CN116414567A - Resource scheduling method, device and equipment for smart car operating system - Google Patents

Abstract

The present application provides a resource scheduling method, device and equipment for a smart car operating system, including: performing at least one performance test on the artificial intelligence model in the smart car operating system, obtaining at least one static performance index, and identifying at least one static performance index Priority static index, set the static strategy corresponding to the priority static index as a static allocation strategy; according to the static allocation strategy, allocate graphics processor resources for each artificial intelligence model in the smart car operating system; run the smart car operation At least one artificial intelligence model in the system, and monitor the dynamic performance index of the running artificial intelligence model; generate a dynamic allocation strategy according to the dynamic performance index, and adjust the graphics processor resources allocated to each artificial intelligence model according to the dynamic allocation strategy. The application avoids the situation that multiple requests compete for the working thread in the graphics processor at the same time, and ensures the stability of the resource call of the graphics processor.

Description

Resource scheduling method, device and equipment for smart car operating system

technical field

The present application relates to the technical field of smart cars, in particular to a resource scheduling method, device and equipment for a smart car operating system.

Background technique

With the development of intelligent driving technology, the automatic driving scene of smart cars has also emerged as the times require. The automatic driving scene is constructed based on the reasoning conclusions generated by the graphics processor through multiple artificial intelligence models; among them, the artificial intelligence model is used to solve A problem or a series of calculations and rules for analyzing a set of data, it is like a flowchart containing step-by-step instructions for posing a problem, but written in the form of mathematical and programming code, so artificial intelligence models are often used to implement Radar perception, visual perception, path planning and other specified tasks; the reasoning conclusion is based on multiple artificial intelligence models to send reasoning requests to the graphics processor, so that the graphics processor calls the artificial intelligence model and generates it based on the collected data loaded into the artificial intelligence model inference results.

However, the inventors found that when multiple artificial intelligence models in the current autonomous driving scene send inference requests to the graphics processor, it is easy for multiple requests to compete for the working thread in the graphics processor at the same time, causing the graphics processor to fail. Worker threads are frequently called by conflicting inference requests, resulting in unstable graphics processor resource calls.

Contents of the invention

This application provides a resource scheduling method, device, and equipment for an intelligent vehicle operating system, which are used to solve the problem that when multiple artificial intelligence models in the current automatic driving scene send reasoning requests to the graphics processor, it is easy to cause multiple requests to compete at the same time. The situation of grabbing the working thread in the graphics processor causes the graphics processor to be frequently invoked by conflicting inference requests due to the working thread, resulting in unstable resource calls of the graphics processor.

In a first aspect, the present application provides a resource scheduling method for a graphics processor of an intelligent vehicle operating system, including:

Perform at least one performance test on the artificial intelligence model in the smart car operating system to obtain at least one static performance index; and identify the priority static index in the at least one static performance index, and set the static strategy corresponding to the priority static index to Static allocation strategy, wherein, there is at least one artificial intelligence model in the smart car operating system, the artificial intelligence model is used to realize the specified task, the specified task is used to realize the automatic driving of the car, and the static performance index reflects the artificial intelligence model. The performance of the model in the performance test, the priority static index is a static performance index that meets the preset priority rules, and the static strategy is used to group the artificial intelligence models and assign work to the grouped artificial intelligence models a computer policy for threads, said worker thread being a sequential GPU for scheduling stream processors and/or compute units in said graphics processor;

According to the static allocation strategy, allocate graphics processor resources for each artificial intelligence model in the intelligent vehicle operating system;

Run at least one artificial intelligence model in the smart car operating system, and monitor the dynamic performance indicators of the running artificial intelligence model, wherein the dynamic performance indicators reflect that the artificial intelligence model is calling the allocated graphics processor resources to perform calculations performance when

A dynamic allocation strategy is generated according to the dynamic performance index, and the graphics processor resources allocated to each artificial intelligence model are adjusted according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used for the artificial intelligence model at runtime according to the dynamic allocation strategy. Performance, allocating GPU resources to running AI models.

In the above scheme, the artificial intelligence model in the smart car operating system conducts at least one performance test to obtain at least one static performance index, including:

Obtaining at least one static strategy, grouping the artificial intelligence models in the smart car operating system according to each static strategy and assigning work threads to each group of artificial intelligence models to obtain at least one static sample, wherein the The static strategy is a computer strategy for grouping artificial intelligence models and assigning working threads to the grouped artificial intelligence models;

Entering the preset test instance into the artificial intelligence model in each of the static samples, and running the test instance in the artificial intelligence model through the worker threads assigned to the artificial intelligence model in each of the static samples, Performance testing is performed on each of the static samples, and at least one static performance indicator corresponding to at least one static strategy is obtained, wherein the test instance is a test case for performance testing of an artificial intelligence model.

In the above scheme, according to each of the static strategies, the artificial intelligence models in the smart car operating system are grouped respectively and work threads are assigned to each group of artificial intelligence models to obtain at least one static sample, including:

Group each artificial intelligence model in the smart car operating system according to the division rules in the static strategy to obtain at least one test group, wherein there is at least one artificial intelligence model in the test group;

Acquiring at least one worker thread in the graphics processor, and assigning a worker thread to each of the test groups according to the resource allocation rules in the static policy;

Summarizing at least one test group and working threads corresponding to each test group to form a static sample corresponding to the static policy.

In the above scheme, each artificial intelligence model in the smart car operating system is grouped according to the division rules in the static strategy to obtain at least one test group, including:

If it is determined that there is at least one directed acyclic graph in the smart car operating system, the artificial intelligence models belonging to the same directed acyclic graph are divided into a test group, wherein the directed acyclic graph reflects the intelligence The logical relationship between two or more artificial intelligence models in the automotive operating system;

If it is determined that the smart car operating system has other artificial intelligence models that do not belong to the directed acyclic graph, then group the other artificial intelligence models according to the model attribute data of each of the other artificial intelligence models to obtain At least one test group, wherein the model attribute data describes the computing power consumed by the artificial intelligence model to achieve a specified task;

If it is determined that there is no directed acyclic graph in the smart car operating system, the artificial intelligence models in the smart car operating system are grouped according to the model attribute data of each artificial intelligence model to obtain at least one test group.

