An intelligent computing resource agile dynamic allocation method, system, device and medium

Abstract

This invention belongs to the field of edge-embedded integrated core intelligent processing, and provides a method, system, device, and medium for agile dynamic allocation of intelligent computing resources. The method includes: establishing a communication link between two PCIe switching domains; writing the current resource information of the local domain into shared memory at a first predetermined period, and retrieving the current resource information of the other domain from the shared memory at a second predetermined period, wherein the first predetermined period is shorter than the second predetermined period; the PCIe switching domain dynamically allocates computing resources based on the current workload, the current resource information of the local domain, and the current resource information of the other domain, according to a mode selection logic, by selecting a computing resource allocation mode. This invention's method can flexibly allocate computing resources, enabling efficient collaboration between hardware and software, and is suitable for multi-chip, high-computing-power application scenarios.

Description

Technical Field

[0001] This invention belongs to the field of edge-embedded integrated core intelligent processing, and relates to multi-intelligent computing node scheduling technology. Specifically, it relates to a method, system, device and medium for agile dynamic allocation of intelligent computing resources based on PC Ie bus. Background Technology

[0002] With the continuous development of intelligent computing technology, its application areas are becoming increasingly broad, especially in fields such as aviation / aerospace, where it can handle tasks such as image processing and large-scale model inference. Currently, intelligent resources are typically allocated through point-to-point interconnections, which has limitations in terms of sufficient and efficient resource allocation. Summary of the Invention

[0003] To address the limitations of existing point-to-point interconnection allocation methods for intelligent resources, which result in insufficient and inefficient resource allocation, this invention discloses an agile dynamic allocation method for intelligent computing resources. The method includes the following steps:

[0004] S4. Connect two PC Ie switching domains on the user terminal and establish a communication link between the two PC Ie switching domains;

[0005] S5. The PC Ie switching domain writes its current resource information into the shared memory at a first set period, and retrieves the current resource information of the other domain from the shared memory at a second set period, wherein the first set period is shorter than the second set period.

[0006] S6. The PC Ie switching domain dynamically allocates computing resources based on the current task volume, the current resource information of the local domain, and the current resource information of the peer domain, according to the mode selection logic, by selecting a computing resource allocation mode.

[0007] Furthermore, the aforementioned agile dynamic allocation method for intelligent computing resources also includes:

[0008] S1. Construct shared memory and two PCIe switching domains, each of which includes a switching domain host. The switching domain host includes a resource planning and orchestration scheduler. One end of the switching domain host is connected to the user terminal, and the other end is connected to each of the downstream co-intelligent processors. The resource planning and orchestration scheduler is connected to the shared memory.

[0009] Furthermore, the aforementioned agile dynamic allocation method for intelligent computing resources also includes:

[0010] S2. Based on the task volume and computing resources, design the computing resource allocation mode, including the exchange domain private mode, the single cross-connection possession mode, and the cross-boundary possession mode.

[0011] S31. When the amount of tasks received by this domain is less than or equal to the computing resources of this domain, the computing resources of this domain are dynamically allocated using the private mode of the exchange domain.

[0012] S32. When the amount of tasks received by this domain is greater than the computing resources of this domain, and the computing resources of the other domain are greater than the difference between the amount of tasks and the computing resources of this domain, the computing resources of this domain and the other domain are dynamically allocated using a single-connection occupancy mode.

[0013] S33. When the local domain receives a new task while executing a task allocated by the other domain, a cross-domain occupancy mode is adopted to dynamically allocate computing resources between the local domain and the other domain, thus completing the construction of the mode selection logic.

[0014] Furthermore, in step S32 above, when the amount of tasks received by the local domain is greater than the local domain's computing resources, and the computing resources of the peer domain are greater than the difference between the amount of tasks and the computing resources of the local domain, a single-connection occupancy mode is used to dynamically allocate computing resources between the local and peer domains, including:

[0015] S321. When the amount of tasks received by this domain is greater than the amount of remaining resources in this domain, and any one of the sub-computing resources in the other domain is greater than the difference between the amount of tasks and the computing resources in this domain, the computing resources of this domain and the other domain are dynamically allocated using a single-connection occupancy mode.

