Control method of computing power unit, computing power chip, electronic device, and storage medium

Abstract

The application relates to the technical field of electronic equipment, and provides a control method of a computing power unit, a computing power chip, electronic equipment and a storage medium. A control unit is connected with multiple computing power units, at least two computing power units are connected with the control unit in different manners, and the method comprises the following steps: obtaining an interaction instruction of a target computing power unit, the target computing power unit is a computing power unit in the multiple computing power units, and the interaction instruction comprises an identifier of the target computing power unit; determining a task execution environment corresponding to the target computing power unit from task execution environments corresponding to the multiple computing power units according to the interaction instruction of the target computing power unit, the task execution environment corresponding to each computing power unit being determined according to the connection manner of each computing power unit and the control unit; determining target information corresponding to the interaction instruction according to the task execution environment corresponding to the target computing power unit; and sending the target information to the target computing power unit so that the target computing power unit executes a task corresponding to the target information. The corresponding computing power unit can be called by directly sending the interaction instruction.

Description

Technical Field

[0001] This application relates to the field of electronic equipment technology, and in particular to control methods for computing units, computing chips, electronic devices, and storage media. Background Technology

[0002] A computing chip typically consists of a control unit and computing power units. The control unit is the core component of the chip, responsible for instruction reading, decoding, timing control, and process management to ensure the chip operates according to program logic. The computing power units directly determine the chip's computing power. Generally, to perform calculations on massive amounts of data, multiple computing chips need to work collaboratively.

[0003] In the operation of multiple computing chips, related technologies generally require users to use different drivers and upper-level software stacks to call computing units with different connection methods, which greatly increases the complexity of user programming. Summary of the Invention

[0004] The main objective of this application is to provide a control method, computing chip, electronic device, and storage medium for a computing power unit, which enables the direct sending of interactive commands to call the corresponding computing power unit without considering the hardware connection method between the computing power unit and the control unit and setting different interactive commands. Therefore, it is not necessary to use different drivers and upper-level software stacks to call computing power units with different connection methods.

[0005] In a first aspect, this application provides a control method for a computing power unit, applied to a control unit, wherein the control unit is connected to multiple computing power units, and at least two of the computing power units are connected to the control unit in different ways; the control method includes: Obtain the interaction command of the target computing power unit, wherein the target computing power unit is a computing power unit among the plurality of computing power units, and the interaction command includes the identifier of the target computing power unit; According to the interaction instructions of the target computing power unit, the task execution environment corresponding to the target computing power unit is determined from the task execution environments corresponding to each of the plurality of computing power units. The task execution environment corresponding to each computing power unit is determined according to the connection method between each computing power unit and the control unit. Based on the task execution environment corresponding to the target computing unit, determine the target information corresponding to the interaction instruction; The target information is sent to the target computing unit so that the target computing unit can execute the task corresponding to the target information.

[0006] Secondly, this application also provides a computing power chip, including a control unit and a computing power unit, wherein the control unit is connected to a plurality of computing power units, wherein at least two of the computing power units are connected to the control unit in different ways, and the control unit controls the computing power units through the computing power unit control method described in the first aspect above.

[0007] Thirdly, this application also provides an electronic device, including a circuit board and the computing chip described in the second aspect; the computing chip is disposed on the circuit board.

[0008] Fourthly, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the control method for the computing unit as described in the first aspect.

[0009] This application provides a control method for a computing power unit, a computing power chip, an electronic device, and a storage medium. A control unit connects to multiple computing power units, wherein at least two of the computing power units are connected to the control unit in different ways. The control method includes: acquiring an interaction instruction from a target computing power unit, wherein the target computing power unit is one of the multiple computing power units, and the interaction instruction includes an identifier of the target computing power unit; determining, based on the interaction instruction, a task execution environment corresponding to the target computing power unit from the task execution environments corresponding to each of the multiple computing power units, wherein the task execution environment corresponding to each computing power unit is determined according to the connection method between each computing power unit and the control unit; determining target information corresponding to the interaction instruction based on the task execution environment corresponding to the target computing power unit; and sending the target information to the target computing power unit so that the target computing power unit executes the task corresponding to the target information. This application first determines the task execution environment corresponding to the target computing power unit based on the identifier in the interaction instruction, then determines the target information corresponding to the interaction instruction based on the task execution environment corresponding to the target computing power unit, and then sends the target information to the target computing power unit corresponding to the interaction instruction. Thus, the target computing power unit can be called from multiple computing power units to execute tasks based on the interaction instruction. The interaction instruction does not need to consider the hardware connection method between the computing power unit and the control unit, so there is no need to use different drivers and upper-level software stacks to call computing power units with different connection methods. Attached Figure Description

