GPU application system and application method thereof, electronic terminal and storage medium

Abstract

The invention provides a GPU application system and an application method thereof, an electronic terminal and a storage medium. The GPU application system comprises a physical machine and an acceleration card, the physical machine is connected with the acceleration card; the GPU module at least comprises a first GPU resource pool deployed on a physical machine and a second GPU resource pool deployed on an accelerator card; the first GPU resource pool comprises at least one container, and the second GPU resource pool comprises at least one business process; and the regulation and control module is used for determining a matched target accelerator card according to the configuration parameters of the to-be-processed task in the target container, so that the target accelerator card processes the to-be-processed task. According to the method, the GPU module is deployed on the physical machine and the acceleration card, so that the container in the first GPU resource pool deployed on the physical machine can occupy CPU and memory resources on the physical machine, and meanwhile, the service process in the second GPU resource pool deployed on the acceleration card can call AI resources on the acceleration card; and therefore, AI resources, memory resources and CPU computing power on the acceleration card can be effectively utilized.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a GPU application system and its application method, an electronic terminal, and a storage medium. Background Technology

[0002] In the era of artificial intelligence, the application of Graphics Processing Units (GPUs) is becoming increasingly widespread. A GPU is a microprocessor that performs image processing tasks in personal computers, workstations, game consoles, and some mobile devices such as tablets and smartphones.

[0003] Although GPU service nodes have good processing performance in the field of image processing, due to the high cost of using GPU service nodes, they are usually not deployed in large numbers. On the other hand, using CPU (Central Processing Unit) service nodes for computing cannot meet the computing performance requirements of large-scale data computing. Summary of the Invention

[0004] The main technical problem addressed by this application is to provide a GPU application system and its application method, electronic terminal, and storage medium, thereby solving the problem in the prior art that GPU application systems cannot effectively utilize the resources on the accelerator card.

[0005] To solve the above-mentioned technical problems, the first technical solution adopted in this application is: to provide a GPU application system, the GPU application system comprising: Physical machine and accelerator card; connection between physical machine and accelerator card; The GPU module includes at least a first GPU resource pool deployed on a physical machine and a second GPU resource pool deployed on an accelerator card; the first GPU resource pool includes at least one container, which encapsulates tasks to be processed; each task to be processed has configuration parameters; the second GPU resource pool includes at least one business process. The control module, connected to the physical machine and accelerator card, is used to determine the matching target accelerator card based on the configuration parameters of the task to be processed in the target container, so that the target accelerator card can process the task to be processed.

[0006] The control module is also used to control the sending of pending tasks and program information from the target container to the matching target accelerator card, so that the target accelerator card can establish a business process on the accelerator card based on the received pending tasks and program information, and process the pending tasks based on the program information in the business process.

[0007] The configuration parameters of the task to be processed include computing power requirements. The control module is also used to compare the computing power requirements of the task to be processed in the target container with the remaining resources and performance of the accelerator card, and to select the accelerator card with remaining resources greater than computing power requirements as the target accelerator card.

[0008] The configuration parameters for the task to be processed also include the task type. In response to the fact that the remaining resources of at least two accelerator cards are greater than the computing power requirement of the task to be processed, the control module is also used to select the accelerator card whose remaining resources are greater than the computing power requirement of the task to be processed and whose task type is inconsistent with the task to be processed as the target accelerator card to match the target container.

[0009] If there are at least two accelerator cards that respond to the computing power requirement of the task to be processed if the remaining resources are greater than the remaining resources, the control module will use the accelerator card with the fewest remaining resources as the target accelerator card for the target container.

[0010] The control module is also used to determine that each accelerator card is connected to the physical machine through a different preset connection method when at least two accelerator cards are connected to the same physical machine; the preset connection methods include PCIe channel and socket channel.

[0011] The control module is also used to determine the server type corresponding to each business process when the target acceleration card contains at least two business processes.

[0012] In response to the completion of pending tasks based on program information during the business process, the control module is used to control the acceleration card to save the pending tasks and the corresponding processing results.