[0016] S322. When the amount of tasks received by this domain is greater than the amount of remaining resources in this domain, and all sub-computing resources in the other domain are less than the difference between the amount of tasks and the computing resources in this domain, it is determined whether the task can be split. If the task can be split, the computing resources of this domain and the other domain are dynamically allocated using a single-connection occupancy mode. If the task cannot be split, the task waits.

[0017] Furthermore, in step S5 above, the PCIe switching domain writes its current resource information into shared memory at a first predetermined period, and retrieves the current resource information of the other domain from the shared memory at a second predetermined period, including:

[0018] S51. Each of the co-intelligent processors connected downstream of the PCIe switching domain reports its current usage status and current sub-computing resources according to a first set period. Based on the current sub-computing resources of all the co-intelligent processors connected downstream of the PCIe switching domain, the current total computing resources are obtained. The current total computing resources and the current usage status and current sub-computing resources of each co-intelligent processor are written into the shared memory as the current resource information of the domain.

[0019] S52. The PCI e switching domain obtains the current resource information of the other domain according to the second set period.

[0020] Furthermore, in step S6 above, the PCIe switching domain dynamically allocates computing resources based on the current workload, the current resource information of its own domain, and the current resource information of the other domain, according to the mode selection logic, by selecting a computing resource allocation mode, including:

[0021] S61. The PCI e switching domain estimates the current task volume based on the received current task, and obtains the current total computing resources and current sub-computing resources of this domain.

[0022] S62. When the current task volume of this domain is less than or equal to the current total computing resources, the current total computing resources of this domain are dynamically allocated using the private mode of the exchange domain.

[0023] S63. When the current task volume of this domain is greater than the current total computing resources, and the difference between the current task volume of this domain and the current total computing resources is less than any of the current sub-computing resources in the other domain, select the single-connection occupancy mode to dynamically allocate computing resources between this domain and the other domain.

[0024] S64. When the current task quantity of this domain is greater than the current total computing resources, and all current sub-computing resources in the other domain are less than the difference between the current task quantity and the current total computing resources of this domain, determine whether the current task of this domain can be split. If it can be split, select the single-connection occupancy mode to dynamically allocate computing resources of this domain and the other domain. If it cannot be split, wait for the task.

[0025] This invention also provides an intelligent computing resource agile dynamic allocation system, including two PCIe switching domains, a communication module, a resource write module, a resource read module, and a computing resource dynamic allocation module.

[0026] The two PC Ie switching domains are connected via NT ports. Each PC Ie switching domain includes a switching domain host and multiple co-intelligent processors. The host is equipped with a resource planning and orchestration scheduler, which is located between the co-intelligent processor and the user terminal.

[0027] The communication module is used to establish a communication link between the two PC Ie switching domains;

[0028] The resource writing module is used to write the current total computing resources of the PC Ie switching domain and the current usage status and current sub-computing resources of each of the downstream co-intelligent processors into the shared memory according to a first set period.

[0029] The resource reading module is used to obtain the current total computing resources of the other domain and the current usage status and current sub-computing resources of each downstream co-intelligent processor from the shared memory according to a second set period, wherein the first set period is shorter than the second set period.

[0030] The dynamic allocation module for computing resources is used by the PC Ie switching domain to dynamically allocate computing resources based on the current workload, the current resource information of the domain, and the current resource information of the peer domain, according to the mode selection logic, by selecting a computing resource allocation mode.

[0031] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned agile dynamic allocation method for intelligent computing resources, thereby solving the limitation that existing intelligent resources cannot be fully and effectively allocated due to the point-to-point interconnection allocation method.

[0032] This invention also provides a computer-readable storage medium storing a computer program that executes any of the above-described intelligent computing resource agile dynamic allocation methods, thereby overcoming the limitations of existing intelligent resource allocation methods that use point-to-point interconnection, which result in insufficient and ineffective resource allocation.

