WO2023050819A1

WO2023050819A1 - System on chip, virtual machine task processing method and device, and storage medium

Info

Publication number: WO2023050819A1
Application number: PCT/CN2022/091613
Authority: WO
Inventors: 胡延隆; 李文广; 郭文
Original assignee: 上海商汤智能科技有限公司
Priority date: 2021-09-29
Filing date: 2022-05-09
Publication date: 2023-04-06
Also published as: CN113821308A; CN113821308B

Abstract

Provided is a virtual machine task processing method, which is applied to a system on chip. The task processing method comprises: acquiring task requests sent by virtual machines (S310), wherein for each task request, the task request comprises a corresponding virtual machine identifier; and on the basis of a pre-established mapping relationship between the virtual machines and processing units of the system on chip, a main control unit of the system on chip distributing, according to the virtual machine identifiers, the task requests to the corresponding processing units for processing (S320).

Description

System on chip, virtual machine task processing method and device, storage medium

Related Application Cross Reference

This application claims the priority of the Chinese patent application with the application number 202111155033.X and the filing date of September 29, 2021. The entire content of the Chinese patent application is hereby incorporated by reference into this application.

technical field

The present disclosure relates to the field of computer technology, and in particular to a system on chip, a virtual machine task processing method and device, and a storage medium.

Background technique

Virtualization refers to the division of physical resources into multiple logical partitions through virtualization management software on a physical machine server. The data resources of each partition are isolated from each other and become independent virtual machines. At present, virtualization technologies mainly include software virtualization and hardware-assisted virtualization. However, the performance of software virtualization solutions is poor, and the hardware-assisted virtualization solutions require pre-adaptation of hardware, and the cost is relatively high.

Contents of the invention

In a first aspect, the embodiment of the present disclosure provides a virtual machine task processing method, which is applied to a system-on-chip. The task processing method includes: acquiring task requests sent by each virtual machine, wherein, for each task request, the The task request includes a corresponding virtual machine identifier; based on the pre-established mapping relationship between the virtual machine and each processing unit of the system-on-chip, the main control unit of the system-on-chip sends each of the task requests according to the virtual machine identifier Distributed to the corresponding processing unit for processing.

In some implementations, for any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to one of the processing units, the main control unit of the system-on-chip will Each of the task requests is distributed to the corresponding processing unit for processing, including: for the task request sent by any one of the virtual machines, the main control unit distributes the task request to the processing unit corresponding to the any one of the virtual machines unit is processed.

In some embodiments, for any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to at least two processing units in the processing units, the main control unit of the system-on-chip Identifying and distributing each task request to a corresponding processing unit for processing includes: for the task request sent by any one of the virtual machines, the main control unit splits and processes the task corresponding to the task request to obtain multiple sub- task; the number of the plurality of subtasks is less than or equal to the number of processing units corresponding to any one virtual machine; the main control unit distributes each subtask to one or more of the processing units corresponding to any one virtual machine Multiple processing units perform processing.

In some implementations, for any one of the processing units, in response to the fact that the any one processing unit corresponds to at least two virtual machines in the virtual machines, the main control unit of the system on chip according to the virtual machine Identifying and distributing each task request to a corresponding processing unit for processing includes: the main control unit obtaining a task request sent by a virtual machine corresponding to any one processing unit; according to the virtual machine corresponding to any one processing unit Priority, for the task request sent by the virtual machine corresponding to any one of the processing units, determine the processing time corresponding to the task request; according to the processing time, the any processing unit corresponds to the any one processing unit Each task request sent by the virtual machine is processed.

In some implementations, according to the priority of the virtual machine corresponding to any one of the processing units, for the task request sent by the virtual machine corresponding to any one of the processing units, determining the processing time corresponding to the task request includes: The priority of each virtual machine is set, and the weight value corresponding to each task request sent by the virtual machine corresponding to the arbitrary processing unit is determined; based on the weight value, the task request sent by the virtual machine corresponding to the arbitrary processing unit is determined. The window processing time of each task request within the current time window; according to the processing time, the any processing unit sequentially processes each task request sent by the virtual machine corresponding to the any processing unit, including: Each task request sent by the virtual machine corresponding to any processing unit is distributed to the task queue for queuing; within the current time window, the any processing unit sequentially requests each task in the task queue according to the window processing time to process.

In some implementations, within the current time window, the any processing unit sequentially processes each task request in the task queue according to the window processing time, including: responding to Within the time window, for each task request in the task queue, if the task request is not processed within the corresponding window processing time, the processing process of the task request is interrupted, and the task request is added to the tail of the task queue .

In some implementation manners, the method further includes: based on the mapping relationship, the main control unit sends the processing results corresponding to the task requests to the virtual machines respectively corresponding to the task requests.

In some implementations, the mapping relationship is stored in a memory of the SoC.

In the second aspect, the embodiment of the present disclosure provides a system on chip, including: a main control unit and at least one processing unit; a storage unit storing a pre-established mapping relationship between a virtual machine and each processing unit; wherein, the The main control unit is configured to obtain task requests sent by each virtual machine, and based on the mapping relationship, distribute each task request to a corresponding processing unit for processing according to the virtual machine identifier included in the task request.

In some implementations, the system-on-a-chip in the implementation of the present disclosure further includes: a second communication module, configured to establish a communicable connection with the first communication module of the host computer, and the main control unit obtains the The task request.

In some implementations, for any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to one of the processing units, the main control unit is configured to: for the any one of the virtual machines The task request sent by the virtual machine is distributed to the processing unit corresponding to any one of the virtual machines for processing.