[0013] To solve the above-mentioned technical problems, the second technical solution adopted in this application is: to provide an application method for a GPU application system, wherein the GPU application system is the same as the GPU application system in the first technical solution, and the application method includes: The physical machine determines the program information for processing the tasks based on the configuration parameters of the tasks to be processed in the target container. The task to be processed and program information are packaged and sent to the matching accelerator card, so that the accelerator card can process the task to be processed based on the program information and send the processing result of the task to the physical machine. The physical machine receives the processing results of the tasks to be processed.

[0014] To address the aforementioned technical problems, the third technical solution adopted in this application is: to provide an application method for a GPU application system, wherein the GPU application system is as described in the first technical solution, and the application method includes: The accelerator card constructs a business process based on the received pending tasks and program information; The task to be processed is processed based on the program information in the business process to obtain the processing result of the task to be processed. Send the physical results of the task to be processed to the physical machine.

[0015] Specifically, the processing of tasks based on program information within the business process yields processing results, including: In response to the accelerator card receiving packaged pending tasks and program information, it stops running and deletes the historical program corresponding to the business process, and enables the processing of pending tasks using the program information.

[0016] To solve the above-mentioned technical problems, the fourth technical solution adopted in this application is: to provide an electronic terminal, characterized in that the electronic terminal includes a memory and a processor coupled to each other, the processor is used to execute program instructions stored in the memory, and the processor is used to execute program data to implement the steps in the application method of the GPU application system as described above.

[0017] To solve the above-mentioned technical problems, the fifth technical solution adopted in this application is: to provide a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the steps in the application method of the GPU application system as described above.

[0018] The beneficial effects of this application are as follows: Unlike existing technologies, this application provides a GPU application system and its application method, electronic terminal, and storage medium. The GPU application system includes: a physical machine and an accelerator card; the physical machine and the accelerator card are connected; a GPU module, including at least a first GPU resource pool deployed on the physical machine and a second GPU resource pool deployed on the accelerator card; the first GPU resource pool includes at least one container, which encapsulates tasks to be processed; each task to be processed has configuration parameters; the second GPU resource pool includes at least one business process; and a control module, connected to the physical machine and the accelerator card, is used to determine a matching target accelerator card based on the configuration parameters of the tasks to be processed in the target container, so that the target accelerator card can process the tasks to be processed. This application, by deploying the GPU module on the physical machine and the accelerator card, allows the containers in the first GPU resource pool deployed on the physical machine to occupy the CPU and memory resources of the physical machine, while the business processes in the second GPU resource pool deployed on the accelerator card can call the AI ​​resources on the accelerator card, thereby effectively utilizing the AI ​​resources, memory resources, and CPU computing power of the accelerator card. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the framework structure of an embodiment of the GPU application system provided in this application; Figure 2 This is a flowchart illustrating an embodiment of the application method of the GPU application system provided in this application; Figure 3 This is a flowchart illustrating another embodiment of the application method of the GPU application system provided in this application; Figure 4 This is a schematic diagram of the framework of an embodiment of the terminal provided in this application; Figure 5 This is a schematic diagram of a framework of an embodiment of the computer-readable storage medium provided in this application.

[0021] In the diagram: GPU application system 100; physical machine 1; accelerator card 2; GPU module 3; first GPU resource pool 31; container 311; second GPU resource pool 32; business process 321; control module 4. Detailed Implementation

[0022] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0023] In the following description, specific details such as particular system architectures, interfaces, and technologies are presented for illustrative purposes rather than for limiting purposes, in order to provide a thorough understanding of this application.

[0024] In this article, the term "and / or" simply describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. Additionally, the character " / " generally indicates that the preceding and following related objects have an "or" relationship. Furthermore, "more" in this article means two or more objects.

[0025] To enable those skilled in the art to better understand the technical solutions of this application, the following describes in further detail a GPU application system and its application method, electronic terminal, and storage medium provided in this application, in conjunction with the accompanying drawings and specific embodiments.

