WO2022111456A1

WO2022111456A1 - Core sharing method and apparatus based on many-core system, electronic device, and medium

Info

Publication number: WO2022111456A1
Application number: PCT/CN2021/132357
Authority: WO
Inventors: 吴臻志; 祝夭龙
Original assignee: 北京灵汐科技有限公司
Priority date: 2020-11-24
Filing date: 2021-11-23
Publication date: 2022-06-02
Also published as: CN114546493A

Abstract

The present disclosure provides a core sharing method and apparatus based on a many-core system, an electronic device, and a medium. The many-core system comprises pre-configured multiple core clusters, and each of the core clusters comprises at least one second processing core. The method comprises: receiving a core sharing request; and determining, according to the core sharing request, at least one second processing core from a first core cluster and/or a predetermined second processing core of a second core cluster to be a shared core between the first core cluster and the second core cluster, and at least one shared core constituting a shared core array, wherein the shared core array is configured to receive and execute tasks respectively corresponding to the first core cluster and the second core cluster, store first task data corresponding to the task of the first core cluster, and store second task data corresponding to the task of the second core cluster.

Description

Core sharing method and device, electronic device and medium based on many-core system

technical field

The present disclosure relates to the field of computer technologies, and in particular, to a core sharing method and apparatus based on a many-core system, an electronic device, and a computer-readable medium.

Background technique

With the development of artificial intelligence technology, the demand for data processing speed is increasing day by day, making many-core systems more and more widely used. A many-core system usually has many cores (also called processing cores). The core is the smallest computing unit in the many-core system that can be independently scheduled and has complete computing power. The core has certain resources such as storage and computing. The core of the many-core system can run program instructions independently, using the ability of parallel computing, can speed up the running speed of the program, and can provide multi-tasking ability.

SUMMARY OF THE INVENTION

The present disclosure provides a core sharing method and apparatus based on a many-core system, an electronic device, and a computer-readable medium.

In a first aspect, the present disclosure provides a core sharing method based on a many-core system, the many-core system includes a plurality of pre-configured core clusters, each of the core clusters includes at least one second processing core, the method Including: receiving a core sharing request; according to the core sharing request, determining at least one second processing core from the second processing cores of the first core cluster and/or the predetermined second core cluster, as the first core cluster A shared core between a core cluster and the second core cluster, at least one shared core constitutes a shared core array; wherein the shared core array is configured to receive and execute communication with the first core cluster and the second core cluster The tasks corresponding to the core clusters respectively store first task data corresponding to the tasks of the first core cluster, and store second task data corresponding to the tasks of the second core cluster.

In some embodiments, the many-core system further includes a first processing core, and the receiving a core sharing request includes: receiving a core sharing request sent by an external host or a first processing core, where the core sharing request is the external Generated by the host or the first processing core based on common tasks corresponding to the first core cluster and the second core cluster; or, receiving a core sharing request sent by the main processing core of the second core cluster, The main processing core of the second core cluster is a pre-designated second processing core in at least one second processing core of the second core cluster.

In some embodiments, after the at least one second processing core is determined, the method further includes: updating a core cluster list corresponding to the shared core, where the updated core cluster list includes Identification information corresponding to a core cluster and the second core cluster respectively.

In some embodiments, after the at least one second processing core is determined, the method further includes: creating a first shared core corresponding to the first core cluster in a storage area corresponding to the shared core array a storage area, the first shared storage area is used to store the first task data; a second shared storage area corresponding to the second core cluster is created in the storage area corresponding to the shared core array, the The second shared storage area is used to store the second task data.

In some embodiments, before creating the first shared storage area corresponding to the first core cluster in the storage area corresponding to the shared core array, the method further includes: judging the storage area corresponding to the shared core array Whether there is available storage space in the storage area; in response to judging that there is available storage space in the storage area corresponding to the shared core array, executing the creation of the first core cluster in the storage area corresponding to the shared core array Corresponding steps of the first shared storage area.

In some embodiments, after determining whether there is available storage space in the storage area corresponding to the shared core array, the method further includes: in response to determining that the storage area corresponding to the shared core array does not exist Available storage space, determine an idle processing core in the processing cores that do not belong to any core cluster in the many-core system; determine at least one of the idle processing cores as the shared core, and add it to the shared core array .