In some implementations, for any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to at least two of the processing units, the main control unit is configured to: A task request sent by a virtual machine is split and processed to obtain multiple subtasks; the number of the multiple subtasks is less than or equal to the number of processing units corresponding to any one of the virtual machines; The subtasks are distributed to one or more processing units in the processing units corresponding to any one virtual machine for processing.

In some implementations, for any one of the processing units, in response to the fact that the any one processing unit corresponds to at least two virtual machines in the virtual machines, the main control unit is configured to: acquire the A task request sent by a virtual machine corresponding to a processing unit; according to the priority of the virtual machine corresponding to any one processing unit, for the task request sent by the virtual machine corresponding to any one processing unit, determine the processing corresponding to the task request Time: according to the processing time, the any one processing unit sequentially processes each task request sent by the virtual machine corresponding to the any one processing unit.

In some implementations, the main control unit is configured to: determine the weight value corresponding to each task request sent by the virtual machine corresponding to any one of the processing units according to the preset priority of each virtual machine; The weight value is used to determine the window processing time of each task request sent by the virtual machine corresponding to the arbitrary processing unit in the current time window; each task sent by the at least two virtual machines corresponding to the arbitrary processing unit The request is distributed to the task queue for queuing; within the current time window, any one of the processing units sequentially processes each task request in the task queue according to the window processing time.

In some implementations, the main control unit is configured to: in response to each task request in the task queue within the current time window, the task request has not been processed within the corresponding window processing time, The processing process of the task request is interrupted, and the task request is added to the tail of the task queue.

In some implementation manners, the main control unit is configured to: based on the mapping relationship, send the processing results corresponding to the task requests to the virtual machines respectively corresponding to the task requests.

In a third aspect, an embodiment of the present disclosure provides a virtual machine task processing device, including: a host machine including at least one pre-established virtual machine; and the system-on-chip according to any embodiment of the second aspect, and the The host computer can communicate with each other.

In a fourth aspect, the embodiments of the present disclosure provide a storage medium storing computer-readable instructions, the computer-readable instructions are used to cause a computer to execute the method according to any embodiment of the first aspect.

In a fifth aspect, an embodiment of the present disclosure provides a computer program product, including a computer program stored in a memory, and when the computer program instructions are executed by a processor, the method described in any one of the embodiments of the first aspect is implemented.

The virtual machine task processing method of the embodiment of the present disclosure includes obtaining the task request sent by each virtual machine, the task request includes the corresponding virtual machine identifier, and based on the pre-established mapping relationship between the virtual machine and each processing unit of the system-on-chip, the on-chip The main control unit of the system distributes each task request to the corresponding processing unit for processing according to the virtual machine identifier. In the embodiments of the present disclosure, the system on chip schedules and distributes tasks based on the mapping relationship between the virtual machine and the hardware resources, realizes the virtualization of the system on chip, reduces virtualization cost and improves the performance of the virtual machine.

Description of drawings

In order to more clearly illustrate the technical solutions in the specific embodiments of the present disclosure, the following will briefly introduce the drawings that need to be used in the specific embodiments. Obviously, the drawings in the following description are some embodiments of the present disclosure. For those skilled in the art, other drawings can also be obtained based on these drawings without creative effort.

Fig. 1 is a schematic structural diagram of a system on a chip according to some embodiments of the present disclosure.

Fig. 2 is a schematic diagram of a virtual machine task processing method according to some implementations of the present disclosure.

Fig. 3 is a flowchart of a virtual machine task processing method according to some implementations of the present disclosure.

Fig. 4 is a flowchart of a virtual machine task processing method according to some implementations of the present disclosure.

Fig. 5 is a flowchart of a virtual machine task processing method according to some implementations of the present disclosure.

Fig. 6 is a flowchart of a method for processing a virtual machine task according to some implementations of the present disclosure.

Detailed ways

The technical solutions of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings. Apparently, the described implementations are part of the implementations of the present disclosure, but not all of them. Based on the implementation manners in the present disclosure, all other implementation manners obtained by persons of ordinary skill in the art without creative efforts shall fall within the protection scope of the present disclosure. In addition, the technical features involved in different embodiments of the present disclosure described below may be combined with each other as long as they do not constitute a conflict with each other.

Virtualization is the foundation of cloud computing. Virtualization refers to the division of physical resources into multiple logical partitions on the physical machine server through virtualization management software. The data resources of each partition are isolated from each other and become independent virtual machines. Among the related technologies, the virtualization technology mainly includes software virtualization and hardware-assisted virtualization.

Software virtualization refers to the interception and simulation of access to physical resources on a physical machine through software methods. A common software virtual machine, such as QEMU, uses software to simulate the addressing, decoding and execution of the physical platform processor, and the instructions of the client (Guest) are not directly executed on the physical machine platform. Because the software virtualization method has no hardware support, all instructions are simulated by software, so the performance is often relatively poor, and only relatively simple processing tasks can be performed.

Hardware-assisted virtualization means that the physical machine platform itself provides hardware support for the interception and redirection of special instructions, and even the hardware itself provides additional physical resources to assist software in virtualizing key hardware resources. It can be understood that since the hardware itself provides resource support for virtualization, hardware-assisted virtualization has better performance. However, when implementing virtualization, it is necessary to perform special adaptation on hardware resources in advance to support the virtualization of physical resources, that is, it is necessary to improve the hardware itself, which increases the cost of virtualization.

SoC (System on Chip, system on chip) refers to a system-level chip with a complete system integrated circuit and embedded software content. Currently, common such as GPU (graphics processing unit, graphics processor) chips, mobile phone chips, etc., can be called SoC, also called system on chip.

SoC often includes a main control unit and multiple processing subsystems. For virtualization scenarios, the host (Host) cannot directly control the processing subsystems in the SoC, so it is impossible to directly virtualize the physical resources of these processing subsystems. Segmentation makes it impossible to leverage existing SoCs for hardware-assisted virtualization.