[0010] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A flowchart illustrating a control method for a computing unit provided in an embodiment of this application; Figure 2 This is a schematic diagram showing the connection between the control unit and the first computing unit provided in an embodiment of this application; Figure 3 This is a schematic diagram showing the connection between the control unit and the second computing unit provided in an embodiment of this application; Figure 4 This is a schematic diagram showing the connection between the control unit and the third computing unit provided in an embodiment of this application; Figure 5 A schematic block diagram of a computing chip provided in this application embodiment; Figure 6 This is a schematic diagram showing the connection between the control unit and multiple computing units provided in an embodiment of this application; Figure 7 This is a schematic block diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] In practical applications, if multiple different connection methods are used to connect the computing unit and the control unit when building the system, users need to use different drivers and upper-level software stacks to call the computing unit corresponding to different hardware connection methods, which increases the complexity of the user process.

[0013] Taking the BM1684x chip as an example, using the computing units within the chip and inserting it as an accelerator card into a PCIe host generally requires two different drivers and upper-layer software stacks. For instance, when connecting two BM1684x chips together via the PCIe bus, the user needs to use two different drivers and runtime libraries to call the computing units within the two BM1684x chips separately, thus increasing the complexity of user programming.

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0015] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0016] This application provides a control method for a computing power unit. This control method can be applied to a control unit, which connects to multiple computing power units, wherein at least two computing power units are connected to the control unit in different ways.

[0017] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0018] Please see Figure 1 , Figure 1 This is a flowchart illustrating a control method for a computing unit provided in an embodiment of this application.

[0019] like Figure 1 As shown, the control method includes steps S101 to S104.

[0020] Step S101: Obtain the interaction instructions of the target computing power unit. The target computing power unit is a computing power unit among multiple computing power units. The interaction instructions include the identifier of the target computing power unit.

[0021] It is understandable that a computing chip consists of a control unit and a computing unit. The computing unit is responsible for the computation of the neural network. The control unit is responsible for running the upper-layer software stack to coordinate the interaction between hardware resources and software instructions. The upper-layer software stack includes a driver layer, a runtime library, a middleware layer, and an application layer. The driver layer interacts directly with the hardware and is responsible for managing the physical resources of the computing chip, such as memory. The runtime library provides high-level programming interfaces, such as APIs (Application Programming Interfaces). The middleware layer provides general services, connecting the runtime library and the application layer. The application layer directly addresses user needs, implementing specific business logic, such as interacting with the user to obtain user input commands.

[0022] It should be noted that APIs allow interaction between different software components. Specifically, the runtime library can use APIs to enumerate the number of computing units connected to the control unit and display the identifiers and types of the computing units at the application layer.

[0023] The control unit in this embodiment can connect to multiple computing power units, such as GPUs, TPUs, NPUs, and FPGAs. The control unit can obtain user-inputted interactive commands for different computing power units by running the application layer in the upper-level software stack. For example, each computing power unit has a unique identifier. The interactive commands for each computing power unit can include the identifier of the computing power unit.

[0024] Step S102: Based on the interaction instructions of the target computing power unit, determine the task execution environment corresponding to the target computing power unit from the task execution environments corresponding to the multiple computing power units. The task execution environment corresponding to each computing power unit is determined according to the connection method between each computing power unit and the control unit.

[0025] It is understandable that the drivers in the driver layer of the upper software stack mainly manage the instantiation of hardware resources. Specifically, the drivers corresponding to the connection method between the computing unit and the control unit can create a task execution environment, i.e., an instance, for each computing unit when the system starts up (hot-swappable).

[0026] The specific process of creating a task execution environment is as follows: In the first stage, hardware detection and initialization are performed. For plug-in devices, the PCIe bus is scanned during startup to identify the computing chip in the connected plug-in device; for embedded devices, their configuration is described through a device tree. In the second stage, the task execution environment is created and resources are allocated. The driver allocates device memory for each computing unit to store instructions, data, and context, and also registers interrupt handling functions for the corresponding task execution environment of each computing unit to respond to hardware events such as computation completion and error interrupts. In the third stage, management is performed. The driver saves the hardware state of each computing unit, such as register values ​​and cache contents, to support fast context switching.

[0027] Furthermore, the hardware connection methods between computing units and control units differ, and therefore the corresponding drivers also differ. After creating a task execution environment for each computing unit, the driver can configure a corresponding identifier for each computing unit through a global counter. Therefore, after obtaining the interaction instructions from the target computing unit, the task execution environment corresponding to the target computing unit can be determined from the task execution environments corresponding to the multiple computing units based on the identifier in the interaction instructions.