In some embodiments, the number of the shared cores is multiple, the storage area corresponding to the shared core array further includes a general shared storage area, the first shared storage area, the second shared storage area and the The general shared storage area corresponds to one or more shared cores respectively; wherein, the shared core corresponding to the first shared storage area is configured to store the first task data; the shared core corresponding to the second shared storage area is configured In order to store the second task data; the shared cores corresponding to the common shared storage area are configured to receive and execute tasks corresponding to the first core cluster and the second core cluster respectively.

In some embodiments, the first task data includes first input data, first output data, and first task configuration information, and the first shared storage area includes a first input data area, a first output data area, and a first a task configuration area; wherein the shared cores in the first input data area are configured to store the first input data; the shared cores in the first output data area are configured to store the first output data ; the shared core in the first task configuration area is configured to store the first task configuration information.

In a second aspect, the present disclosure provides a core sharing apparatus, which is applied to a many-core system, where the many-core system includes a plurality of preconfigured core clusters, each core cluster includes at least one second processing core, the The core sharing apparatus includes: a receiving module for receiving a core sharing request; a sharing determining module for, according to the core sharing request, from the second processing cores of the first core cluster and/or the predetermined second core cluster, determining at least one second processing core as a shared core between the first core cluster and the second core cluster, and at least one shared core forms a shared core array; wherein the shared core array is configured to receive and execute the tasks corresponding to the first core cluster and the second core cluster respectively, store the first task data corresponding to the tasks of the first core cluster, and store the tasks corresponding to the second core cluster of the second task data.

In a third aspect, the present disclosure provides an electronic device comprising: a plurality of processing cores; and an on-chip network configured to exchange data among the plurality of processing cores and external data; wherein one or more One or more instructions are stored in each of the processing cores, and the one or more instructions are executed by the one or more processing cores, so that the one or more processing cores can execute the above-mentioned core sharing method.

In a fourth aspect, the present disclosure provides a computer-readable medium on which a computer program is stored, the computer program implementing the above-mentioned core sharing method when executed by a processing core.

In a fifth aspect, the present disclosure provides a computer program product that, when running on a computer, causes the computer to execute the above-mentioned core sharing method.

According to the embodiments of the present disclosure, at least one second processing core in one or more core clusters can be used as a shared core among multiple core clusters according to the received core sharing request, and multiple core clusters can share the shared core In order to realize the cross-cluster business between multiple core clusters in the many-core system, meet the needs of multiple core clusters sharing the data of some processing cores, improve the control ability of some processing cores in the many-core system, and improve the many-core system. Flexibility for task processing.

It should be understood that what is described in this section is not intended to identify key or critical features of embodiments of the disclosure, nor is it intended to limit the scope of the disclosure. Other features of the present disclosure will become readily understood from the following description.

Description of drawings

The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification, and together with the embodiments of the present disclosure, they are used to explain the present disclosure, and are not intended to limit the present disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings, in which:

1 is a flowchart of a core sharing method based on a many-core system according to an embodiment of the present disclosure;

FIG. 2 is a block diagram of the composition of a many-core system according to an embodiment of the present disclosure;

FIG. 3 is a flowchart of a core sharing method according to an embodiment of the present disclosure;

FIG. 4 is a flowchart of a core sharing method according to an embodiment of the present disclosure;

FIG. 5 is a schematic structural diagram of a shared core array according to an embodiment of the disclosure;

FIG. 6 is a block diagram of the composition of a core sharing apparatus according to an embodiment of the present disclosure;

FIG. 7 is a block diagram of the composition of an electronic device according to an embodiment of the present disclosure.

Detailed ways

The present disclosure will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure, but not to limit the present disclosure. In addition, it should be noted that, for the convenience of description, the drawings only show some but not all structures related to the present disclosure.

In order for those skilled in the art to better understand the technical solutions of the present disclosure, the exemplary embodiments of the present disclosure are described below with reference to the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, and they should be considered to be exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted from the following description for clarity and conciseness.