In related technologies, it is necessary to perform special adaptation work on the SoC, that is, to perform corresponding hardware improvements on the SoC, so that host instructions can directly access the processing subsystem of the SoC, which undoubtedly increases the cost of virtualization.

Based on the defects existing in the above-mentioned related technologies, embodiments of the present disclosure provide a virtual machine task processing method, device, SoC chip, and storage medium. The embodiment of the present disclosure aims to realize virtualization without hardware adaptation of the SoC, and improve the performance of the virtual machine.

Embodiments of the present disclosure provide a virtual machine task processing method, which can be applied to a system on chip (hereinafter collectively referred to as "SoC").

FIG. 1 shows the basic structure of SoC. As shown in FIG. 1 , a common SoC includes a main control unit 110 and a plurality of processing units 120 . The main control unit 110 may be, for example, a CPU (central processing unit, central processing unit). Multiple processing units 120 are hardware subsystems (Subsys) of the SoC, and are used to complete task processing.

Based on the foregoing, it can be seen that the host (Host) can only communicate with the processing unit 120 through the main control unit 110 , and cannot directly control each processing unit 120 . Therefore, the Host side cannot directly virtualize the processing unit 120 of the SoC.

In the embodiment of the present disclosure, the mapping relationship between the host-side virtual machine and the processing unit 120 of the SoC is established in advance, and then the main control unit 110 completes the distribution and scheduling of virtual machine tasks based on the mapping relationship, and realizes multiple processing Virtualized partitioning of unit 120 . FIG. 2 shows a functional block diagram for implementing virtualization in an embodiment of the present disclosure, which will be described below with reference to FIG. 2 .

As shown in FIG. 2, in some implementation manners, the SoC 100 establishes a communicable connection with the host host 200 through the communication bus 300. In some embodiments, the host machine 200 is provided with a first communication module 210, the main control unit 110 of the SoC 100 is integrated with a second communication module 111, and the first communication module 210 and the second communication module 111 establish a communication bus 300 through the communication bus 300. communication connection. The first communication module 210 and the second communication module 111 may adopt an inter-core communication manner and use a self-defined communication protocol for communication.

For example, in an example, the communication bus 300 is a PCIe bus, and the corresponding first communication module 210 and the second communication module 111 are also communication modules of the PCIe communication protocol.

Those skilled in the art can understand that the form of communication between the host machine 200 and the SoC 100 is not limited to PCIe, and can also be any other form suitable for implementation, such as USB, eth-net, etc., which is not limited in the present disclosure.

In addition, it is worth noting that FIG. 2 only shows an example in which only one SoC 100 is mounted on the host computer 200. In other embodiments, multiple SoCs can also be mounted on the host computer 200 according to specific requirements. The principle As such, the present disclosure is not limited thereto.

Continuing to refer to FIG. 2 , the host machine 200 can create multiple virtual machines 220 through virtualization management software (VMM), such as VMO, VM1, VM2, . . . , VMn. In some implementations, each virtual machine 220 can run an independent operating system, and various applications can also be installed on the upper layer of each virtual machine, so that the client (Guest) can directly use the services corresponding to the applications. Those skilled in the art can understand and implement this, and the present disclosure will not repeat it.

In the embodiment of the present disclosure, when implementing virtualization, instead of directly dividing the physical resources of each processing unit 120 of the SoC 100 and then allocating them to each virtual machine 220, a relationship between each virtual machine 220 and the processing unit 120 is established. The mapping relationship among them can be stored in the memory of the SoC 100 (not shown in FIG. 2 ), so as to be called and maintained by the main control unit 110. When the main control unit 110 of the SoC 100 receives the task request, it can determine the processing unit corresponding to the virtual machine sending the task request based on the mapping relationship, and then send the task request to the corresponding processing unit for processing.

FIG. 3 shows some implementations of the method for processing virtual machine tasks of the present disclosure, which will be described below in conjunction with FIG. 3 .

As shown in FIG. 3 , in some implementation manners, the method for processing a virtual machine task in the example of the present disclosure includes step S310 to step S320 .

S310. Obtain a task request sent by each virtual machine.

In some implementations, with the system architecture shown in FIG. 2 , when the client end uses the services provided by the virtual machine 220, the virtual machine 220 can send the corresponding information to the SoC 100 through the first communication module 210 using the communication bus 300 as a medium. task request, so that the main control unit 110 of the SoC 100 can obtain the task request sent by each virtual machine through the second communication module 111.

The task request includes relevant data about executing the task and a virtual machine identifier, and the virtual machine identifier is used to uniquely point to the ID of the virtual machine sending the task request.

S320. Based on the pre-established mapping relationship between the virtual machine and each processing unit of the SoC, the main control unit of the SoC distributes each task request to the corresponding processing unit for processing according to the virtual machine identifier.

In some implementation manners, the mapping relationship between the virtual machine 220 and each processing unit 120 of the SoC 100 needs to be established in advance.

In some implementations, the mapping relationship between the virtual machine 220 and the processing unit 120 can be set based on the user's scenario requirements, purchase conditions, and the like. And after the mapping relationship is set, the mapping relationship can also be adjusted at any time according to the usage requirements and purchase conditions of subsequent users, which will not be described in this disclosure.

The mapping relationship between the virtual machine 220 and the processing unit 120 can be mainly divided into one-to-one mapping, one-to-many mapping and many-to-one mapping.

1) One-to-one mapping

Each virtual machine 220 corresponds to a processing unit in the SoC 100. For example, in the example of FIG. 2, virtual machine VM0 corresponds to processing unit 1200 (hardware subsystem 0), and virtual machine VM1 corresponds to processing unit 1201 (hardware subsystem 1). , . . . , the virtual machine VMn corresponds to the processing unit 120n (hardware subsystem n).