[0028] Step S103: Determine the target information corresponding to the interaction instruction based on the task execution environment corresponding to the target computing unit.

[0029] It is understood that the task execution environment corresponding to each computing power unit includes a message sending component for sending messages to the corresponding computing power unit. In this embodiment, data from the interaction instructions of the target computing power unit can be input into the message sending component in the corresponding task execution environment to integrate and obtain the target information to be sent to the target computing power unit.

[0030] Specifically, the interaction instructions may include identifiers and information related to the execution of the task. After the interaction instructions are input into the message sending component in the task execution environment corresponding to the target computing unit, the target information suitable for sending to the target computing unit can be determined.

[0031] Step S104: Send the target information to the target computing unit so that the target computing unit can execute the task corresponding to the target information.

[0032] After obtaining the target information corresponding to the interactive command, the target information is sent to the target computing unit corresponding to the identifier in the interactive command, so as to call the target computing unit to execute the task corresponding to the target information. This enables the computing unit corresponding to the identifier to be called directly according to the identifier in the interactive command, without having to consider the hardware connection method between the computing unit and the control unit. Therefore, it is also not necessary to use different drivers and upper-level software stacks to call computing units with different connection methods.

[0033] The control method for computing power units provided in the above embodiments is applied to a control unit. The control unit connects to multiple computing power units, wherein at least two computing power units are connected to the control unit in different ways. The control method includes: acquiring an interaction instruction from a target computing power unit, wherein the target computing power unit is one of the multiple computing power units, and the interaction instruction includes an identifier of the target computing power unit; determining the task execution environment corresponding to the target computing power unit from the task execution environments corresponding to each of the multiple computing power units based on the interaction instruction of the target computing power unit, wherein the task execution environment corresponding to each computing power unit is determined according to the connection method between each computing power unit and the control unit; determining the target information corresponding to the interaction instruction based on the task execution environment corresponding to the target computing power unit; and sending the target information to the target computing power unit so that the target computing power unit executes the task corresponding to the target information. In this embodiment, the task execution environment corresponding to the target computing power unit is first determined based on the identifier in the interaction instruction. Then, the target information corresponding to the interaction instruction is determined based on the task execution environment corresponding to the target computing power unit. Finally, the target information is sent to the target computing power unit corresponding to the interaction instruction. Thus, the target computing power unit can be called from multiple computing power units to execute tasks based on the interaction instruction. The interaction instruction does not need to consider the hardware connection method between the computing power unit and the control unit, so there is no need to use different drivers and upper-level software stacks to call computing power units with different connection methods.

[0034] In one exemplary implementation, such as Figures 2 to 4 As shown, the multiple computing units include at least two of the following: a first computing unit 120, a second computing unit 130, and a third computing unit 140. The first computing unit 120 is integrated with the control unit 110 on the same chip and is connected to the control unit 110 via a first type bus. The second computing unit 130 is integrated on the first chip, and the control unit 110 is integrated on the second chip; the first chip and the second chip are connected via a second type bus. The third computing unit 140 is disposed on the first device, and the control unit 110 is disposed on the second device; the first device and the second device are connected via a preset network protocol.

[0035] Understandably, the first computing unit 120 and the control unit 110 are integrated on the same chip and connected to the control unit 110 via an AMBA bus. The second computing unit 130 is integrated on the first chip, and the control unit 110 is integrated on the second chip. The first and second chips are connected via a PCIe bus, enabling the second computing unit 130 on the first chip to connect with the control unit 110 on the second chip. The third computing unit 140 is located on the first device, and the control unit 110 is located on the second device. The first and second devices are connected via Ethernet, enabling the third computing unit 140 on the first device to connect with the control unit 110 on the second device.

[0036] In this application embodiment, when multiple computing chips are used in combination, the control unit of one of the computing chips can be connected to the computing unit of the same chip, the computing units of different computing chips on the same device, and the computing units on other devices to realize the calculation of massive amounts of data.

[0037] In one exemplary embodiment, step S103 includes steps S1031 and S1032.

[0038] Step S1031: Obtain the message sending component corresponding to the interaction instruction from the task execution environment corresponding to the target computing unit.

[0039] Step S1032: Determine the target information corresponding to the interaction instruction based on the message sending component and the interaction instruction.

[0040] For example, different connection methods between computing units and control units correspond to different drivers. The driver corresponding to the connection method between the computing unit and the control unit can create a task execution environment (instance) for each computing unit when the system starts up (hot-plugged). For example, the driver corresponding to the computing unit connected to the control unit via the PCIe bus, i.e., the PCIe driver, can create a task execution environment for each computing unit connected to the control unit via the PCIe bus, and the task execution environment corresponding to each connection method can include a corresponding message sending component.