[0033] Compared with the prior art, the beneficial effects that the above-mentioned at least one technical solution adopted in the embodiments of this specification can achieve include at least the following: The method of the present invention constructs two PC Ie switching domains at the user end, each switching domain having a switching domain host and multiple co-intelligent processors. The two PC Ie switching domains communicate via mapped NT (non-transparent) ports. At the same time, a resource planning and orchestration scheduler is designed in each switching domain host to achieve the shielding of underlying resources. By constructing a resource exchange network and using a shared memory communication method, the underlying resource quantity value is perceived in real time. The switching domain private, single-connection occupation, cross-boundary occupation and other modes are flexibly configured. The present invention solves the problems of low computing power resources, inflexible resource scheduling, underutilization of resources and complex upper-layer application design in existing edge embedded intelligent computing centers. By constructing a shieldable resource planning and orchestration scheduler, the load balancing of computing resources in intelligent computing centers is improved, and resource scheduling is flexible. At the same time, the dual-switching domain cross-connection communication mode improves computing power configuration. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the 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.

[0035] Figure 1 This is a flowchart of an embodiment of the intelligent computing resource agile dynamic allocation method disclosed in this invention;

[0036] Figure 2 This is a schematic diagram of the agile dynamic allocation method for intelligent computing resources disclosed in an embodiment of the present invention;

[0037] Figure 3 This is a schematic diagram illustrating a computing architecture built on the PCIe bus, as disclosed in an embodiment of the present invention.

[0038] Figure 4 This is a schematic diagram illustrating the allocation of computing resources according to an embodiment of the present invention. Detailed Implementation

[0039] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0040] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features of the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0041] This invention discloses an agile dynamic allocation method for intelligent computing resources. See [link to relevant documentation]. Figure 1 and Figure 2 As shown, the method includes the following steps:

[0042] S4. Connect two PC Ie switching domains on the user terminal and establish a communication link between the two PC Ie switching domains;

[0043] S5. The PC Ie switching domain writes its current resource information into the shared memory at a first set period, and retrieves the current resource information of the other domain from the shared memory at a second set period, wherein the first set period is shorter than the second set period.

[0044] S6. The PC Ie switching domain dynamically allocates computing resources based on the current task volume, the current resource information of the local domain, and the current resource information of the peer domain, according to the mode selection logic, by selecting a computing resource allocation mode.

[0045] Furthermore, for the aforementioned agile dynamic allocation method of intelligent computing resources, see [link to relevant documentation]. Figure 1 and Figure 2 As shown, it also includes:

[0046] S1. Construct shared memory and two PCIe switching domains, each of which includes a switching domain host. The switching domain host includes a resource planning and orchestration scheduler. One end of the switching domain host is connected to the user terminal, and the other end is connected to each of the downstream co-intelligent processors. The resource planning and orchestration scheduler is connected to the shared memory.

[0047] Specifically, such as Figure 3 and Figure 4 As shown, each PCIe switching domain has one switching domain host and three co-intelligent processors (EPs). Communication between two PCIe switching domains occurs via NT ports. The switching domain host can access and invoke the co-intelligent processors across domains. Setting up a resource planning, orchestration, and scheduler in the switching domain host can shield upper-layer applications from lower-layer resources, providing only a data input interface for user applications without requiring them to concern themselves with the resource scheduling of the underlying co-intelligent processors.

[0048] Meanwhile, the switching domain host can dynamically monitor the computing resources of downstream co-intelligent processors in real time. This is achieved using shared memory communication, where a shared memory segment is allocated on each of the two switching domain hosts to the host and all EPs. Each EP writes its usage status and resource availability into the shared memory, while the host writes instructions and other information into its corresponding shared memory. For example, when each PCIe switch domain has three co-intelligent processors connected, seven shared memory segments can be configured, belonging to local node 1, local node 2, local node 3, host 2, cross-domain node 1, cross-domain node 2, and cross-domain node 3. Local nodes write the `XX_POI NT_BUSY / FREE` instruction to the corresponding shared memory area every 200ms, and the host node reads the value from the shared memory every 205ms. Receiving the `XX_POI NT_FREE` command indicates that the node's computing resources are idle. When the XX_POI NT_BUSY instruction is received, it indicates that the node's resources are being utilized. At this time, the host writes the resource quantity query instruction XX_POI NT_LEVEL to the corresponding shared memory region. After writing, it waits for the slave node to reply with an instruction, thereby obtaining the number of idle NPU cores of the slave node and sensing the progress of the underlying computing tasks in real time through shared memory.