2) One-to-many mapping

Each virtual machine 220 corresponds to at least two processing units in SoC 100. For example, in the example of FIG. 2 , the virtual machine VM0 corresponds to the processing unit 1200 (hardware subsystem 0), the processing unit 1201 (hardware subsystem 1) and the processing unit 1202 (hardware subsystem 2). Other numbers of one-to-many mappings are also possible, and are not limited to this example.

3) Many-to-one mapping

Multiple (at least two) virtual machines 220 correspond to one processing unit in SoC 100. For example, in the example of FIG. 2 , the virtual machine VM0 and the virtual machine VM1 jointly correspond to the processing unit 1200 (hardware subsystem 0). Of course, other numbers of many-to-one mappings are also possible, and are not limited to this example.

Based on the above mapping relationship, a corresponding relationship between each virtual machine 220 and the processing unit 120 may be established. When the main control unit 110 of the SoC 100 receives the task request, it can determine the processing unit corresponding to the virtual machine sending the task request according to the virtual machine identifier in the task request and based on the mapping relationship established above. Furthermore, the main control unit 110 can distribute the task request to one or more corresponding processing units through a preset scheduling and distribution algorithm for processing.

The process of scheduling and distributing each task request by the main control unit 110 will be described in the following embodiments of the present disclosure, and will not be described in detail here.

In the embodiments of the present disclosure, by establishing a mapping relationship between a virtual machine and a hardware resource, the main control unit of the SoC is used to implement scheduling and distribution of tasks based on the mapping relationship, and the virtualization of the system on chip is realized by combining software with hardware. Compared with the software virtualization solution in the related art, the embodiment of the present disclosure has higher performance due to the use of hardware-assisted virtualization. Compared with the hardware-assisted virtualization solution in the related art, the embodiment of the present disclosure does not require special adaptation of hardware resources, and virtualization can be realized by using a general-purpose SoC, thereby reducing virtualization costs.

From the above, it can be seen that the virtual machine task processing method of the embodiment of the present disclosure establishes the mapping relationship between the virtual machine and the hardware resource, and uses the main control unit of the SoC to schedule and distribute tasks based on the mapping relationship, realizing the SoC. Virtualization reduces virtualization costs and improves virtual machine performance.

In some implementation manners, the above mapping relationship may be stored and maintained by the host side and the SoC side together, that is, the host side and the SoC side adopt the same mapping relationship. For example, when a user pays to purchase new hardware resources, the host machine 200 can update the original mapping relationship based on the user's purchase request, and at the same time pass the updated mapping relationship through the first communication module 210 using the communication bus 300 as a medium send to SoC 100, so that SoC 100 updates the original mapping relationship according to the received new mapping relationship. Those skilled in the art can understand this, and the present disclosure will not repeat it here.

Based on the foregoing, in the embodiments of the present disclosure, the mapping relationship between the virtual machine 220 and the mapping relationship processing unit 120 includes various situations. The task processing method of the present disclosure will be described below based on the foregoing three situations respectively.

1) One-to-one mapping. As shown in FIG. 2 , in an example, the virtual machine VM0 corresponds to the processing unit 1200 (hardware subsystem 0 ) in the SoC in the mapping relationship.

When the user uses the service of the virtual machine VMO through the client (Guest), the virtual machine VMO sends a task request to the SoC 100 through the first communication module 210 using the communication bus 300 as a communication medium. The task request carries the identifier of the virtual machine sending the task request, for example, the identifier of the virtual machine is VMO.

After receiving the task request through the second communication module 111, the SoC 100 invokes the pre-stored mapping relationship according to the virtual machine identifier VM0 carried in the task request, and determines that the hardware subsystem corresponding to the virtual machine V0 is the hardware subsystem 0. Then the main control unit 110 can distribute the task corresponding to the task request to the hardware subsystem 0 for processing.

2) One-to-many mapping. As shown in FIG. 2 , in an example, the virtual machine VMO corresponds to hardware subsystem 0 , hardware subsystem 1 and hardware subsystem 2 in the SoC in a mapping relationship.

After receiving the task request through the second communication module 111, the SoC 100 invokes the pre-stored mapping relationship according to the virtual machine identifier VM0 carried in the task request, and determines that the hardware subsystem corresponding to the virtual machine V0 is the hardware subsystem 0, Hardware Subsystem 1 and Hardware Subsystem 2.

Since the virtual machine VM0 corresponds to three processing units, in order to improve task processing efficiency, the main control unit 110 may split and process the task requested by the task, so as to obtain three subtasks. Then the three subtasks are respectively distributed to three processing units of hardware subsystem 0, hardware subsystem 1 and hardware subsystem 2 for processing. After each processing unit finishes processing the subtasks, the main control unit 110 obtains the final task processing results according to the processing results of the three subtasks.

In an example, the main control unit 110 receives the task request sent by the virtual machine VM0, parses to obtain the task data included in the task request, and then divides the task data based on the number of processing units 120 corresponding to the virtual machine VM0 to obtain Three subtasks are divided, and the three subtasks are respectively sent to three processing units of hardware subsystem 0, hardware subsystem 1 and hardware subsystem 2 for processing. After each processing unit processes the subtasks, the corresponding processing results can be obtained. For example, hardware subsystem 0 obtains the first result data after processing the subtasks, and hardware subsystem 1 obtains the second result data after processing the subtasks. Subsystem 2 obtains third result data after processing the subtasks. The main control unit 110 then fuses the first result data, the second result data and the third result data to obtain a task processing result corresponding to the task request. In the embodiment of this formula, when the virtual machine VM0 corresponds to 3 processing units, in order to improve the task processing efficiency, the main control unit 110 may split and process the task requested by the task, so as to obtain 2 subtasks. Then distribute these two subtasks to at least one processing unit in hardware subsystem 0, hardware subsystem 1, and hardware subsystem 2 for processing, for example, distribute these two subtasks to hardware subsystem 0, hardware subsystem respectively 1. These two processing units perform processing. After the two processing units process the subtasks, the main control unit 110 obtains the final task processing result according to the processing results of the two subtasks. Therefore, in one-to-many mapping, the number of multiple subtasks is less than or equal to the number of processing units corresponding to one virtual machine.