In the above solution, identifying the priority static index in the at least one static performance index includes:

Extracting the first index element in each static performance index, and sorting the first index elements to obtain a target sequence, wherein there is at least one static index element in the static performance index, and the static index element reflects the artificial intelligence The performance of the model on a performance dimension in the performance test, the first index element is a static index element in the static performance index;

Determine the performance value of each first index element according to the position of each first index element in the target sequence, wherein the performance value reflects the degree of performance of the first index element;

Obtaining the comprehensive performance value of each static performance index according to the performance value of each static index element in each static performance index;

The static performance index with the highest comprehensive performance value is set as the priority static index.

In the above scheme, according to the static allocation strategy, the allocation of graphics processor resources for each artificial intelligence model in the intelligent vehicle operating system includes:

According to the division rules in the static allocation strategy, group the artificial intelligence models in the smart car operating system to obtain at least one operation group;

Acquiring at least one worker thread in the graphics processor, assigning a worker thread to each operation group according to the resource allocation rules in the static allocation strategy, so as to allocate resources to each artificial intelligence model in each operation group Graphics processor resources.

In the above solution, generating a dynamic allocation strategy according to the dynamic performance index includes:

Extracting a second index element in the dynamic performance index, and obtaining an index rule corresponding to the second index element, wherein the dynamic performance index has at least one index element, and the second index element is the An index element in the dynamic performance index, the index rule is a computer rule used to define normal and abnormal index elements;

If it is determined that the second index element conforms to the index rule, setting the second index element as a normal index element;

If it is determined that the second index element does not comply with the index rule, setting the second index element as an abnormal index element;

If it is determined that the number of normal index elements in the dynamic performance index does not reach a preset normal threshold, or the number of abnormal index elements reaches a preset abnormal threshold, then determine that the dynamic performance index is an abnormal performance index;

If it is determined that the number of normal index elements in the dynamic performance index reaches a preset normal threshold, or the number of abnormal index elements does not reach a preset abnormal threshold, then determine that the dynamic performance index is a normal performance index;

A dynamic allocation policy is generated according to the normal performance index and the abnormal performance index.

In the above solution, a dynamic allocation strategy is generated according to the normal performance index and the abnormal performance index, including:

The artificial intelligence model corresponding to the normal performance index is set as a normal model, the artificial intelligence model corresponding to the abnormal performance index is set as an abnormal model, and the operating group where the normal model is located and the operating group of the smart car operating system are not The operating group where the artificial intelligence model is located is set as a normal group, and the operating group where the abnormal model is located is an abnormal group;

If it is determined that there is a logical relationship between the abnormal model and other artificial intelligence models in the abnormal group; then adjust the division rules in the static allocation strategy or the dynamic allocation strategy, so that the adjusted division rules are used to divide The independent model in the abnormal group is adjusted to a normal group; and/or the resource allocation rule in the static allocation strategy or the dynamic allocation strategy is adjusted so that the adjusted resource allocation rule is used to map the abnormal group to The working thread is adjusted to a corresponding working thread of a normal group; wherein, the independent model is an artificial intelligence model that has no logical relationship with other artificial intelligence models in the abnormal group;

If it is determined that there is no logical relationship between the abnormal model and other artificial intelligence models in the abnormal group; then adjust the division rules in the static allocation strategy or the dynamic allocation strategy, so that the adjusted division rules are used for Adjusting the exception model to a normal group; and/or adjusting the resource allocation rules in the static allocation strategy or the dynamic allocation strategy, so that the adjusted resource allocation rules are used to allocate the worker threads corresponding to the exception group , adjusted to a worker thread corresponding to a normal group;

Generate a dynamic allocation policy according to the adjusted division rule and/or the adjusted resource allocation rule.

In a second aspect, the present application provides a resource scheduling device for a graphics processor of a smart car operating system, including:

The static test module is used to perform at least one performance test on the artificial intelligence model in the smart car operating system to obtain at least one static performance index; and identify the priority static index in the at least one static performance index, and use the priority static index The corresponding static strategy is set as a static allocation strategy, wherein, there is at least one artificial intelligence model in the smart car operating system, and the artificial intelligence model is used to realize a specified task, and the specified task is used to realize the automatic driving of the car. The performance index reflects the performance of the artificial intelligence model in the performance test. The priority static index is a static performance index that meets the preset priority rules. The static strategy is used to group the artificial intelligence models and group the A computer strategy for assigning worker threads of an artificial intelligence model that is a sequential graphics processor for scheduling stream processors and/or compute units in the graphics processor;

A static allocation module, configured to allocate graphics processor resources to each artificial intelligence model in the smart car operating system according to the static allocation strategy;

A dynamic monitoring module, configured to run at least one artificial intelligence model in the smart car operating system, and monitor the dynamic performance indicators of the running artificial intelligence model, wherein the dynamic performance indicators reflect that the artificial intelligence model is invoking the assigned The performance of graphics processor resources when performing calculations;

A dynamic allocation module, configured to generate a dynamic allocation strategy according to the dynamic performance index, and adjust graphics processor resources allocated to each artificial intelligence model according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used to The performance of the intelligent model at runtime, and allocate graphics processor resources to the running artificial intelligence model.

In a third aspect, the present application provides a computer device, including: a processor and a memory communicatively connected to the processor;

the memory stores computer-executable instructions;

The processor executes the computer-executed instructions stored in the memory, so as to implement the resource scheduling method for a graphics processor as claimed in the claims.

In a fourth aspect, the present application provides a computer-readable storage medium, where computer-executable instructions are stored in the computer-readable storage medium, and when the computer-executable instructions are executed by a processor, they are used to implement the resource scheduling of the above-mentioned graphics processor method.

In a fifth aspect, the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the above resource scheduling method for a graphics processor is implemented.

A resource scheduling method, device, and equipment for a smart car operating system provided by the present application, by performing at least one performance test on the artificial intelligence model in the smart car operating system, before running the smart car operating system, the artificial intelligence in it Static performance testing is performed on the model to obtain at least one static performance index, so as to determine the performance of each artificial intelligence model in different groups and assign different working threads to it in the smart car operating system.

By identifying the priority static index in the at least one static performance index, to identify the best performance grouping method and working thread matching method, and then optimize the rationality of the allocation of graphics processor resources to the greatest extent before running the smart car system , and set the optimal grouping method and worker thread matching method as a static allocation strategy.