[0026] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0027] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0028] Please see Figure 1 , Figure 1 This is a schematic diagram of the framework structure of an embodiment of the GPU application system provided in this application.

[0029] This embodiment provides a GPU application system 100, which includes a physical machine 1, an accelerator card 2, a GPU module 3, and a control module 4. The physical machine 1 is connected to the accelerator card 2; the control module 4 is connected to both the physical machine 1 and the accelerator card 2. The accelerator card 2 provides network function acceleration for the physical machine 1.

[0030] In one embodiment, the GPU module 3 includes at least a first GPU resource pool 31 deployed on the physical machine 1 and a second GPU resource pool 32 deployed on the accelerator card 2.

[0031] The first GPU resource pool 31 includes at least one container 311. That is, the first GPU resource pool 31 may include one, two, or three containers 311, etc. The three containers 311 can be deployed on the same physical machine 1 or on different physical machines 1. Each container 311 encapsulates a task to be processed; each task has configuration parameters. Container 311 refers to the GPU application front-end deployed on physical machine 1 and managed using container 311 technology, primarily responsible for base library management, task scheduling, and communication with the accelerator card 2. For example, the task to be processed can be a feature comparison application, a feature extraction application, etc., specifically configured according to the actual situation. The configuration parameters of the task to be processed include computing power requirements and task type.

[0032] In this embodiment, the first GPU resource pool 31 manages the CPU and memory resources of the physical machine 1 containing the container 311.

[0033] The second GPU resource pool 32 includes at least one business process 321. That is, the second GPU resource pool 32 may include one, two, or three business processes 321, etc. The business process 321 refers to the GPU application backend deployed on the accelerator card 2 and running as a normal process, primarily responsible for large-scale database storage and actual AI computing tasks. The three business processes 321 can be contained on the same accelerator card 2 or on different accelerator cards 2.

[0034] In one embodiment, the accelerator card 2 may not contain the service process 321, that is, the remaining resources of the accelerator card 2 are equal to the initial resources of the accelerator card 2.

[0035] In this embodiment, the service process 321 deployed on the accelerator card 2 uses the AI ​​resources on the accelerator card 2.

[0036] The control module 4 is used to determine the matching target accelerator card 2 according to the configuration parameters of the task to be processed in the target container 311, so that the target accelerator card 2 can process the task to be processed.

[0037] This application deploys GPU module 3 on physical machine 1 and accelerator card 2, so that container 311 in the first GPU resource pool 31 deployed on physical machine 1 can occupy CPU and memory resources on physical machine 1, while business process 321 in the second GPU resource pool 32 deployed on accelerator card 2 can call AI resources on accelerator card 2, thereby effectively utilizing AI resources, memory resources and CPU computing power on accelerator card 2.

[0038] In one embodiment, the control module 4 is further configured to control the transmission of the pending tasks and program information in the target container 311 to the matched target accelerator card 2, so that the target accelerator card 2 establishes a business process 321 on the accelerator card 2 based on the received pending tasks and program information, and processes the pending tasks in the business process 321 based on the program information. By processing the pending tasks based on the AI ​​resources, memory resources, and CPU computing power on the accelerator card 2 using the program information, the effective utilization of the AI ​​resources, memory resources, and CPU computing power on the accelerator card 2 is achieved.

[0039] In one specific embodiment, the configuration parameters of the task to be processed include computing power requirements. The control module 4 is also used to compare the computing power requirements of the task to be processed in the target container 311 with the remaining resources and performance of the accelerator card 2, and to select the accelerator card 2 with remaining resources greater than computing power requirements as the target accelerator card 2.

[0040] Furthermore, in one embodiment, the configuration parameters of the task to be processed also include the task type. In response to the situation where the remaining resources of at least two accelerator cards 2 are greater than the computing power requirement of the task to be processed, the control module 4 is further configured to select an accelerator card 2 whose remaining resources are greater than the computing power requirement of the task to be processed, and whose task type is inconsistent with the task type of the task to be processed, as the target accelerator card 2 to match the target container 311. By deploying tasks of different task types on the same accelerator card 2, the utilization of AI resources on the accelerator card 2 can be maximized.