[0041] Specifically, the driver for each physical connection method can define a set of function pointers. Among them, the function pointer used to send information, i.e., the message sending component, is named "send". For example, the driver for the computing unit that connects to the control unit via the PCIe bus is a PCIe driver, and the message sending component defined by this PCIe driver is named "pcie_send".

[0042] The interaction instructions in this embodiment may include information related to task execution, such as identifiers, types, task flow numbers, and task numbers. After determining the corresponding task execution environment based on the identifier of the target computing unit, a message sending component is obtained from the task execution environment. The data carried by the interaction instructions is then input into the message sending component, which integrates the data carried by the interaction instructions to obtain the corresponding target information.

[0043] In one exemplary embodiment, step S104 specifically includes sending target information to the target computing unit through a message sending component, based on the connection method between the target computing unit and the control unit.

[0044] Understandably, after determining the task execution environment corresponding to the target computing unit based on the identifier in the interaction instructions, the target information can be sent to the target computing unit through the message sending component in the task execution environment corresponding to the target computing unit. Specifically, the message sending component can send the target information to the target computing unit using either ring buffer transmission or socket transmission, depending on the connection method between the target computing unit and the control unit. Ring buffer transmission is suitable for high-performance scenarios within a process or shared memory. Socket transmission is suitable for scenarios involving cross-process or cross-network communication.

[0045] In one exemplary implementation, the message sending components in the task execution environment corresponding to each computing unit are of the same type.

[0046] In the embodiments of this application, the name of the message sending component defined by the driver for sending information is "send". That is, each computing unit uses the same communication API. Therefore, there is no need to consider the hardware connection method between each computing unit and the control unit, and thus there is no need to use different drivers and upper-level software stacks to call computing units with different connection methods.

[0047] In one exemplary implementation, the target information includes the field contents corresponding to the number field, type field, task field, and data field.

[0048] Understandably, the target information transmitted between the control unit and the computing unit can be sent according to a preset information format. For example, the target information may include the content of each of the following fields: number field, type field, task field, and data field. The number field includes the identifier of the computing unit. The type field includes task operations such as matrix calculation tasks and data transfer tasks. The task field may include a task flow number field and a task number field; based on the content of these fields, a task can be identified between the control unit and all computing units. The data field may include an address field, a function name field, a module name field, and a task data field. Different task types require different information sent to the computing unit; for example, a data transfer task requires the content of the address field, while a matrix calculation task requires the content of the function name field and the corresponding address field.

[0049] Furthermore, if some fields in the transmitted target information lack corresponding content, zeros can be filled in the corresponding fields to avoid affecting the accuracy of the transmitted target information.

[0050] In this embodiment of the application, after the data carried by the interactive instruction is input into the message sending component, the target information can be determined by the message sending component according to the preset information format, so that the target information includes the field content corresponding to the number field, type field, task field and data field, thereby ensuring that the target information transmitted between the control unit and the computing unit is in the preset information format.

[0051] In one exemplary embodiment, step S104 is followed by steps S105 and S106.

[0052] Step S105: Obtain the task completion information sent by the target computing unit.

[0053] Step S106: Determine the result of the target computing unit executing the task corresponding to the target information based on the task completion information.

[0054] Understandably, the target computing unit can send task completion information to the control unit after executing its task. This task completion information includes the result of the target computing unit executing the task corresponding to the target information, so that the control unit can determine the result of the target computing unit executing the task corresponding to the target information based on the task completion information, such as success, failure, abort, and interruption.

[0055] Please see Figure 5 , Figure 5 This is a schematic block diagram of a computing chip provided in an embodiment of this application.

[0056] like Figure 5As shown, the computing chip 100 in this embodiment includes a control unit 110 and a computing unit. The control unit 110 is connected to multiple computing units. At least two computing units are connected to the control unit 110 in different ways, and the first computing unit 120 and the control unit 110 are integrated on the same computing chip 100.

[0057] It is understood that the control unit 110 in this application embodiment can control multiple computing units connected to it through the control method of the computing unit in the above embodiment.

[0058] For example, such as Figure 6 As shown, the first computing unit 120 and the second computing unit 130 are both located on the local device, and the first computing unit 120 and the second computing unit 130 are respectively located on different computing chips. The third computing unit 140 is located on the remote device. The control unit 110 is connected to the first computing unit 120 integrated with it on the same computing chip via an AMBA bus; it is connected to the computing chip where the second computing unit 130 is located via a PCIe bus, thereby connecting to the second computing unit 130; and it is connected to the remote device via Ethernet, thereby connecting to the third computing unit 140 located on the remote device.