Various embodiments of the present disclosure and various features of the embodiments may be combined with each other without conflict.

As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

The terminology used herein is used to describe particular embodiments only and is not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are intended to include the plural forms as well, unless the context clearly dictates otherwise. It will also be understood that when the terms "comprising" and/or "made of" are used in this specification, the stated features, integers, steps, operations, elements and/or components are specified to be present, but not precluded or Add one or more other features, integers, steps, operations, elements, components, and/or groups thereof. Words like "connected" or "connected" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect.

Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in common dictionaries should be construed as having meanings consistent with their meanings in the context of the related art and the present disclosure, and will not be construed as having idealized or over-formal meanings, unless expressly so limited herein.

In the embodiment of the present disclosure, the many-core system includes a plurality of preconfigured core clusters, each core cluster includes at least one second processing core, and each core cluster is respectively used for executing corresponding computing tasks. During the process of each core cluster performing the corresponding task, there may exist data and parameters (eg weights) of some processing cores, which are required by multiple core clusters at the same time. As an example, in a deep learning network, for the processing core corresponding to the backbone network and the processing core corresponding to the branch network, the processing core corresponding to the branch network needs to use the data of the processing core corresponding to the backbone network to perform prediction processing. When the processing cores corresponding to the branch network and the backbone network are located in different core clusters, how to effectively implement cross-cluster services between core clusters is a technical problem that needs to be solved urgently in the core cluster scenario of the many-core system.

FIG. 1 is a flowchart of a core sharing method based on a many-core system according to an embodiment of the present disclosure.

1 , an embodiment of the present disclosure provides a core sharing method based on a many-core system, wherein the many-core system includes a plurality of pre-configured core clusters, each core cluster includes at least one second processing core, and the method can be implemented by a core To execute by sharing a device, the device can be implemented by means of software and/or hardware, and the core sharing method includes:

Step S1, receiving a core sharing request.

Step S2, according to the core sharing request, determine at least one second processing core from the second processing cores of the first core cluster and/or the predetermined second core cluster, as the first core cluster and the second core cluster. Shared cores between core clusters, at least one shared core forms a shared core array.

The shared core array is configured to receive and execute tasks corresponding to the first core cluster and the second core cluster respectively, store first task data corresponding to the tasks of the first core cluster, and store tasks corresponding to the second core cluster Corresponding second task data. The first task data includes first input data, first output data, and first task configuration information, and the second task data includes second input data, second output data, and second task configuration information.

The input data can be, for example, the data required to execute the task, the output data can be, for example, the result of executing the task, and the task configuration information is the configuration required to execute the task, such as memory configuration, routing configuration, operation sequence configuration, operation Time configuration, operator configuration and other information. The present disclosure does not limit the specific categories of input data, output data, and task configuration information.

In the embodiment of the present disclosure, the first core cluster is any core cluster among the multiple core clusters in the many-core system, the second core cluster is any core cluster other than the first core cluster in the many-core system, and the second core cluster is any core cluster other than the first core cluster in the many-core system. The number of clusters can be one or more.

In the core sharing method based on the many-core system provided by the embodiments of the present disclosure, in the business scenario where the many-core system includes multiple core clusters, for multiple core clusters that have cross-cluster business requirements, any core cluster or multiple core clusters can be used for multiple core clusters. At least one second processing core in the core clusters is used as a shared core among multiple core clusters, and multiple core clusters can share the data of the shared core (for example, parameters required in the deep learning network, etc.), thereby realizing multiple core clusters. The cross-cluster business between multiple core clusters of the core system satisfies the requirement that multiple core clusters need to share the data of some processing cores, improves the control ability of some processing cores in the many-core system, and improves the task processing of the many-core system. flexibility.

FIG. 2 is a block diagram of the composition of a many-core system provided by an embodiment of the present disclosure. In an embodiment of the present disclosure, referring to FIG. 2 , each core cluster has a main processing core, and the main processing core is the core cluster. A pre-designated second processing core among the at least one second processing core. The core sharing method of the embodiment of the present disclosure can be applied to the main processing core of any core cluster in the many-core system, that is, the core sharing method of the embodiment of the present disclosure is implemented based on the main processing core of any core cluster, and the main processing core can be used for It handles the tasks corresponding to its own core cluster, and can also be used for task allocation and management within the cluster.