Allocate graphics processor resources to each artificial intelligence model through the static allocation strategy corresponding to the priority static index, so as to realize that the graphics processor resources with optimal performance have been allocated to each artificial intelligence model before the operating system of the smart car is running .

The dynamic performance indicators of the running artificial intelligence model are monitored through the performance acquisition module. The indicator elements in the dynamic performance indicators include: CPU usage, memory usage, disk IO, system average load, delay, and frame rate; according to the dynamic performance The index generates a dynamic allocation strategy to realize the allocation of graphics processor resources to the artificial intelligence model based on the performance of the current artificial intelligence model at runtime, so as to ensure that each running artificial intelligence model has enough graphics processor resources for calling, solving It prevents the inference requests generated by the artificial intelligence model from conflicting on the invocation of the working thread of the graphics processor, avoids the situation where multiple requests compete for the working thread of the graphics processor at the same time, and ensures the multiple artificial intelligence models in the automatic driving scene performance, thereby ensuring the stability of graphics processor resource calls.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application.

FIG. 1 is a schematic diagram of an application scenario provided by an embodiment of the present application;

FIG. 2 is a flow chart of Embodiment 1 of a resource scheduling method for a graphics processor provided in an embodiment of the present application;

3 is a schematic diagram of program modules of a resource scheduling device for a graphics processor provided by the present invention;

Fig. 4 is a schematic diagram of the hardware structure of the computer device in the computer device of the present invention.

By means of the above drawings, specific embodiments of the present application have been shown, which will be described in more detail hereinafter. These drawings and text descriptions are not intended to limit the scope of the concept of the application in any way, but to illustrate the concept of the application for those skilled in the art by referring to specific embodiments.

Detailed ways

Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numerals in different drawings refer to the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with aspects of the present application as recited in the appended claims.

Please refer to Figure 1, the specific application scenario of this application is:

The control unit 11 running the resource scheduling method of the graphics processor of the smart car operating system is installed in the smart car operating system 1 , and the control unit 11 is connected with the artificial intelligence model 12 and the graphics processor 13 in the smart car operating system 1 .

The control unit 11 performs at least one performance test on the artificial intelligence model 12 in the smart car operating system 1, obtains at least one static performance index, identifies the priority static index in the at least one static performance index, and sets the static strategy corresponding to the priority static index to static allocation strategy

The control unit 11 allocates graphics processor 13 resources to each artificial intelligence model 12 in the intelligent vehicle operating system 1 according to a static allocation strategy.

The control unit 11 runs at least one artificial intelligence model 12 in the smart car operating system 1, and monitors the dynamic performance index of the running artificial intelligence model 12.

The control unit 11 generates a dynamic allocation strategy according to the dynamic performance index, and adjusts the graphics processor resources allocated to each artificial intelligence model 12 according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used to perform according to the performance of the artificial intelligence model 12 at runtime, Allocate graphics processor resources to the running artificial intelligence model 12 .

The control unit 11 is a computer module installed in a circuit or chip with the smart car operating system 1 , and the graphics processor 13 is a graphics accelerator installed in the circuit or chip as a part of the smart car operating system 1 .

The technical solution of the present application and how the technical solution of the present application solves the problems in the prior art will be described in detail below with specific embodiments. The following specific embodiments may be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below in conjunction with the accompanying drawings.

Example 1:

Please refer to Fig. 2, the present application provides a kind of resource scheduling method of the graphic processor of intelligent car operating system, including:

S201: Perform at least one performance test on the artificial intelligence model in the smart car operating system to obtain at least one static performance index; and identify the priority static index in at least one static performance index, and set the static policy corresponding to the priority static index as static allocation Strategy, wherein the smart car operating system has at least one artificial intelligence model, the artificial intelligence model is used to realize the specified task, the specified task is used to realize the automatic driving of the car, and the static performance index reflects the performance of the artificial intelligence model in the performance test , the priority static index is a static performance index that satisfies the preset priority rules. The static strategy is a computer strategy for grouping artificial intelligence models and assigning work threads to the grouped artificial intelligence models. Work threads are used to schedule graphs. A stream processor in a processor and/or a sequential graphics processor in a compute unit.

In this step, the smart car operating system is the underlying operating system (Operating System, OS) of the smart car, which is used to control and manage the hardware and software resources of the entire smart car, and provide interfaces and environments for users and other software. Current smart car operating systems include: autopilot OS and smart cockpit OS. Autonomous driving OS has very high requirements on safety, real-time performance and stability, while smart cockpit OS pays more attention to openness and compatibility.

The autonomous driving OS is mainly used for vehicle chassis and power control to realize basic driving functions such as accelerator, steering, gear shifting, and braking. Autopilot OS is the core of L3 and above-level autopilot functions that are currently under development. It includes high-performance complex embedded systems, artificial intelligence chips and algorithms, high-speed networks, massive data processing, and cloud-machine collaboration. .

The smart cockpit OS is a complete system composed of different cockpit electronics. The smart cockpit is mainly divided into 5 parts: vehicle infotainment system, streaming media central rearview mirror, head-up display system HUD, full LCD instrument, car networking module. The smart cockpit realizes human-computer interaction through multi-screen integration. It uses LCD instrumentation, HUD, central control panel, central control vehicle information terminal, rear seat HMI entertainment screen, and interior and exterior rearview mirrors as carriers to realize voice control and gesture operations. A smarter way of interacting. In the future, it is possible to integrate artificial intelligence, AR, ADAS, VR and other technologies into it.

The worker thread is used to schedule the stream processor (SP, streaming processor, which is the most basic processing unit in the graphics processor, also known as CUDA core) in the graphics processor. The specific instructions and tasks in the graphics processor are all is processed on the SP.

The worker thread is also used to schedule computing units in the graphics processor (SM, streaming multiprocessor, also known as the large core of the graphics processor). SM is composed of multiple SPs plus some other resources to form a large graphics processor core. Among them, other resources such as: warp scheduler, register, shared memory, etc., SM can be regarded as the heart of the graphics processor (compared to the CPU core), Register and shared memory are scarce resources of SM.