It can be known from the above that, in the embodiments of the present disclosure, when one virtual machine corresponds to multiple processing units, the tasks requested by the task are split and processed so that multiple processing units can process tasks in parallel, thereby improving task processing efficiency.

3) Many-to-one mapping. As shown in FIG. 2 , in an example, in the mapping relationship, the virtual machines VM0 and VM1 both correspond to the hardware subsystem 0 in the SoC.

In some implementations, when only one of the virtual machines VM0 and VM1 sends a task request, the main control unit 110 distributes the task request to the hardware subsystem 0 for processing based on a process similar to one-to-one mapping. I won't repeat it publicly.

In some other implementations, when the virtual machines VM0 and VM1 send task requests at the same time, the main control unit 110 can determine the processing time corresponding to each task request according to the priority, so as to distribute and process multiple task requests, as shown below in conjunction with FIG. 4 Embodiments will be described.

As shown in FIG. 4 , in some implementations, in the virtual machine task processing method of the example of the present disclosure, the process of distributing and processing multiple task requests includes S410 to S430.

S410. The main control unit acquires the task request sent by the virtual machine corresponding to the processing unit.

In some implementation manners, taking the foregoing example as an example, both the virtual machines VM0 and VM1 in the mapping relationship correspond to the hardware subsystem 0 in the SoC. When the virtual machines VM0 and VM1 send task requests at the same time, the main control unit 110 may receive the task requests sent by the virtual machines VM0 and VM1.

S420. According to the priority of the virtual machine corresponding to the processing unit, for the task request sent by the virtual machine corresponding to the processing unit, determine the processing time corresponding to the task request.

In some implementation manners, the priority of the virtual machine indicates the priority of processing the task request sent by each virtual machine. Different virtual machines have their own priorities, and the priorities can be preset based on the use requirements and purchase conditions of the client machines corresponding to the virtual machines. For example, in the above example, the priority of the virtual machine VM0 is P1, and the priority of the virtual machine VM1 is P2.

The processing time indicates the resource occupation of the processing unit by the task request. The higher the priority of the virtual machine, the longer the resource occupation time of the task request on the processing unit is, that is, the corresponding processing time within a time window is longer.

In the embodiments of the present disclosure, the processing time of each task request can be obtained according to the priority of each virtual machine, so that the processing unit can process the task request according to the processing time. For this process, reference may be made to the following implementation manners, which will not be expanded here.

S430. According to the processing time, the processing unit sequentially processes each task request sent by the virtual machine corresponding to the processing unit.

In some implementation manners, after the main control unit determines the processing time of each task request according to the priority of each virtual machine, it may distribute each task request to a task queue for queuing. The processing unit may sequentially retrieve a task request at the head of the queue from the task queue, and process the task request according to the processing time corresponding to the task request.

It can be seen from the above that in the embodiments of the present disclosure, when multiple virtual machines correspond to one processing unit, by setting priorities, task requests are processed based on the priorities of the virtual machines, physical resources are allocated reasonably, and task processing efficiency is improved.

In some implementations of the present disclosure, on the basis of the implementation in FIG. 4 , when distributing and processing the task requests of the virtual machines V0 and VM1 , the other is not processed until one of them is processed. Instead, time is divided into consecutive time windows at equal intervals in the time domain. For example, in an example, every 100 ms may be used as a time window. Within a time window, multiple task requests are processed sequentially, but the processing time corresponding to each task request is not necessarily the same, but is determined based on priority.

As shown in FIG. 5 , in some implementations, in the virtual machine task processing method of the example of the present disclosure, the process of determining the processing time corresponding to each task request includes S510 to S520.

S510. Determine, according to the preset priority of each virtual machine, a weight value corresponding to each task request sent by the virtual machine corresponding to the processing unit.

S520. Based on the weight value, determine the window processing time of each task request sent by the virtual machine corresponding to the processing unit within the current time window.

In some implementations, still taking the foregoing example as an example, the virtual machines VM0 and VM1 in the mapping relationship correspond to the processing unit 1200 (hardware subsystem 0) in the SoC, and the main control unit 110 simultaneously receives the information sent by the virtual machines VM0 and VM1. task request.

Each virtual machine is preset with its own priority, for example, the priority of the virtual machine VM0 is P1, and the priority of the virtual machine VM1 is P2. The weight value w1 of the task request of the virtual machine VM0 can be expressed as: w1=P1/(P1+P2). Similarly, the weight value w2 corresponding to the task request of the virtual machine VM1 can be expressed as: w2=P2/(P1 +P2).

After obtaining the weight value corresponding to each task request, the processing time of each task request within the current time window can be determined according to the total duration of the time window.

In the above example, the total duration of each time window is s, then the processing duration s1 corresponding to the task request of the virtual machine VM0 can be expressed as: s1=w1*s=P1*s/(P1+P2), and In principle, the processing duration s2 corresponding to the task request of the virtual machine VM1 can be expressed as: s2=w2*s=P2*s/(P1+P2).