[0049] Furthermore, for the aforementioned agile dynamic allocation method of intelligent computing resources, see [link to relevant documentation]. Figure 1 and Figure 2 As shown, it also includes:

[0050] S2. Based on the task load and computing resources, design the computing resource allocation modes, including the private switching domain mode, the single-connection occupancy mode, and the cross-connection occupancy mode. Specifically, the computing resource allocation modes are designed based on the principle of minimizing resource overhead.

[0051] S3, Design Pattern Selection Logic, including:

[0052] S31. When the amount of tasks received by this domain is less than or equal to the computing resources of this domain, the computing resources of this domain are dynamically allocated using the private mode of the exchange domain.

[0053] S32. When the amount of tasks received by this domain is greater than the computing resources of this domain, and the computing resources of the other domain are greater than the difference between the amount of tasks and the computing resources of this domain, the computing resources of this domain and the other domain are dynamically allocated using a single-connection occupancy mode.

[0054] S33. When the local domain receives a new task while executing a task allocated by the other domain, a cross-domain occupancy mode is adopted to dynamically allocate computing resources between the local domain and the other domain, thus completing the construction of the mode selection logic.

[0055] Furthermore, in step S32 above, when the amount of tasks received by the local domain is greater than the local domain's computing resources, and the computing resources of the peer domain are greater than the difference between the amount of tasks and the computing resources of the local domain, a single-connection occupancy mode is used to dynamically allocate computing resources between the local and peer domains, including:

[0056] S321. When the amount of tasks received by this domain is greater than the amount of remaining resources in this domain, and any one of the sub-computing resources in the other domain is greater than the difference between the amount of tasks and the computing resources in this domain, the computing resources of this domain and the other domain are dynamically allocated using a single-connection occupancy mode.

[0057] S322. When the amount of tasks received by this domain is greater than the amount of remaining resources in this domain, and all sub-computing resources in the other domain are less than the difference between the amount of tasks and the computing resources in this domain, it is determined whether the task can be split. If the task can be split, the computing resources of this domain and the other domain are dynamically allocated using a single-connection occupancy mode. If the task cannot be split, the task waits.

[0058] Specifically, in steps S31 to S33 above, the dynamic allocation of computing resources includes the following process, wherein the two switching domain hosts can be distinguished by cross-domain hosts and task hosts respectively:

[0059] (1) Estimate the current task quantity for the current task received by the task host. If the resource N required by the task (i.e. the task quantity, i.e. the current task quantity in steps S61 to S64 below) is less than the computing resource M in this domain perceived at this moment (i.e. the computing resource, i.e. the current total computing resource in steps S61 to S64 below), i.e. N < M, the resource orchestration scheduler starts the private mode of the switching domain, i.e., completes the task scheduling and allocation in this domain.

[0060] (2) Based on the current task volume, if the resource N required by the task is greater than the computing resource M perceived in the local domain at this moment (i.e., N > M), the resource orchestration scheduler will start the single-connection occupancy mode and write the required resource value NM (i.e., the difference) into the shared memory area between the cross-domain host and the task host. After the cross-domain host receives the task request with this value, it compares it with the perceived local domain resource value P at this moment (i.e., the current sub-computing resource in steps S61 to S64 below). If NM < P, the idle slave node permission will be released, and the task host will access the cross-domain slave node resource through the NT port to implement task deployment.