[0041] In one embodiment, if there are at least two accelerator cards 2 whose remaining resources exceed the computing power requirements of the task to be processed, the control module 4 is used to select the accelerator card 2 with the fewest remaining resources as the target accelerator card 2 corresponding to the target container 311. By prioritizing the full utilization of the remaining resources on the accelerator card 2, the resource fragmentation problem of the second GPU resource pool 32 on the accelerator card 2 can be effectively improved.

[0042] In one embodiment, the control module 4 is further configured to, in response to at least two accelerator cards 2 being connected to the same physical machine 1, determine that each accelerator card 2 is connected to the physical machine 1 through a different preset connection method; the preset connection method includes a PCIe channel or a Socket channel. The PCIe channel refers to the Peripheral Component Interconnect Express (PCIE) and is used for hardware-level internal chassis communication. The Socket channel is used for software-level network communication. The Socket channel is an important communication mechanism. It allows two programs to communicate in real time on different computers, transmitting data through the creation and use of sockets.

[0043] In this embodiment, the accelerator card 2 is connected to the physical machine 1 via a PCIe channel or a Socket channel, enabling the accelerator card 2 and the physical machine 1 to work together. This allows the data of the large database to be directly stored on the accelerator card 2, solving the problem of frequent data transfer during CPU and GPU mixed computing.

[0044] In one embodiment, the control module 4 is further configured to determine that each business process 321 corresponds to a different server type when the target accelerator card 2 contains at least two business processes 321. The server type can be configured by the control module 4. For example, if one business process 321 in the target accelerator card 2 contains a "feature comparison task," then the server type used by that business process 321 is server number 14; if another business process 321 in the target accelerator card 2 contains a "feature extraction task," then the server type used by that business process 321 is server number 15.

[0045] In one embodiment, in response to the completion of a pending task based on program information in business process 321, the control module 4 controls the accelerator card 2 to save the pending task and its corresponding processing result. By saving the pending task and processing result in business process 321, data in the large database can be synchronized with data on the accelerator card 2, further solving the problem of frequent data transfer in CPU and GPU hybrid computing.

[0046] In this embodiment, by sending the tasks to be processed in the target container 311 to the connected acceleration card 2, the tasks to be processed can be flexibly deployed, and the business processes 321 on the acceleration card 2 can also be expanded and upgraded.

[0047] Please see Figure 2 , Figure 2 This is a flowchart illustrating an embodiment of the application method of the GPU application system provided in this application.

[0048] This application provides an application method for a GPU application system, such as the GPU application system in the above embodiments, wherein the execution subject of this application method is a physical machine. The application method for the GPU application system includes the following steps.

[0049] S11: The physical machine determines the program information for processing the task based on the configuration parameters of the task to be processed in the target container.

[0050] S12: Package the task to be processed and program information and send them to the matching accelerator card, so that the accelerator card can process the task to be processed based on the program information and send the processing result of the task to be processed to the physical machine.

[0051] S13: The physical machine receives the processing results of the task to be processed.

[0052] In this embodiment, by sending the tasks to be processed in the target container and the corresponding program information to the accelerator card, the AI ​​resources, memory resources and CPU computing power on the accelerator card can be effectively utilized to complete the processing of the tasks to be processed.

[0053] In one embodiment, each container in the first GPU resource pool is deployed on the same physical machine or different physical machines, so that the containers in the first GPU resource pool can occupy the CPU and memory resources on the physical machine.

[0054] In one embodiment, the process of determining the program information for processing the task based on the configuration parameters of the task to be processed in the target container in step S11 specifically includes the following implementation methods.

[0055] Specifically, in response to a user request, the physical machine launches one or more containers from the first GPU resource pool.

[0056] The configuration parameters for the task to be processed include the task type. The task type can be feature extraction, feature comparison, etc.