After obtaining the processing time of each task request, the main control unit 110 can sequentially distribute each task request to the task queue for queuing, and wait for the processing unit to call and complete the processing.

S530. Within the current time window, the processing unit sequentially processes each task request in the task queue according to the window processing time.

In some implementations, after calculating the processing time of each task request within the current time window, the main control unit 110 sequentially distributes each task request to the task queue for queuing, and the processing unit 120 sequentially calls the tasks queued in the task queue. The task request at the head of the team is processed according to the processing time corresponding to the task request.

Taking the foregoing example as an example, the task request of the virtual machine VM0 and the task request of the virtual machine VM1 are arranged in sequence in the task queue. In the current time window s duration, the hardware subsystem 0 may first call the task request processing s1 duration of the virtual machine VM0, and then call the task request processing s2 duration of the virtual machine VM1.

In the embodiment of the present disclosure, if a certain task request has not been processed within the processing duration of a time window, it is necessary to interrupt the processing process of the task request, and re-enter the task queue at the end of the task queue for queuing .

For example, in the above example, in the current time window s duration, the hardware subsystem 0 first calls the task request of the virtual machine VM0 to process the s1 duration, but the task request has not been processed, and at this time it is necessary to interrupt the task request of the virtual machine VM0 The processing process starts to enter the task request processing process of the virtual machine VM1.

At the same time, the task request of the interrupted virtual machine VM0 re-enters the queue at the end of the task queue and waits for the next time window. At the next time window, the hardware subsystem 0 calls the task request again to continue processing. For example, in the current time window s, the hardware subsystem 0 first calls the task request of the virtual machine VM0 to process s1 and then interrupts, adds the task request of the virtual machine VM0 to the tail of the task queue for queuing, and then calls the virtual machine VM1 The task request is interrupted after processing s2 and enters the next time window. In the next time window, the hardware subsystem 0 adds the task request of the virtual machine VM1 to the end of the task queue to continue queuing, and at the same time calls the task request of the virtual machine VM0 again for processing. This cycle is executed until the task request processing is completed.

The above example only shows the case where two virtual machines correspond to the same processing unit. In other implementations, other numbers of virtual machines may correspond to the same processing unit in the mapping relationship. The principle of task processing is the same as that of the above example Similarly, this disclosure does not repeat it.

In addition, in some implementations of the present disclosure, on the basis of the implementation in FIG. 5 , when the task processing method of the present disclosure distributes task requests in each time window, the processing duration of each task request in the current time window is also Dynamic adjustment can be made according to the actual processing time of each task request in the previous time window.

For example, in the time window T1, the total duration of T1 is 100ms, the task processing time of the virtual machine VM0 is 30ms, and the task processing time of the virtual machine VM1 is 70ms. However, in actual task processing, the actual task processing time of the virtual machine VM0 takes up 30ms, but the actual task processing time of the virtual machine VM1 only uses 50ms. Therefore, within 100 ms of the next time window T2, the task processing time of the virtual machine VM0 can be adjusted from 30 ms to 50 ms, and the task processing time of the virtual machine VM1 can be adjusted from 70 ms to 50 ms. That is, the processing duration of each task request in each time window can be dynamically fine-tuned based on the task processing situation in the previous time window. Those skilled in the art can understand this, which is not repeated in the present disclosure.

From the above, it can be seen that in the embodiment of the present disclosure, when multiple virtual machines correspond to one processing unit, by setting the priority of the virtual machines, the task request can be processed based on the priority of the virtual machines, so that physical resources can be allocated reasonably, and the processing speed can be improved. task processing efficiency.

FIG. 6 shows some implementations of the task processing method of the present disclosure, and the method of the present disclosure will be further described below with reference to FIG. 6 .

As shown in FIG. 6 , in some implementation manners, the task processing method of the virtual machine in the example of the present disclosure includes S610 to S640.

S610. Pre-establish a mapping relationship between each virtual machine and each processing unit of the SoC.

In some implementations, referring to the system architecture shown in FIG. 2 , the administrator can pre-allocate physical resources for each virtual machine based on the scenario requirements of each virtual machine user, purchase payment situation, etc., that is, establish each virtual machine and each processing unit mapping relationship between them.

In an example, for a virtual machine that requires more hardware resources or pays more for purchase, a larger number of processing units may be set for the virtual machine in the mapping relationship. For a virtual machine that requires less hardware resources or pays less, one processing unit can be set for the virtual machine in the mapping relationship, or multiple virtual machines correspond to the same processing unit. A person skilled in the art may perform settings according to specific scenario requirements, which will not be described in detail in this disclosure.

S620. The main control unit of the SoC obtains the task request sent by each virtual machine on the host machine side.

In some embodiments, referring to the system architecture shown in FIG. 2 , when a user uses the services provided by each virtual machine on the host computer side, each virtual machine sends a task to the SoC 100 through the first communication module 210 using the communication bus 300 as a medium. ask. The main control unit 110 of the SoC 100 can receive each task request through the second communication module 111.

S630. Based on the mapping relationship, the main control unit of the system on chip distributes each task request to a corresponding processing unit for processing according to the virtual machine identifier in the task request.

In some implementation manners, the process of scheduling and distributing task requests by the main control unit 110 can be referred to the above implementation manners, which will not be repeated here.

S640. Based on the mapping relationship, the main control unit sends the processing results corresponding to the task requests to the virtual machines respectively corresponding to the task requests.

In some implementations, after the processing of each task request is completed, the processing unit 120 feeds back the processing result corresponding to the task request to the main control unit 110, and then the main control unit 110 sends the processing result to the communication bus through the second communication module 11 300 is sending the media to the host, the host receives the processing result, and based on the above mapping relationship, sends the processing result to the corresponding virtual machine, so that the client (Guest) using the virtual machine can receive the processing result.