[0061] (3) If NM > P, the cross-domain host writes the value of P into the shared memory to inform the task host with the task. The resource orchestration scheduler of the task host then determines whether the task can be further split. If it can be split, the NMP resources are split and distributed to the cross-domain host. The host continues to sense the resources of the local node in real time and queries the cross-domain host for the remaining resource value through the shared memory. When there are idle resources, they are allocated immediately.

[0062] (4) If a single bridging event is occurring, and the host being bridging receives a new computing task with a required resource value of Q, and senses the resource value of the local domain at this moment R, if Q > R, then the required resource value QR needs to be communicated to the other host through shared memory. If the other host senses the resource value at this moment as S, and S > QR, then a cross-bridging occupancy mode occurs. At this time, the resource orchestration scheduler schedules the task in a more complex manner.

[0063] (5) When all tasks have completed their computation tasks, all co-intelligent processors write the XX_POI NT_FREE command to their corresponding switching domain host for the release of sub-resources, and for returning the usage rights of the slave node to the switching domain host after a single cross-connection occupation or cross-boundary occupation occurs.

[0064] Furthermore, in step S5 above, the PC Ie switching domain writes its current resource information into shared memory at a first predetermined period, and retrieves the current resource information of the other domain from the shared memory at a second predetermined period, including:

[0065] S51. Each of the co-intelligent processors connected downstream of the PC Ie switching domain reports its current usage status and current sub-computing resources according to a first set period. Based on the current sub-computing resources of all the co-intelligent processors connected downstream of the PC Ie switching domain, the current total computing resources are obtained. The current total computing resources and the current usage status and current sub-computing resources of each co-intelligent processor are written into the shared memory as the current resource information of the domain.

[0066] S52. The PC Ie switching domain obtains the current resource information of the other domain according to the second set period.

[0067] Furthermore, in step S6 above, the PC Ie switching domain dynamically allocates computing resources based on the current task volume, the current resource information of its own domain, and the current resource information of the other domain, according to the mode selection logic, by selecting a computing resource allocation mode, including:

[0068] S61. The PC Ie switching domain estimates the current task volume based on the received current task, and obtains the current total computing resources and current sub-computing resources of the domain.

[0069] S62. When the current task volume of this domain is less than or equal to the current total computing resources, the current total computing resources of this domain are dynamically allocated using the private mode of the exchange domain.

[0070] S63. When the current task volume of this domain is greater than the current total computing resources, and the difference between the current task volume of this domain and the current total computing resources is less than any of the current sub-computing resources in the other domain, select the single-connection occupancy mode to dynamically allocate computing resources between this domain and the other domain.

[0071] S64. When the current task quantity of this domain is greater than the current total computing resources, and all current sub-computing resources in the other domain are less than the difference between the current task quantity and the current total computing resources of this domain, determine whether the current task of this domain can be split. If it can be split, select the single-connection occupancy mode to dynamically allocate computing resources of this domain and the other domain. If it cannot be split, wait for the task.

[0072] Based on the same inventive concept, this invention also provides an intelligent computing resource agile dynamic allocation system, as described in the following embodiments. Since the principle of the intelligent computing resource agile dynamic allocation system in solving the problem is similar to the intelligent computing resource agile dynamic allocation method provided in the above embodiments, the implementation of the intelligent computing resource agile dynamic allocation system can refer to the implementation of the intelligent computing resource agile dynamic allocation method, and repeated details will not be elaborated further. As used below, the terms "unit" or "module" can refer to a combination of software and / or hardware that implements a predetermined function. Although the apparatus described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0073] This invention discloses an intelligent computing resource agile dynamic allocation system, which includes two PCIe switching domains, a communication module, a resource write module, a resource read module, and a computing resource dynamic allocation module. The structure is described below.

[0074] The two PC Ie switching domains are connected via NT ports. Each PC Ie switching domain includes a switching domain host and multiple co-intelligent processors. The host is equipped with a resource planning and orchestration scheduler, which is located between the co-intelligent processor and the user terminal.