[0057] When the feature type of the task to be processed is feature extraction, the program information that matches the task to be processed can be executed or the feature extraction task can be completed.

[0058] When the feature type of the task to be processed is feature comparison, the program information that matches the task to be processed can be executed or the feature comparison task can be completed.

[0059] In one embodiment, the step S12 of packaging the task to be processed and program information and sending it to the matching accelerator card specifically includes the following implementation method.

[0060] The tasks to be processed in the container, along with program information and performance metrics, are packaged into an installation package and sent to the matching accelerator card. The matching accelerator card is determined based on the configuration information of the tasks to be processed, as described in the above embodiments and will not be repeated here.

[0061] In one embodiment, the physical machine receiving the processing result of the task to be processed in step S13 specifically includes the following implementation method.

[0062] Specifically, the physical machine stores the processing data of the received tasks in the large database to achieve synchronization between the large database and the data on the accelerator card.

[0063] Please see Figure 3 , Figure 3 This is a flowchart illustrating another embodiment of the application method of the GPU application system provided in this application.

[0064] This application provides an application method for a GPU application system, such as the GPU application system in the above embodiments, wherein the execution subject of this application method is an accelerator card. The application method for the GPU application system includes the following steps.

[0065] S21: The accelerator card constructs a business process based on the received pending tasks and program information.

[0066] S22: Process the task to be processed based on the program information in the business process to obtain the processing result of the task to be processed.

[0067] S23: Send the physical results of the task to be processed to the physical machine.

[0068] In one embodiment, the accelerator card can detect at preset intervals whether it has received a new installation package sent by the physical machine.

[0069] In one embodiment, the step S21 in which the accelerator card constructs a service process based on the received task to be processed and program information specifically includes the following implementation methods.

[0070] Specifically, in response to the accelerator card detecting that it has received a new installation package, the accelerator card constructs a new business process based on the pending tasks and program information in the new installation package.

[0071] In one embodiment, step S22, which processes the task to be processed based on the program information in the business process to obtain the processing result of the task to be processed, specifically includes the following implementation methods.

[0072] Specifically, in response to the accelerator card receiving the packaged pending tasks and program information, it stops running and deletes the historical program corresponding to the business process, and uses the program information to process the pending tasks.

[0073] Specifically, after deleting the historical programs on the accelerator card, the new installation package is extracted, the program information in the new installation package is installed, and the program information is launched to process the pending tasks.

[0074] In one embodiment, the accelerator card obtains the processing result of the task after completing the task to be processed.

[0075] The accelerator card transmits the processing results of the tasks to be processed to the physical machine through the corresponding channel, so that the physical machine can save the tasks and results. The accelerator card can also save the tasks and their corresponding results to achieve data synchronization.

[0076] Please see Figure 4 , Figure 4 This is a schematic diagram of a framework of an embodiment of the electronic terminal provided in this application. The electronic terminal 80 includes a memory 81 and a processor 82 coupled to each other. The processor 82 is used to execute program instructions stored in the memory 81 to implement the steps of any of the above-described GPU application system application method embodiments. In a specific implementation scenario, the electronic terminal 80 may include, but is not limited to, a microcomputer or a server. In addition, the electronic terminal 80 may also include mobile devices such as laptops and tablets, which are not limited here.

[0077] Specifically, processor 82 controls itself and memory 81 to implement the steps of any of the above-described GPU application system application method embodiments. Processor 82 can also be referred to as a CPU (Central Processing Unit). Processor 82 may be an integrated circuit chip with signal processing capabilities. Processor 82 can also be a general-purpose processor, digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor. Furthermore, processor 82 can be implemented using integrated circuit chips.

[0078] Please see Figure 5 , Figure 5 This is a schematic diagram of a framework of an embodiment of the computer-readable storage medium provided in this application. The computer-readable storage medium 90 stores program instructions 901 that can be executed by a processor. The program instructions 901 are used to implement the steps of any of the application method embodiments of the GPU application system described above.