In this step, at least one performance test is performed on the artificial intelligence model in the smart car operating system, so as to perform a static performance test on the artificial intelligence model in the smart car operating system before running the smart car operating system to obtain at least one static performance index to determine the intelligence The performance of each artificial intelligence model in the automotive operating system when it is divided into different groups and assigned different working threads.

By identifying the priority static index in at least one static performance index, to identify the grouping method and the worker thread matching method with the best performance, and then optimize the rationality of the allocation of graphics processor resources to the greatest extent before running the smart car system, and Set the optimal grouping method and worker thread matching method as a static allocation strategy.

In a preferred embodiment, at least one performance test is carried out to the artificial intelligence model in the smart car operating system to obtain at least one static performance index, including:

Obtaining at least one static strategy, obtaining at least one static strategy, grouping the artificial intelligence models in the smart car operating system according to each static strategy and assigning work threads to each group of artificial intelligence models to obtain at least one static sample, wherein, The static strategy includes division rules and resource allocation rules. The division rules are used to group the artificial intelligence models in the smart car operating system, and the resource allocation rules are used to allocate working threads to each group of artificial intelligence models.

Specifically, according to each static strategy, the artificial intelligence models in the smart car operating system are grouped respectively and work threads are assigned to each group of artificial intelligence models to obtain at least one static sample, including:

Group each artificial intelligence model in the smart car operating system according to the division rules in the static strategy to obtain at least one test group, wherein there is at least one artificial intelligence model in the test group;

Obtain at least one worker thread in the graphics processor, and assign a worker thread to each test group according to the resource allocation rules in the static policy;

Summarize at least one test group and the worker threads corresponding to each test group to form a static sample corresponding to the static strategy.

Exemplarily, for example: the artificial intelligence models in the smart car operating system include: Model 1, Model 2, Model 3, Model 4, and Model 5.

If the static strategy 1 is to group the models in pairs, if there are any leftovers, group them separately, then you will get: test group 1: model 1, model 2; test group 2: model 3, model 4; test group 3: model 5.

If static strategy 2 is to list a specified model as a group (for example: model 3) and other models in pairs, then you will get: test group 1: model 3; test group 2: model 1, model 2 ;Test Group 3: Model 4, Model 5.

There is at least one worker thread in the graphics processor, and each worker thread invokes its corresponding stream processor or computing unit. Therefore, the graphics processor resources that can be invoked by each worker thread are different. Assume that the working threads in the graphics processor include: thread 1 and thread 2; then, static strategy 1 assigns test group 1 and test group 2 to thread 1, and assigns test group 3 to thread 2; static strategy 2 assigns test group 1 Assign to thread 1, assign test group 2 and test group 3 to thread 2, and so on.

Therefore, based on the above example, static samples corresponding to static strategy 1 can be obtained: model 1, model 2, model 3, model 4 assigning thread 1, model 5 assigning thread 2, and static samples corresponding to static strategy 2: model 3 Thread 1 is allocated, model 1, model 2, model 4, and model 5 allocate static samples of thread 2, and so on, which will not be described here.

Further, according to the division rules in the static strategy, each artificial intelligence model in the smart car operating system is grouped to obtain at least one test group, including:

If it is determined that there is at least one directed acyclic graph in the smart car operating system, the artificial intelligence models belonging to the same directed acyclic graph are divided into a test group, wherein the directed acyclic graph reflects the The logical relationship between two or more artificial intelligence models;

If it is determined that there are other artificial intelligence models that do not belong to the directed acyclic graph in the smart car operating system, then group other artificial intelligence models according to the model attribute data of each other artificial intelligence model to obtain at least one test group, wherein the model Attribute data describes the computing power consumed by the artificial intelligence model to complete the specified task;

If it is determined that there is no directed acyclic graph in the smart car operating system, the artificial intelligence models in the smart car operating system are grouped according to the model attribute data of each artificial intelligence model to obtain at least one test group.

Exemplarily, if it is determined that the directed acyclic graph in the smart car operating system includes: a first graph and a second graph, the first graph represents model 1 pointing to model 2, and the second graph represents model 3 pointing to model 4, then, Model 1 and Model 2 are divided into a test group, and Model 3 and Model 4 are divided into a test group.

If it is determined that there are other artificial intelligence models in the smart car operating system that do not belong to the directed acyclic graph, such as model 5 and model 6, and model 5 and model 6 neither belong to the first graph nor to the second graph, then according to the model 5 and the attribute data of model 6 to group model 5.

The model attribute data describes the computing power consumed by the artificial intelligence model to achieve the specified task; for example: the specified task used by model 5 is radar perception, and the computing power consumed by it is M1; model 6 is the specified task used to realize The task is visual perception, and the computing power it consumes is M2.

It should be noted that a directed acyclic graph is composed of a finite number of vertices and "directed edges". Starting from any vertex, passing through several directed edges, it is impossible to return to the vertex. This kind of graph is a directed acyclic graph. Graph, in this embodiment, the artificial intelligence model is a vertex in a directed acyclic graph, and the directed edge in the directed acyclic graph is used to describe the relationship between two artificial intelligence models, such as: dependency, Association relationship, aggregation relationship, combination relationship, etc.; the artificial intelligence models belonging to the same directed acyclic graph are interdependent, and the specified tasks realized by the artificial intelligence model in the previous directed acyclic graph are the artificial intelligence models in the next one. The input of the intelligent model makes each artificial intelligence model execute in sequence, and finally realizes the combined task, which is the total task completed based on the specified tasks realized by each artificial intelligence model in the directed acyclic graph.

Optionally, artificial intelligence models are grouped according to model attribute information and model attribute data, including:

If it is determined that the sum of the model attribute data of two or more artificial intelligence models does not exceed the preset computing power threshold, then divide the two or more artificial intelligence models into a test group;

If it is determined that the sum of the model attribute data of the first artificial intelligence model and the model attribute data of other artificial intelligence models in the smart car operating system exceeds the preset computing power, the first artificial intelligence model is divided into a test group, wherein, The first artificial intelligence model is an artificial intelligence model in the smart car operating system.

If the sum of the computing power consumed by Model 5 and Model 6 exceeds the preset computing power threshold M3, set Model 5 as a test group and Model 6 as a test group;

If the sum of the computing power consumed by model 5 and model 6 does not exceed the computing power preset M3, then set model 5 and model 6 together as a test group.