Embodiments of the present disclosure provide a system-on-chip. In some embodiments, the system-on-chip of the disclosed example may include: a main control unit and at least one processing unit; a storage unit storing a relationship between a pre-established virtual machine and each processing unit The mapping relationship among them; wherein, the main control unit is configured to obtain the task requests sent by each virtual machine, and based on the mapping relationship, distribute each task request to the corresponding processing unit for processing according to the virtual machine identifier included in the task request.

In some implementations, the system-on-a-chip of the example of the present disclosure may be shown as SoC 100 in FIG. 1 , including a main control unit 110 and a plurality of processing units 120. At the same time, the SoC 100 also includes a storage unit for storing the pre-established mapping relationship between the virtual machine and each processing unit in the above implementation manner.

For the process of realizing SoC virtualization by using the mapping relationship of the main control unit 110 , those skilled in the art can refer to the method in any of the above-mentioned implementation manners, which will not be repeated in this disclosure.

From the above, it can be seen that in the embodiments of the present disclosure, the mapping relationship between the virtual machine and the hardware resource is established, and the main control unit of the SoC is used to schedule and distribute tasks based on the mapping relationship, so as to realize the virtualization of the SoC and reduce the cost of virtualization. Improve virtual machine performance.

In some embodiments, the system-on-a-chip of the embodiment of the present disclosure further includes: a second communication module, configured to establish a communicable connection with the first communication module of the host computer, and the main control unit obtains the task request through the second communication module.

In some implementations, for any one of the virtual machines, if any one of the virtual machines corresponds to one of the processing units, the main control unit is configured to: send to any one of the virtual machines The task request is distributed to the processing unit corresponding to any virtual machine for processing.

In some implementations, for any one of the virtual machines, if any one of the virtual machines corresponds to at least two processing units in the processing units, the main control unit is configured to: for any one of the virtual machines The task request sent by the virtual machine, the main control unit splits and processes the task corresponding to the task request to obtain multiple subtasks; the number of multiple subtasks is less than or equal to the number of processing units corresponding to any virtual machine; the main control unit divides each The subtasks are distributed to one or more processing units in the processing units corresponding to any virtual machine for processing.

In some implementations, for any one of the processing units, if any one of the processing units corresponds to at least two of the virtual machines, the main control unit is configured to: obtain the any one processing A task request sent by the virtual machine corresponding to the unit; according to the priority of the virtual machine corresponding to the any one processing unit, for the task request sent by the virtual machine corresponding to the any one processing unit, determine the processing time corresponding to the task request; According to the processing time, any one of the processing units sequentially processes each task request sent by the virtual machine corresponding to the any one of the processing units.

In some implementations, the main control unit is configured to: determine the weight value corresponding to each task request sent by the virtual machine corresponding to any one of the processing units according to the preset priority of each virtual machine; based on the weight value, Determining the window processing time of each task request sent by the virtual machine corresponding to any one processing unit within the current time window; distributing the task requests sent by the at least two virtual machines corresponding to any one processing unit to tasks Queue queuing; within the current time window, any processing unit sequentially processes each task request in the task queue according to the window processing time.

In some implementations, the main control unit is configured to: in response to each task request in the task queue within the current time window, the task request is not processed within the corresponding window processing time, interrupt the task request processing process, and add the task request to the tail of the task queue.

In some implementations, the main control unit is configured to: based on the mapping relationship, send the processing result corresponding to the task request to the virtual machine corresponding to the task request.

From the above, it can be seen that the system on chip in the embodiment of the present disclosure establishes the mapping relationship between the virtual machine and the hardware resource, and uses the main control unit of the SoC to schedule and distribute tasks based on the mapping relationship, so as to realize the virtualization of the SoC and reduce the cost of virtualization And improve virtual machine performance.

The embodiment of the present disclosure provides a virtual machine task processing device, including: a host machine including at least one pre-established virtual machine; and a system-on-chip, communicably connected to the host machine.

In some implementations, the virtual machine task processing device of the example of the present disclosure may be shown in FIG. 2 , including a host machine 200 and a system-on-chip 100. The host machine 200 is pre-established with multiple virtual machines 220, and the multiple virtual machines 220 can be configured by The first communication module 210 establishes a communication connection with the SoC 100 .

For the process of implementing SoC virtualization by the system-on-chip 100 and the host machine 200 using the mapping relationship, those skilled in the art may refer to the method in any of the above-mentioned implementation manners, which will not be repeated in this disclosure.

In some embodiments, the embodiments of the present disclosure provide a storage medium storing computer-readable instructions, the computer-readable instructions are used to cause a computer to execute the virtual machine task processing method according to any one of the above-mentioned embodiments.

Apparently, the above-mentioned implementation manners are only examples for clear description, rather than limiting the implementation manners. For those of ordinary skill in the art, on the basis of the above description, other changes or changes in different forms can also be made. It is not necessary and impossible to exhaustively list all the implementation manners here. And the obvious changes or changes derived therefrom are still within the scope of protection of the present disclosure.