[0079] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0080] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0081] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0082] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0083] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0084] The above are merely embodiments of this application and do not limit the scope of patent protection of this application. Any equivalent structural or procedural changes made using the content of this application’s specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of this application.

Claims

1. A GPU application system, characterized in that, The GPU application system includes: A physical machine and an accelerator card; the physical machine is connected to the accelerator card; The GPU module includes at least a first GPU resource pool deployed on the physical machine and a second GPU resource pool deployed on the accelerator card; the first GPU resource pool includes at least one container, which encapsulates tasks to be processed; each task to be processed has configuration parameters; the second GPU resource pool includes at least one business process. The control module, connected to the physical machine and the accelerator card, is used to determine the matching target accelerator card according to the configuration parameters of the task to be processed in the target container, so that the target accelerator card can process the task to be processed.

2. The GPU application system according to claim 1, characterized in that, The control module is also used to control the sending of the pending tasks and program information in the target container to the matching target acceleration card, so that the target acceleration card establishes the business process on the acceleration card based on the received pending tasks and program information, so as to process the pending tasks based on the program information in the business process.

3. The GPU application system according to claim 1, characterized in that, The configuration parameters of the task to be processed include computing power requirements. The control module is also used to compare the computing power requirements of the task to be processed in the target container with the remaining resources and performance of the accelerator card, and to select the accelerator card with the remaining resources greater than the computing power requirements as the target accelerator card.

4. The GPU application system according to claim 3, characterized in that, The configuration parameters of the task to be processed also include the task type. In response to the fact that the remaining resources of at least two of the accelerator cards are greater than the computing power requirement of the task to be processed, the control module is further configured to use the accelerator card whose remaining resources are greater than the computing power requirement of the task to be processed and whose task type is inconsistent with the task type of the task to be processed as the target accelerator card to be matched with the target container.

5. The GPU application system according to claim 3 or 4, characterized in that, If there are at least two accelerator cards whose remaining resources exceed the computing power requirement of the task to be processed, the control module is used to select the accelerator card with the fewest remaining resources as the target accelerator card for the target container.

6. The GPU application system according to claim 1, characterized in that, The control module is further configured to, in response to at least two of the accelerator cards being connected to the same physical machine, determine that each of the accelerator cards is connected to the physical machine through a different preset connection method; the preset connection method includes a PCIe channel or a Socket channel; and / or, the control module is further configured to, in response to the target accelerator card containing at least two of the service processes, determine that each of the service processes corresponds to a different server type.

7. The GPU application system according to claim 1, characterized in that, In response to the completion of the pending task based on the program information in the business process, the control module is used to control the acceleration card to save the pending task and the corresponding processing result.

8. An application method for a GPU application system, characterized in that, The GPU application system is the GPU application system as described in any one of claims 1 to 7, wherein the application method includes: The physical machine determines the program information for processing the task based on the configuration parameters of the task to be processed in the target container; The task to be processed and the program information are packaged and sent to the matching accelerator card, so that the accelerator card processes the task to be processed based on the program information and sends the processing result of the task to be processed to the physical machine. The physical machine receives the processing result of the task to be processed.

9. An application method for a GPU application system, characterized in that, The GPU application system is the GPU application system as described in any one of claims 1 to 7, wherein the application method includes: The accelerator card constructs a business process based on the received pending tasks and program information; The task to be processed is processed based on the program information in the business process to obtain the processing result of the task to be processed. The physical results of the task to be processed are sent to the physical machine.

10. The application method according to claim 9, characterized in that, The process of processing the task to be processed based on the program information in the business process to obtain the processing result of the task to be processed includes: In response to the accelerator card receiving the packaged task to be processed and the program information, it stops running and deletes the historical program corresponding to the business process, and uses the program information to process the task to be processed.

11. An electronic terminal, characterized in that, The electronic terminal includes a memory and a processor coupled to each other. The processor is used to execute program instructions stored in the memory and to execute program data to implement the steps of the application method of the GPU application system as described in any one of claims 8 to 10.

12. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the application method of the GPU application system as described in any one of claims 8 to 10.

Images