If it is determined that there is no directed acyclic graph in the smart car operating system, then the artificial intelligence models are grouped according to the model attribute information model attribute data, and the model 1, model 2, model 3, model 4, model 5 and model 6 are realized Perform grouping to obtain at least one test group.

Enter the preset test instance into the artificial intelligence model in each static sample, and run the test instance in the artificial intelligence model through the worker thread assigned to the artificial intelligence model in each static sample to test each static sample Performance testing is performed, and at least one static performance index corresponding to at least one static strategy is obtained, wherein the test instance is a test case for performance testing of the artificial intelligence model.

In a preferred embodiment, identifying priority static indicators among at least one static performance indicator comprises:

Extract the first index element in each static performance index, sort the first index elements to obtain the target sequence, wherein, there is at least one static index element in the static performance index, and the static index element reflects the performance of the artificial intelligence model in the performance test. For performance in a performance dimension, the first index element is a static index element in the static performance index.

Exemplarily, the index elements in the static performance index include: running time, memory occupation, SM utilization rate, and power consumption.

Assuming that three static performance indicators are generated based on three static strategies: static performance indicator 1, static performance indicator 2, static performance indicator 3, and the running time of the first indicator element, then the target sequence will be obtained:

target sequence at run time operation hours Static performance index 1 0.1s Static performance index 2 0.2s Static performance index 3 0.3s

According to the position of each first index element in the target sequence, the performance value of each first index element is determined, wherein the performance value reflects the degree of performance of the first index element.

In this embodiment, the higher the rank, the better the performance of the first index element, and the lower the rank, the worse the performance of the first index element. Therefore, the target sequence of the running time is arranged in ascending order, which indicates that the running time is shorter , the better the performance of the index element; the target sequence of memory occupation is in descending order, which indicates that the smaller the memory occupation, the better the performance of the index element; the target sequence of SM utilization is in descending order, which indicates that the higher the SM occupancy rate, the better the performance of the index element. The better; the target sequence of power consumption is in ascending order, and the smaller the power consumption, the better the performance of the index element.

According to the performance value of each static index element in each static performance index, the comprehensive performance value of each static performance index is obtained.

Exemplarily, based on the above example: assuming that the performance value of the first rank is 3, the performance value of the second rank is 2, and the performance value of the third rank is 1, the performance value of each static index element in each performance index is brought into In the preset weighting function, calculate the weighting function to obtain the comprehensive performance value.

The weighting function is: S=a*x+b*y+c*z+d*m

Among them, x is the performance value of running time, a is the weight of running time; y is the performance value of memory usage, b is the weight of memory usage; z is the performance value of SM occupancy rate, c is the weight of SM occupancy rate; m is the performance value of power consumption, d is the weight of power consumption; S is the comprehensive performance value.

The static performance index with the highest comprehensive performance value is set as the priority static index.

S202: Allocate graphics processor resources for each artificial intelligence model in the smart car operating system according to the static allocation policy.

In this step, the graphics processor resources are allocated to each artificial intelligence model through the static allocation strategy corresponding to the priority static index, so as to realize that each artificial intelligence model has been allocated with the optimal performance configuration before the operation of the intelligent vehicle operating system. Graphics processor resources.

In a preferred embodiment, according to the static allocation strategy, allocate graphics processor resources for each artificial intelligence model in the intelligent vehicle operating system, including:

According to the division rules in the static allocation strategy, the artificial intelligence models in the smart car operating system are grouped to obtain at least one operation group;

Obtain at least one worker thread in the graphics processor, and assign a worker thread to each operation group according to the resource allocation rules in the static allocation strategy, so as to allocate a graphics processor to each artificial intelligence model resource in each operation group resources, where the worker thread is a sequence for invoking graphics processor resources in the graphics processor to run the artificial intelligence model.

S203: Run at least one artificial intelligence model in the smart car operating system, and monitor the dynamic performance index of the running artificial intelligence model, wherein the dynamic performance index reflects the performance of the artificial intelligence model when calling the allocated graphics processor resources for calculation performance.

In this step, monitor the dynamic performance indicators of the running artificial intelligence model through the performance acquisition module. The indicator elements in the dynamic performance indicators include: CPU usage, memory usage, disk IO, system average load, delay, and frame rate.

Use kylinTOP testing and monitoring platform, or LoadRunner, or kylinPET, or Apache JMeter, or NeoLoad, or WebLOAD, or Loadster, or Loadstorm, or Load impact, or OpenSTA, or Telegraf as the performance acquisition module.

S204: Generate a dynamic allocation strategy according to the dynamic performance index, and adjust the graphics processor resources allocated to each artificial intelligence model according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used to optimize the running performance of the artificial intelligence model according to the performance of the artificial intelligence model at runtime. AI models allocate GPU resources.

In this step, a dynamic allocation strategy is generated according to the dynamic performance index, so as to allocate graphics processor resources to the AI model based on the performance of the AI model currently running, and ensure that each running AI model has enough graphics processing Server resources are used to call, which solves the conflict between the reasoning requests generated by the artificial intelligence model and the calling of the working thread of the graphics processor, avoids the situation where multiple requests compete for the working thread of the graphics processor at the same time, and ensures the automatic driving scene The performance of multiple artificial intelligence models in .

In a preferred embodiment, generating a dynamic allocation strategy according to a dynamic performance index includes:

Extract the second index element in the dynamic performance index, and obtain the index rule corresponding to the second index element, wherein the dynamic performance index has at least one index element, the second index element is an index element in the dynamic performance index, and the index Rules are computer rules that define what is normal and abnormal for an indicator element.

If it is determined that the second index element conforms to the index rule, the second index element is set as a normal index element.

If it is determined that the second index element does not conform to the index rule, the second index element is set as an abnormal index element.

If it is determined that the number of normal index elements in the dynamic performance index does not reach the preset normal threshold, or the number of abnormal index elements reaches the preset abnormal threshold, then it is determined that the dynamic performance index is an abnormal performance index.

If it is determined that the number of normal index elements in the dynamic performance index reaches a preset normal threshold, or the number of abnormal index elements does not reach a preset abnormal threshold, then it is determined that the dynamic performance index is a normal performance index.