Claims

A virtual machine task processing method applied to a system on chip, the task processing method comprising:

Acquiring task requests sent by each virtual machine, wherein, for each task request, the task request includes a corresponding virtual machine identifier;

Based on the pre-established mapping relationship between the virtual machine and each processing unit of the SoC, the main control unit of the SoC distributes each task request to a corresponding processing unit according to the virtual machine identifier for processing.
The method according to claim 1, wherein, for any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to one of the processing units,

The main control unit of the system on chip distributes each of the task requests to corresponding processing units for processing according to the virtual machine identifier, including:

For the task request sent by the any one virtual machine, the main control unit distributes the task request to the processing unit corresponding to the any one virtual machine for processing.
The method according to claim 1, wherein, for any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to at least two of the processing units,

The main control unit of the system on chip distributes each of the task requests to corresponding processing units for processing according to the virtual machine identifier, including:

For the task request sent by any one of the virtual machines, the main control unit splits and processes the task corresponding to the task request to obtain multiple subtasks; the number of the multiple subtasks is less than or equal to that of any one of the virtual machines The number of corresponding processing units;

The main control unit distributes each subtask to one or more processing units in the processing units corresponding to any one virtual machine for processing.
The method according to claim 1, wherein, for any one of the processing units, in response to the fact that the any one processing unit corresponds to at least two virtual machines in the virtual machines,

The main control unit of the system on chip distributes each of the task requests to corresponding processing units for processing according to the virtual machine identifier, including:

The main control unit acquires a task request sent by a virtual machine corresponding to any one of the processing units;

According to the priority of the virtual machine corresponding to any one processing unit, for the task request sent by the virtual machine corresponding to any one processing unit, determine the processing time corresponding to the task request;

According to the processing time, the any one processing unit sequentially processes each task request sent by the virtual machine corresponding to the any one processing unit.
The method according to claim 4, wherein, according to the priority of the virtual machine corresponding to any one of the processing units, for the task request sent by the virtual machine corresponding to any one of the processing units, the processing time corresponding to the task request is determined ,include:

According to the preset priority of each virtual machine, determine the weight value corresponding to each task request sent by the virtual machine corresponding to any one of the processing units;

Based on the weight value, determine the window processing time of each task request sent by the virtual machine corresponding to the arbitrary processing unit within the current time window;

According to the processing time, the arbitrary processing unit sequentially processes each task request sent by the virtual machine corresponding to the arbitrary processing unit, including:

Distributing each task request sent by the virtual machine corresponding to any one of the processing units to a task queue for queuing;

Within the current time window, any one of the processing units sequentially processes each task request in the task queue according to the window processing time.
The method according to claim 5, wherein, within the current time window, any one of the processing units sequentially processes each task request in the task queue according to the window processing time, including:

In response to each task request in the task queue within the current time window, the task request is not processed within the corresponding window processing time, interrupt the processing process of the task request, and add the task request to the The tail of the task queue described above.
The method according to any one of claims 1 to 6, further comprising:

Based on the mapping relationship, the main control unit sends the processing results corresponding to the task requests to the virtual machines respectively corresponding to the task requests.
The method according to any one of claims 1 to 7, wherein the mapping relationship is stored in a memory of the system-on-chip.
A system on a chip comprising:

a main control unit and at least one processing unit;

The storage unit stores a pre-established mapping relationship between the virtual machine and each processing unit;

Wherein, the main control unit is configured to obtain task requests sent by each virtual machine, and based on the mapping relationship, distribute each task request to a corresponding processing unit for processing according to the virtual machine identifier included in the task request .
The system-on-chip according to claim 9, further comprising:

The second communication module is configured to establish a communicable connection with the first communication module of the host computer, and the main control unit obtains the task request through the second communication module.
The system on chip according to claim 9, wherein,

For any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to one of the processing units, the main control unit is configured to: for the task request sent by the any one of the virtual machines, Distributing the task request to the processing unit corresponding to any one virtual machine for processing.
The system on chip according to claim 9, wherein,

For any one of the virtual machines, in response to the fact that the any one of the virtual machines corresponds to at least two of the processing units, the main control unit is configured to: send a task for the any one of the virtual machines Request, split and process the task corresponding to the task request to obtain multiple subtasks; the number of the multiple subtasks is less than or equal to the number of processing units corresponding to any one of the virtual machines; distribute each subtask to the One or more processing units in the processing units corresponding to any virtual machine perform processing.
The system on chip according to claim 9, wherein, for any one of the processing units, in response to the fact that the any one processing unit corresponds to at least two virtual machines in the virtual machines, the main control unit is configured for:

Obtaining a task request sent by a virtual machine corresponding to any one of the processing units;

According to the priority of the virtual machine corresponding to any one processing unit, for the task request sent by the virtual machine corresponding to any one processing unit, determine the processing time corresponding to the task request;

According to the processing time, the any one processing unit sequentially processes each task request sent by the virtual machine corresponding to the any one processing unit.
The system on chip according to claim 13, wherein the main control unit is configured to:

According to the preset priority of each virtual machine, determine the weight value corresponding to each task request sent by the virtual machine corresponding to any one of the processing units;

Based on the weight value, determine the window processing time of each task request sent by the virtual machine corresponding to the arbitrary processing unit within the current time window;

Distributing each task request sent by the virtual machine corresponding to any one of the processing units to a task queue for queuing;

Within the current time window, any one of the processing units sequentially processes each task request in the task queue according to the window processing time.
The system on chip according to claim 14, wherein the main control unit is configured to:

In response to each task request in the task queue within the current time window, the task request is not processed within the corresponding window processing time, interrupt the processing process of the task request, and add the task request to the The tail of the task queue described above.
The system on chip according to claim 9, wherein the main control unit is configured to:

Based on the mapping relationship, the processing results corresponding to the task requests are sent to the virtual machines respectively corresponding to the task requests.
A virtual machine task processing device, comprising:

a host machine, including at least one pre-built virtual machine; and

The system-on-a-chip according to any one of claims 9 to 16, which is communicably connected to the host computer.
A storage medium storing computer-readable instructions for causing a computer to execute the method according to any one of claims 1-8.
A computer program product, comprising a computer program stored in a memory, when the computer program instructions are executed by a processor, the method according to any one of claims 1-8 is implemented.