[0075] The communication module is used to establish a communication link between the two PC Ie switching domains;

[0076] The resource writing module is used to write the current total computing resources of the PC Ie switching domain and the current usage status and current sub-computing resources of each of the downstream co-intelligent processors into the shared memory according to a first set period.

[0077] The resource reading module is used to obtain the current total computing resources of the other domain and the current usage status and current sub-computing resources of each downstream co-intelligent processor from the shared memory according to a second set period, wherein the first set period is shorter than the second set period.

[0078] The dynamic allocation module for computing resources is used by the PC Ie switching domain to dynamically allocate computing resources based on the current workload, the current resource information of the domain, and the current resource information of the peer domain, according to the mode selection logic, by selecting a computing resource allocation mode.

[0079] The method of this invention constructs two PCIe switching domains at the user end. Each switching domain has one switching domain host and multiple co-intelligent processors. The two PCIe switching domains communicate via mapped NT (non-transparent) ports. Simultaneously, a resource planning and orchestration scheduler is designed in each switching domain host to achieve shielding of underlying resources. By constructing a resource exchange network and using shared memory communication, the underlying resource quantity is perceived in real time. Flexible configuration modes such as private switching domain, single-connection occupation, and cross-boundary occupation are possible. This invention solves the problems of low computing power, inflexible resource scheduling, underutilization of resources, and complex upper-layer application design in existing edge embedded intelligent computing centers. By constructing a shieldable resource planning and orchestration scheduler, the load balancing of computing resources in intelligent computing centers is improved, and resource scheduling is flexible. Furthermore, the dual-switching domain cross-connection communication mode enhances computing power configuration.

[0080] In this embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the above-mentioned intelligent computing resource agile dynamic allocation methods.

[0081] Specifically, the computer device can be a computer terminal, a server, or a similar computing device.

[0082] In this embodiment, a computer-readable storage medium is provided, which stores a computer program that executes any of the above-described intelligent computing resource agile dynamic allocation methods.

[0083] Specifically, computer-readable storage media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. As defined herein, computer-readable storage media do not include transient computer-readable media, such as modulated data signals and carrier waves.

[0084] Obviously, those skilled in the art should understand that the modules or steps of the above-described embodiments of the present invention can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the embodiments of the present invention are not limited to any particular hardware and software combination.

[0085] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, various modifications and variations can be made to the embodiments of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for agile dynamic allocation of intelligent computing resources, characterized in that, The method comprises: constructing a shared memory and two PCIe exchange domains, each of the PCIe exchange domains comprising an exchange domain host, the exchange domain host comprising a resource planning and arrangement scheduler, one end of the exchange domain host being connected with a user end and the other end being connected with each of the intelligent processors downstream, the resource planning and arrangement scheduler being connected with the shared memory; connecting the two PCIe exchange domains on the user end and establishing a communication link between the two PCIe exchange domains; the PCIe exchange domain writing current resource information of the domain into the shared memory at a first set period and obtaining current resource information of the opposite domain from the shared memory at a second set period, the first set period being less than the second set period; the PCIe exchange domain performing dynamic allocation of computing resources according to a current task amount, the current resource information of the domain and the current resource information of the opposite domain according to mode selection logic, including designing the computing resource allocation mode including exchange domain private mode, single cross-connection possession mode and cross-border possession mode according to the task amount and computing resources of the task; defining that the task amount received by the domain is less than or equal to the computing resources of the domain, the computing resources of the domain are dynamically allocated by the exchange domain private mode; defining that the task amount received by the domain is greater than the computing resources of the domain, and the computing resources of the opposite domain are greater than the difference between the task amount and the computing resources of the domain, the computing resources of the domain and the opposite domain are dynamically allocated by the single cross-connection possession mode; defining that the domain receives a new task while executing the task allocated by the opposite domain, the computing resources of the domain and the opposite domain are dynamically allocated by the cross-border possession mode, and the mode selection logic is completed.