Generate dynamic allocation policies based on normal performance metrics and abnormal performance metrics.

Exemplarily, the obtained dynamic performance indicators include: CPU usage, memory usage, disk IO, system average load, delay, and frame rate, assuming that the second indicator element is the system average load, wherein the system average load is used to describe Graphics processor usage. When the system average load is equal to 1.0, it means that the graphics processor usage is the highest; when the system average load is less than 1.0, it means that the graphics processor usage is idle; when the system average load is greater than 1.0, it means Graphics processor usage has been exceeded.

The index rule of hypothetical load is: if the average system load belongs to [0,1], it is determined that the average system load is a normal index element; if the average system load does not belong to [0,1], it is determined that the average system load is an abnormal index element.

Assume that there are 6 dynamic performance indicators, the normal threshold is 4, and the abnormal threshold is 3.

If the dynamic performance index has 5 normal index elements and 1 abnormal index element, it is determined that the dynamic performance index is a normal performance index.

If the dynamic performance index has 2 normal index elements and 4 abnormal index elements, it is determined that the dynamic performance index is an abnormal performance index.

Specifically, a dynamic allocation strategy is generated according to normal performance indicators and abnormal performance indicators, including:

Set the artificial intelligence model corresponding to the normal performance index as the normal model, set the artificial intelligence model corresponding to the abnormal performance index as the abnormal model, and set the operating group where the normal model is located and the operating group where the artificial intelligence model that is not running in the smart car operating system is located. The group is set as the normal group, and the operation group where the abnormal model is located is the abnormal group;

If it is determined that there is a logical relationship between the abnormal model and other artificial intelligence models in the abnormal group; then adjust the division rules in the static allocation strategy or the dynamic allocation strategy, so that the adjusted division rules are used to adjust the independent models in the abnormal group to A normal group; and/or adjust the resource allocation rules in the static allocation policy or the dynamic allocation policy, so that the adjusted resource allocation rules are used to adjust the worker threads corresponding to the abnormal group to the worker threads corresponding to a normal group; wherein, An independent model is an artificial intelligence model that has no logical relationship with other artificial intelligence models in the abnormal group;

If it is determined that there is no logical relationship between the abnormal model and other artificial intelligence models in the abnormal group; adjust the division rules in the static allocation strategy or the dynamic allocation strategy, so that the adjusted division rules are used to adjust the abnormal model to a normal group ; and/or adjust the resource allocation rules in the static allocation strategy or the dynamic allocation strategy, so that the adjusted resource allocation rules are used to adjust the worker threads corresponding to the abnormal group to the worker threads corresponding to a normal group;

Generate a dynamic allocation policy according to the adjusted division rule and/or the adjusted resource allocation rule.

Exemplarily, the exception model running in the exception group is model 1, and the exception group includes: model 1, model 2, and model 3, assuming that there is a logical relationship between model 1 and model 2, such as: dependency relationship, association relationship, aggregation relationship, combination relationship, etc. Therefore, Model 3 at this time is an artificial intelligence model with independent relationships.

Assume that the working thread corresponding to the abnormal group is thread 1; the normal group 1 includes: model 4 and model 5, and the corresponding working thread is thread 2; the normal group 2 includes: model 6, and the corresponding working thread is thread 3, among which, Model 6 is a non-running AI model. Model 3 can be tuned to Normal Group 1 or Normal Group 2, and Thread 2 or Thread 3 can be assigned to the Abnormal Group.

The abnormal group includes: model 1, model 2 and model 3, and there is no logical relationship among them. Assume that the abnormal model is model 1. Based on the above example, model 2 and/or model 3 can be adjusted to normal group 1 or normal Group 2, Thread 2 or Thread 3 can also be assigned to exception groups.

Example 2:

Referring to Fig. 3, the present application provides a resource scheduling device 3 of a graphic processor of an intelligent vehicle operating system, including:

The static test module 31 is used to perform at least one performance test on the artificial intelligence model in the smart car operating system to obtain at least one static performance index; The strategy is set as a static allocation strategy, wherein the smart car operating system has at least one artificial intelligence model, the artificial intelligence model is used to achieve the specified task, the specified task is used to realize the automatic driving of the car, and the static performance index reflects the performance of the artificial intelligence model. The performance in the test, the priority static index is the static performance index that meets the preset priority rules, and the static strategy is a computer strategy for grouping artificial intelligence models and assigning work threads to the grouped artificial intelligence models. is a sequential GPU for scheduling stream processors and/or compute units in the GPU;

The static allocation module 32 is used to allocate graphics processor resources for each artificial intelligence model in the smart car operating system according to the static allocation strategy;

The dynamic monitoring module 33 is used to run at least one artificial intelligence model in the smart car operating system, and monitor the dynamic performance index of the artificial intelligence model running, wherein the dynamic performance index reflects the graphics processor allocated to the artificial intelligence model when calling The performance of resources when computing;

The dynamic allocation module 34 is used to generate a dynamic allocation strategy according to the dynamic performance index, and adjust the graphics processor resources allocated to each artificial intelligence model according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used to determine the performance of the artificial intelligence model at runtime Performance, allocating GPU resources to running AI models.

Example 3:

To achieve the above purpose, the present application also provides a computer device 4, including: a processor 42 and a memory 41 communicatively connected to the processor 42; the memory stores computer-executed instructions;

The processor executes the computer-executed instructions stored in the memory 41 to implement the resource scheduling method of the above-mentioned graphics processor, wherein the components of the resource scheduling device of the graphics processor can be dispersed in different computer devices, and the computer device 4 can be executed Smartphones, tablet computers, laptops, desktop computers, rack servers, blade servers, tower servers, or cabinet servers (including independent servers, or server clusters composed of multiple application servers), etc. The computer device in this embodiment at least includes but is not limited to: a memory 41 and a processor 42 that can be communicatively connected to each other through a system bus, as shown in FIG. 4 . It should be noted that FIG. 4 only shows a computer device with components - but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead. In this embodiment, the memory 41 (that is, a readable storage medium) includes a flash memory, a hard disk, a multimedia card, a card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Programmable Read Only Memory (PROM), Magnetic Memory, Magnetic Disk, Optical Disk, etc. In some embodiments, the memory 41 may be an internal storage unit of a computer device, such as a hard disk or internal memory of the computer device. In some other embodiments, the memory 41 can also be an external storage device of the computer equipment, such as a plug-in hard disk equipped on the computer equipment, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card , Flash Card (Flash Card) and so on. Of course, the memory 41 may also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed in the computer equipment, such as the program code of the resource scheduling device of the graphics processor in the third embodiment. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output. The processor 42 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device. In this embodiment, the processor 42 is used to run program codes stored in the memory 41 or process data, for example, to run a resource scheduling device for a graphics processor, so as to implement the resource scheduling method for a graphics processor in the above embodiment.