[0059] The computing chip in this application embodiment has all the technical effects of the control method of the computing unit described above. That is, it realizes that the corresponding computing unit can be called by directly sending interactive instructions, without having to consider the hardware connection method between the computing unit and the control unit and set different interactive instructions, so that different drivers and upper-level software stacks are not needed to call computing units with different connection methods.

[0060] Please see Figure 7 , Figure 7 This is a schematic block diagram of the structure of an electronic device provided in an embodiment of this application.

[0061] like Figure 7 As shown, the electronic device 300 of this application embodiment includes a circuit board 200 and a computing chip 100 as described in the above embodiment; the computing chip 100 is disposed on the circuit board 200.

[0062] It is understood that the electronic device 300 of this application embodiment has all the technical effects of the computing chip described above.

[0063] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the following steps: Obtain the interaction command for the target computing power unit, which is a computing power unit among multiple computing power units. The interaction command includes the identifier of the target computing power unit. Based on the interaction instructions of the target computing power unit, the task execution environment corresponding to the target computing power unit is determined from the task execution environments corresponding to the multiple computing power units. The task execution environment corresponding to each computing power unit is determined according to the connection method between each computing power unit and the control unit. Based on the task execution environment corresponding to the target computing unit, determine the target information corresponding to the interaction command; The target information is sent to the target computing unit so that the target computing unit can execute the task corresponding to the target information.

[0064] The computer-readable storage medium can be an internal storage unit of the computer device described in the foregoing embodiments, such as a hard disk or memory of the computer device. Alternatively, it can be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., provided on the computer device.

[0065] It should be noted that the functions or steps that the computer-readable storage medium can achieve as described above can be referred to the embodiments of the aforementioned control method for computing power units.

[0066] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0067] It should also be understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. It should be noted that, herein, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0068] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0069] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0070] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A control method for a computing unit, characterized in that, The control method is applied to a control unit, which is connected to multiple computing units, wherein at least two of the computing units are connected to the control unit in different ways. Obtain the interaction command of the target computing power unit, wherein the target computing power unit is a computing power unit among a plurality of computing power units, and the interaction command includes the identifier of the target computing power unit; Based on the interaction instructions of the target computing power unit, the task execution environment corresponding to the target computing power unit is determined from the task execution environments corresponding to each of the multiple computing power units. The task execution environment corresponding to each computing power unit is determined according to the connection method between each computing power unit and the control unit. The message sending component corresponding to the interaction instruction is obtained from the task execution environment corresponding to the target computing power unit. The type of message sending component in the task execution environment corresponding to each computing power unit is the same. Based on the message sending component and the interaction instruction, determine the target information corresponding to the interaction instruction; The target information is sent to the target computing unit so that the target computing unit can execute the task corresponding to the target information.

2. The control method for the computing unit according to claim 1, characterized in that, Sending the target information to the target computing unit includes: The target information is sent to the target computing unit via the message sending component, based on the connection method between the target computing unit and the control unit.

3. The control method for the computing unit according to any one of claims 1 to 2, characterized in that, The target information includes the content of the corresponding fields for the number field, type field, task field, and data field.

4. The control method for the computing unit according to any one of claims 1 to 2, characterized in that, After the target computing unit executes the task corresponding to the target information, the process further includes: Obtain the task completion information sent by the target computing unit; The result of the target computing unit executing the task corresponding to the target information is determined based on the task completion information.

5. The control method for the computing unit according to any one of claims 1 to 2, characterized in that, The plurality of computing units include at least two of the following: The first computing unit is integrated on the same chip as the control unit and is connected to the control unit via a first type of bus. The second computing unit is integrated on the first chip, the control unit is integrated on the second chip, and the first chip and the second chip are connected through a second type of bus; The third computing unit is located on the first device, and the control unit is located on the second device. The first device and the second device are connected through a preset network protocol.

6. A computing chip, characterized in that, The device includes a control unit and a computing power unit. The control unit is connected to multiple computing power units, wherein at least two of the computing power units are connected to the control unit in different ways, and the control unit controls the computing power units using the control method for computing power units as described in any one of claims 1 to 5.

7. An electronic device, characterized in that, It includes a circuit board and the computing chip as described in claim 6; the computing chip is disposed on the circuit board.

8. A computer-readable storage medium storing a computer program thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the control method for the computing unit as described in any one of claims 1 to 5.

Images