2. The intelligent computing resource agile dynamic allocation method according to claim 1, characterized in that, defining that the task amount received by the domain is greater than the computing resources of the domain, and the computing resources of the opposite domain are greater than the difference between the task amount and the computing resources of the domain, the computing resources of the domain and the opposite domain are dynamically allocated by the single cross-connection possession mode, including: defining that the task amount received by the domain is greater than the remaining resource amount of the domain, and any one of the sub-computing resources of the opposite domain is greater than the difference between the task amount and the computing resources of the domain, the computing resources of the domain and the opposite domain are dynamically allocated by the single cross-connection possession mode; defining that the task amount received by the domain is greater than the remaining resource amount of the domain, and all the sub-computing resources of the opposite domain are less than the difference between the task amount and the computing resources of the domain, whether the task can be split is judged, if the task can be split, the computing resources of the domain and the opposite domain are dynamically allocated by the single cross-connection possession mode, and if the task cannot be split, the task is waited. 3.The intelligent computing resource agile dynamic allocation method according to claim 1 or 2, characterized in that, the PCIe exchange domain writing current resource information of the domain into the shared memory at a first set period and obtaining current resource information of the opposite domain from the shared memory at a second set period, including: Each of the co-intelligent processors connected downstream of the PCIe exchange domain reports current usage status and current sub-computing resources at a first set period, obtains current total computing resources from current sub-computing resources of all the co-intelligent processors connected downstream of the PCIe exchange domain, and writes the current total computing resources and the current usage status and the current sub-computing resources of each of the co-intelligent processors into the shared memory as the current domain resource information; The PCIe exchange domain obtains the opposite domain current resource information of the opposite domain at the second set period. 4.The intelligent computing resource agile dynamic allocation method according to claim 1 or 2, characterized in that, The PCIe exchange domain selects a computing resource allocation mode according to a current task amount, the current domain resource information and the opposite domain current resource information according to mode selection logic to perform dynamic allocation of computing resources, including: The PCIe exchange domain estimates the current task amount by estimating the received current task, and obtains the current total computing resources and the current sub-computing resources of the domain; When the current task amount of the domain is less than or equal to the current total computing resources, the current total computing resources of the domain are dynamically allocated in the exchange domain private mode; When the current task amount of the domain is greater than the current total computing resources, and the difference between the current task amount of the domain and the current total computing resources is less than any one of the current sub-computing resources in the opposite domain, the computing resources of the domain and the opposite domain are dynamically allocated in the single cross-occupy mode; When the current task amount of the domain is greater than the current total computing resources, and all the current sub-computing resources in the opposite domain are less than the difference between the current task amount of the domain and the current total computing resources, it is judged whether the current task of the domain is splittable, if splittable, the computing resources of the domain and the opposite domain are dynamically allocated in the single cross-occupy mode, if not splittable, the task is waited.

5. An intelligent computing resource agile dynamic allocation system, characterized by, To implement the method of any one of claims 1 to 4, the system comprises: Two PCIe exchange domains connected by an NT port between the two PCIe exchange domains, the PCIe exchange domain comprising an exchange domain host and a plurality of co-intelligent processors, the host being provided with a resource planning and arrangement scheduler, the resource planning and arrangement scheduler being located between the co-intelligent processors and a user end; A communication module for establishing a communication link between the two PCIe exchange domains; A resource writing module for writing current total computing resources of the PCIe exchange domain and current usage status and current sub-computing resources of each of the co-intelligent processors connected downstream of the PCIe exchange domain into a shared memory at a first set period; A resource reading module for obtaining current total computing resources of the opposite domain and current usage status and current sub-computing resources of each of the co-intelligent processors connected downstream of the opposite domain from the shared memory at a second set period, the first set period being less than the second set period; A computing resource dynamic allocation module is configured to dynamically allocate computing resources according to a current task amount, current resource information of the local domain and current resource information of the opposite domain, and according to mode selection logic to select a computing resource allocation mode.

6. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the intelligent computing resource agile dynamic allocation method in any one of claims 1 to 4 when executing the computer program.

7. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program for implementing the intelligent computing resource agile dynamic allocation method in any one of claims 1 to 4.

Images