The above-mentioned integrated modules implemented in the form of software function modules can be stored in a computer-readable storage medium. The above-mentioned software function modules are stored in a storage medium, and include several instructions to make a computer device (which may be a personal computer, server, or network device, etc.) or a processor execute some steps of the methods in various embodiments of the present application. It should be understood that the above-mentioned processor may be a central processing unit (Central Processing Unit, referred to as CPU), and may also be other general-purpose processors, a digital signal processor (Digital Signal Processor, referred to as DSP), an application specific integrated circuit (Application Specific Integrated Circuit, referred to as ASIC) and so on. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like. The steps of the method disclosed in conjunction with the application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor. The storage may include a high-speed RAM memory, and may also include a non-volatile storage NVM, such as at least one disk storage, and may also be a U disk, a mobile hard disk, a read-only memory, a magnetic disk, or an optical disk.

To achieve the above object, the present application also provides a computer-readable storage medium, such as flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory ( SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, server, App application store, etc., on which storage The computer executes instructions, and the program realizes corresponding functions when executed by the processor 42 . The computer-readable storage medium in this embodiment is used to store computer-executable instructions for implementing the resource scheduling method for a graphics processor, and when executed by the processor 42, implements the resource scheduling method for a graphics processor in the above-mentioned embodiment.

The above-mentioned storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable In addition to programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. A storage media may be any available media that can be accessed by a general purpose or special purpose computer.

An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. Of course, the storage medium may also be a component of the processor. The processor and the storage medium may be located in application specific integrated circuits (Application Specific Integrated Circuits, ASIC for short). Of course, the processor and the storage medium can also exist in the electronic device or the main control device as discrete components.

The present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, the above resource scheduling method for a graphics processor is implemented.

It should be noted that, in this document, the term "comprising", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, It also includes other elements not expressly listed, or elements inherent in the process, method, article, or device. Without further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising that element.

Other embodiments of the present application will be readily apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any modification, use or adaptation of the application, these modifications, uses or adaptations follow the general principles of the application and include common knowledge or conventional technical means in the technical field not disclosed in the application . The specification and examples are to be considered exemplary only, with a true scope and spirit of the application indicated by the following claims.

It should be understood that the present application is not limited to the precise constructions which have been described above and shown in the accompanying drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims (10)

1. A method for scheduling resources of a graphics processor of an intelligent automobile operating system, comprising:

performing at least one performance test on an artificial intelligent model in an intelligent automobile operating system to obtain at least one static performance index; the method comprises the steps of identifying a priority static index in at least one static performance index, setting a static strategy corresponding to the priority static index as a static allocation strategy, wherein an intelligent automobile operating system is provided with at least one artificial intelligent model, the artificial intelligent model is used for realizing a specified task, the specified task is used for realizing automatic driving of an automobile, the static performance index reflects performance of the artificial intelligent model in a performance test, the priority static index is a static performance index meeting a preset priority rule, the static strategy is a computer strategy used for grouping the artificial intelligent model, and allocating working threads to the grouped artificial intelligent model, and the working threads are sequences used for scheduling a flow processor and/or a computing unit in a graphic processor;

Distributing graphics processor resources for each artificial intelligent model in the intelligent automobile operating system according to the static distribution strategy;

running at least one artificial intelligent model in the intelligent automobile operating system, and monitoring dynamic performance indexes of the running artificial intelligent model, wherein the dynamic performance indexes reflect performance of the artificial intelligent model when the distributed graphic processor resources are called for operation;

generating a dynamic allocation strategy according to the dynamic performance index, and adjusting the graphics processor resources allocated to each artificial intelligent model according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used for allocating the graphics processor resources to the running artificial intelligent model according to the performance of the artificial intelligent model during running.

2. The resource scheduling method of a graphic processor according to claim 1, wherein performing at least one performance test on an artificial intelligent model in an intelligent automobile operating system to obtain at least one static performance index comprises:

acquiring at least one static strategy, respectively grouping artificial intelligent models in the intelligent automobile operating system according to each static strategy, and distributing working threads to each group of artificial intelligent models to obtain at least one static sample;

Recording preset test cases into artificial intelligent models in each static sample, and running the test cases in the artificial intelligent models through working threads distributed by the artificial intelligent models in each static sample so as to perform performance test on each static sample and obtain at least one static performance index corresponding to at least one static strategy, wherein the test cases are test cases for performing performance test on the artificial intelligent models.

3. The resource scheduling method of a graphic processor according to claim 2, wherein grouping the artificial intelligent models in the intelligent automobile operating system and assigning a work thread to each group of artificial intelligent models according to each static policy, respectively, to obtain at least one static sample, comprises:

grouping each artificial intelligent model in the intelligent automobile operating system according to the partitioning rule in the static strategy to obtain at least one test group, wherein the test group is provided with at least one artificial intelligent model;

acquiring at least one working thread in the graphic processor, and distributing one working thread to each test group according to a resource distribution rule in the static strategy;

And summarizing at least one test group and the working threads corresponding to each test group to form a static sample corresponding to the static strategy.

4. A method for scheduling resources of a graphics processor according to claim 3, wherein grouping each artificial intelligence model in an intelligent automobile operating system according to a partitioning rule in the static policy, to obtain at least one test group, comprises:

if at least one directed acyclic graph is determined to exist in the intelligent automobile operating system, dividing the artificial intelligent models belonging to the same directed acyclic graph into a test group, wherein the directed acyclic graph reflects the logic relationship between two or more artificial intelligent models in the intelligent automobile operating system;

if it is determined that other artificial intelligence models which do not belong to the directed acyclic graph exist in the intelligent automobile operating system, grouping the other artificial intelligence models according to model attribute data of each other artificial intelligence model to obtain at least one test group, wherein the model attribute data describe calculation power consumed by the artificial intelligence model for realizing a specified task;

And if the intelligent automobile operation system is determined to not have the directed acyclic graph, grouping the artificial intelligent models in the intelligent automobile operation system according to the model attribute data of each artificial intelligent model to obtain at least one test group.

