocess the original data to obtain original data features, label part of the original data features according to preset label values to obtain labeled part of the original data features, and use the original data features and the labeled part of the original data features as training data; A model training module configured to input the training data into an initialized AI model to train the AI model and obtain a trained AI model; A second acquisition module configured to acquire real-time state data of the domestic chip platform and real-time load characteristics of the CAE software; An optimization decision module configured to input the real-time state data and the real-time load characteristics into the trained AI model to obtain an optimization decision, wherein the optimization decision comprises dynamically adjusting parameters of software runtime, adjusting chip resource configuration and selecting or recommending to use a specific algorithm, to translate the optimization decision into specific control instructions, send the specific control instructions to chip hardware to automatically adjust hardware behavior and / or automatically adjust software behavior through a software running interface.

[0037] In a possible design, the domestic chip platform of this embodiment is a server / workstation equipped with Haiguang C86 series CPUs and / or Haiguang DCUs. The latest compatible drivers and firmware are deployed on the domestic chip platform. When first running, the system is automatically or manually configured by an administrator to identify key hardware information of the chip, such as the specific model, the number of cores, the clock frequency, the memory capacity and hierarchy, the DCU model, and the HBM capacity. The memory capacity and hierarchy include the number of NUMA nodes and the topology.

[0038] In a possible design, the version of the CAE software is a preset specific version.

[0039] In a possible design, according to the cloud-edge-end collaborative architecture design, each component of the collaborative optimization system is installed on a corresponding node, such as a user terminal, an edge server, and a cloud platform. The components include but are not limited to a hardware perception agent, a software monitoring agent, an AI decision service, an optimization decision executor, a data processing module, an AI model library, and a knowledge base. Meanwhile, the start parameters and dependency relationships of each component are configured.

[0040] In this embodiment, the network connection between the cloud, the edge, and the end is configured to ensure that reliable and secure data and control instruction transmission can be performed between components, such as using HTTPS, AMQP, and MQTT protocols. Firewall rules and access control policies are set to ensure system security.

[0041] In a possible design, the pre-trained AI model is loaded into a corresponding inference engine, such as an end-side, edge, and cloud, in advance. The pre-built knowledge base is initialized, which includes but is not limited to general optimization rules based on expert experience or preliminary configuration suggestions extracted from a small amount of prior data.

[0042] In a possible design, if the optimization strategy involves algorithm-level optimization, the system can automatically call a pre-optimized algorithm library function or prompt the user to switch to a recommended optimization module.

[0043] The working process, working details, and technical effects of the domestic chip and CAE software collaborative optimization system provided in the second aspect of this embodiment in combination with the AI technology can be referred to the method for collaborative optimization of the domestic chip and CAE software in combination with the AI technology described in the first aspect, which will not be repeated here.

[0044] The third aspect of the embodiment provides a computer device, comprising a memory, a processor and a transceiver connected in sequence, wherein the memory is used for storing a computer program, the transceiver is used for transmitting and receiving messages, and the processor is used for reading the computer program and executing the method for collaborative optimization of a domestic chip and CAE software in combination with an AI technology according to the first aspect. Specifically, the memory can include, but is not limited to, a random access memory (RAM), a read-only memory (ROM), a flash memory, a first input first output (FIFO) memory and / or a first input last output (FILO) memory, etc.; and the processor can be, but is not limited to, a microprocessor with a model number of STM32F105 series. In addition, the computer device can further include, but is not limited to, a power module, a display screen and other necessary components.

[0045] The working process, working details and technical effects of the aforementioned computer device provided by the third aspect of the embodiment can be referred to the method for collaborative optimization of a domestic chip and CAE software in combination with an AI technology according to the first aspect, which will not be described here.

[0046] The fourth aspect of the embodiment provides a computer program product comprising a computer program or instructions, which, when executed by a computer, is used to implement the method for collaborative optimization of a domestic chip and CAE software in combination with an AI technology according to the first aspect.

[0047] The working process, working details and technical effects of the aforementioned computer program product provided by the embodiment can be referred to the method for collaborative optimization of a domestic chip and CAE software in combination with an AI technology according to the first aspect, which will not be described here.

[0048] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for collaborative optimization of domestically produced chips and CAE software combining AI technology, characterized in that, include: Obtain raw data, which includes historical performance data and / or benchmark test data of CAE software running on a domestic chip platform. The benchmark test data includes hardware and software status, software operation sequence and corresponding performance indicators. The original data is preprocessed to obtain the original data features. Some of the original data features are labeled according to the preset label values ​​to obtain the labeled partial original data features. The original data features and the labeled partial original data features are used as training data. The training data is input into the initialized AI model for training, resulting in the trained AI model. Obtain real-time status data of domestic chip platforms and real-time load characteristics of CAE software; Real-time status data and real-time load characteristics are input into the trained AI model to obtain optimization decisions. These optimization decisions include dynamically adjusting the parameters during software operation, adjusting chip resource configuration, and selecting or recommending the use of specific algorithms. The optimization decisions are translated into specific control instructions, which are then sent to the chip hardware to automatically adjust hardware behavior and / or automatically adjust software behavior through the software operation interface.

2. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology according to claim 1, characterized in that, The raw data is preprocessed to obtain its features, including: The original data is cleaned, which includes removing outliers and filling in missing values, to obtain the cleaned original data. The cleaned raw data is then transformed to obtain the transformed raw data. Feature engineering is performed on the transformed original data to obtain the original data features.

3. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology according to claim 1, characterized in that, The preset label values ​​include performance metrics and performance bottleneck types; the step of inputting training data into the initialized AI model for training to obtain the trained AI model includes: The training data is input into the initialized AI model for training. The training data is used as the input value of the initialized AI model, and the predicted performance indicators and performance bottleneck types are used as the output values ​​of the initialized AI model. The loss value is calculated based on the loss function, the predicted performance metrics, the performance bottleneck type, and the performance metrics and performance bottleneck types in the preset label values. The parameters of the initialized AI model are then updated based on the loss value. Repeat the above update steps until the preset iteration termination condition is met to obtain the trained AI model.

4. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology according to claim 1, characterized in that, The AI ​​model is an AI-driven software and hardware interaction model, including a perception module, a prediction module, and a decision module. The perception module is used to extract features from real-time status data and real-time load characteristics to obtain real-time feature vectors. The prediction module is used to predict the real-time feature vectors to obtain prediction results. The decision module is used to generate optimization decisions based on the prediction results.

5. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology according to claim 1, characterized in that, After obtaining the trained AI model, the process also includes evaluating the trained AI model using cross-validation and adjusting the hyperparameters of the trained AI model based on the evaluation results.

6. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology according to claim 1, characterized in that, After obtaining the optimized decision, it also includes: The pre-built knowledge base, which includes optimization rules, is updated based on the acquired domain expert experience, the trained AI model, benchmark test data, and the domain expert experience. The optimized rules, optimized decisions, and trained AI models are sent to various nodes of the pre-built cloud-edge-device collaborative architecture through a cloud-edge-device collaboration mechanism, forming a closed-loop data optimization mechanism.

7. The method for collaborative optimization of domestically produced chips and CAE software combining AI technology according to claim 6, characterized in that, The cloud-edge-device collaborative architecture includes a cloud data platform, and the method further includes: Within a preset time interval, the system acquires operational data, optimization records, and effect evaluation data, and uploads these data to a cloud data platform. The operational data refers to the operational data of the CAE software on a domestic chip platform, the optimization records are the data recorded after executing optimization decisions, and the effect evaluation data is the data used to evaluate the CAE software after executing optimization decisions. The AI ​​model is periodically trained based on the operational data, optimization records, and performance evaluation data stored in the cloud data platform to obtain the periodically trained AI model. The AI ​​model, after being periodically trained, is used again to process the chip's real-time status data and the real-time load characteristics of the CAE software to obtain new optimization decisions.

8. A domestically produced chip and CAE software collaborative optimization system combining AI technology, used to implement the method according to any one of claims 1 to 7, characterized in that, include: The first acquisition module is used to acquire raw data, which includes historical performance data and / or benchmark test data of CAE software running on a domestic chip platform. The benchmark test data includes hardware and software status, software operation sequence and corresponding performance indicators. The data preprocessing module is used to preprocess the raw data to obtain raw data features, label some raw data features according to preset label values ​​to obtain labeled raw data features, and use the raw data features and labeled raw data features as training data. The model training module is used to input training data into the initialized AI model for training, and obtain the trained AI model. The second acquisition module is used to acquire real-time status data of the domestic chip platform and real-time load characteristics of the CAE software. The optimization decision module is used to input real-time status data and real-time load characteristics into the trained AI model to obtain optimization decisions. The optimization decisions include dynamically adjusting the parameters of the software runtime, adjusting the chip resource configuration, and selecting or recommending the use of specific algorithms. The optimization decisions are translated into instructions to obtain specific control instructions, which are then sent to the chip hardware to automatically adjust hardware behavior and / or automatically adjust software behavior through the software runtime interface.

9. A computer device, characterized in that, The device includes a memory, a processor, and a transceiver that are sequentially and communicatively connected. The memory is used to store computer programs, the transceiver is used to send and receive messages, and the processor is used to read the computer programs and execute the collaborative optimization method of domestically produced chips and CAE software that combines AI technology as described in any one of claims 1 to 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or the instructions are executed by the computer, they implement the collaborative optimization method of domestically produced chips and CAE software that combines AI technology as described in any one of claims 1 to 7.

Images