5. The resource scheduling method of a graphics processor of claim 1, wherein identifying a priority static indicator of the at least one static performance indicator comprises:

extracting a first index element in each static performance index, and sequencing the first index element to obtain a target sequence, wherein the static performance index is provided with at least one static index element, the static index element reflects the performance of an artificial intelligent model in one performance dimension in a performance test, and the first index element is one static index element in the static performance index;

determining a performance value of each first index element according to the rank of each first index element in the target sequence, wherein the performance value reflects the performance quality degree of the first index element;

obtaining the comprehensive performance value of each static performance index according to the performance value of each static index element in each static performance index;

And setting the static performance index with the highest comprehensive performance value as a priority static index.

6. The method of resource scheduling for a graphics processor of claim 1, wherein allocating graphics processor resources for each artificial intelligence model in the intelligent vehicle operating system according to the static allocation policy comprises:

grouping the artificial intelligent models in the intelligent automobile operating system according to the partitioning rules in the static allocation strategy to obtain at least one running group;

and acquiring at least one working thread in the graphic processor, and distributing one working thread to each running group according to the resource distribution rule in the static distribution strategy so as to distribute the graphic processor resource to each artificial intelligent model resource in each running group.

7. The resource scheduling method of a graphics processor according to claim 6, wherein generating a dynamic allocation policy according to the dynamic performance index comprises:

extracting a second index element in the dynamic performance index, and acquiring an index rule corresponding to the second index element, wherein the dynamic performance index is provided with at least one index element, the second index element is one index element in the dynamic performance index, and the index rule is a computer rule for defining the normal and abnormal index elements;

If the second index element is determined to accord with the index rule, setting the second index element as a normal index element;

if the second index element is determined not to accord with the index rule, setting the second index element as an abnormal index element;

if the number of the normal index elements in the dynamic performance index is not up to the preset normal threshold value or the number of the abnormal index elements is up to the preset abnormal threshold value, determining the dynamic performance index as an abnormal performance index;

if the number of the normal index elements in the dynamic performance index is determined to reach a preset normal threshold value or the number of the abnormal index elements is determined to not reach a preset abnormal threshold value, determining the dynamic performance index as a normal performance index;

and generating a dynamic allocation strategy according to the normal performance index and the abnormal performance index.

8. The resource scheduling method of a graphic processor according to claim 7, wherein generating a dynamic allocation policy according to the normal performance index and the abnormal performance index comprises:

setting an artificial intelligent model corresponding to the normal performance index as a normal model, setting an artificial intelligent model corresponding to the abnormal performance index as an abnormal model, setting an operation group in which the normal model is located and an operation group in which an artificial intelligent model which does not operate in the intelligent automobile operating system is located as a normal group, and setting the operation group in which the abnormal model is located as an abnormal group;

If the abnormal model and other artificial intelligent models in the abnormal group are determined to have a logic relationship; adjusting the partitioning rule in the static partitioning strategy or the dynamic partitioning strategy to enable the adjusted partitioning rule to be used for adjusting the independent model in the abnormal group to a normal group; and/or adjusting the resource allocation rule in the static allocation policy or the dynamic allocation policy, so that the adjusted resource allocation rule is used for adjusting the working thread corresponding to the abnormal group into a working thread corresponding to the normal group; wherein the independent model is an artificial intelligent model which has no logical relationship with other artificial intelligent models in the anomaly group;

if it is determined that the abnormal model and other artificial intelligence models in the abnormal group do not have a logic relationship; adjusting the partitioning rule in the static allocation strategy or the dynamic allocation strategy to enable the adjusted partitioning rule to be used for adjusting the abnormal model to a normal group; and/or adjusting the resource allocation rule in the static allocation policy or the dynamic allocation policy, so that the adjusted resource allocation rule is used for adjusting the working thread corresponding to the abnormal group into a working thread corresponding to the normal group;

And generating a dynamic allocation strategy according to the adjusted partitioning rule and/or the adjusted resource allocation rule.

9. A resource scheduling device for a graphics processor of an intelligent automobile operating system, comprising:

the static test module is used for performing at least one performance test on the artificial intelligent model in the intelligent automobile operating system to obtain at least one static performance index; the method comprises the steps of identifying a priority static index in at least one static performance index, setting a static strategy corresponding to the priority static index as a static allocation strategy, wherein an intelligent automobile operating system is provided with at least one artificial intelligent model, the artificial intelligent model is used for realizing a specified task, the specified task is used for realizing automatic driving of an automobile, the static performance index reflects performance of the artificial intelligent model in a performance test, the priority static index is a static performance index meeting a preset priority rule, the static strategy is a computer strategy used for grouping the artificial intelligent model, and allocating working threads to the grouped artificial intelligent model, and the working threads are sequence graphic processors used for scheduling a flow processor and/or a computing unit in a graphic processor;

The static allocation module is used for allocating graphics processor resources for each artificial intelligent model in the intelligent automobile operating system according to the static allocation strategy;

the dynamic monitoring module is used for running at least one artificial intelligent model in the intelligent automobile operating system and monitoring dynamic performance indexes of the running artificial intelligent model, wherein the dynamic performance indexes reflect performance of the artificial intelligent model when the distributed graphic processor resource is called for operation;

and the dynamic allocation module is used for generating a dynamic allocation strategy according to the dynamic performance index, and adjusting the graphics processor resources allocated to each artificial intelligent model according to the dynamic allocation strategy, wherein the dynamic allocation strategy is used for allocating the graphics processor resources to the running artificial intelligent model according to the performance of the artificial intelligent model during running.

10. A computer device, comprising: a processor and a memory communicatively coupled to the processor;

the memory stores computer-executable instructions;

the processor executes computer-executable instructions stored in the memory to implement a resource scheduling method for a graphics processor as claimed in any one of claims 1 to 8.

Images