In some embodiments, when at least one second core cluster in the many-core system needs to use data (eg, parameters) of one or more second processing cores in the first core cluster, the main processing core of the second core cluster may Send a core sharing request to the main processing core of the first core cluster. The core sharing request may include but is not limited to: identification information, address information corresponding to the second core cluster, one or more second core clusters to be shared in the first core cluster Processing core information, etc. In this case, in step S1, the main processing core of the first core cluster receives the core sharing request sent by the main processing core of the second core cluster, wherein the main processing core of the second core cluster is at least one of the second core cluster. A pre-designated second processing core among the second processing cores.

In some embodiments, referring to FIG. 2 , the many-core system includes a plurality of processing cores, wherein one processing core is pre-designated as the first processing core, and the remaining processing cores are the second processing cores, and the first processing core is used for receiving data from The tasks to be processed sent by the external host, and the core cluster for processing the tasks to be processed is formed in the many-core system, and the tasks to be processed are scheduled and managed in the many-core system.

When the external host or the first processing core needs to establish or assign a common task to multiple core clusters, the external host or the first processing core can select any one of the multiple core clusters as the first core cluster, and the rest As the core cluster is the second core cluster, the external host or the first processing core can first send a core sharing request to the main processing core of the first core cluster. The core sharing request may include but not limited to common tasks and the identifier corresponding to the second core cluster. information, address information, etc., so that the main processing cores of the first core cluster form shared cores for processing common tasks.

In this case, in step S1, the main processing core of the first core cluster receives the core sharing request sent by the external host or the first processing core, and the core sharing request is based on the external host or the first processing core and the first core cluster and the core sharing request. It is generated by the public tasks corresponding to the second core cluster.

Among them, common tasks refer to tasks that both the first core cluster and the second core cluster need to perform. For example, the first core cluster needs to perform fast Fourier transform tasks (FFT) and window function filtering tasks, and the second core cluster needs to perform fast Fourier transform task (FFT) and frequency domain multiplication task, the fast Fourier transform task (FFT) is the common task corresponding to the first core cluster and the second core cluster, for example, the common task can be the backbone of the neural network Complex mathematical operations (such as matrix factorization, FFT, etc.) tasks in the network.

In some embodiments, in step S2, after receiving the core sharing request from the main processing core of the second core cluster, the main processing core of the first core cluster, according to the core sharing request, from the first core where it is located. At least one second processing core is determined from at least one second processing core of the cluster, the determined second processing core is used as a shared core, and the at least one shared core forms a shared core array. The number of shared cores may be specifically determined according to the number of second processing cores requested to be shared by the main processing cores of the second core cluster, which is not limited in this embodiment of the present disclosure.

In this way, cross-cluster services between multiple core clusters are realized, the data requirements of multiple core clusters sharing some processing cores are met, and some processing cores in the many-core system (such as the main processing core of the core cluster) are improved. ) and the flexibility of task processing for each cluster.

In some embodiments, in step S2, after receiving the core sharing request from the external host or the first processing core in the many-core system, the main processing core of the first core cluster, At least one second processing core is determined in at least one second processing core of the first core cluster, the determined second processing core is used as a shared core, and the at least one shared core forms a shared core array. The number of shared cores may be determined according to the resource requirements required by the common tasks, which is not limited in the embodiments of the present disclosure, as long as the total remaining resources of the formed shared cores can meet (greater than or equal to) the required resources for the common tasks. resource requirements.

The resource requirement required by the common task refers to the amount of resources required to perform the common task, for example, the resources may be computing resources, storage resources, and bandwidth resources. For multiple core clusters that need to process common tasks, by forming shared cores among multiple core clusters to process common tasks, the processing efficiency and flexibility of the tasks are improved, and multiple core clusters of the many-core system are realized at the same time. The cross-cluster business between multiple core clusters satisfies the requirement that multiple core clusters need to share some processing core data.

In some embodiments, the main processing core of the first core cluster selects an idle second processing core as a shared core. In some embodiments, in the case of receiving a core sharing request from the main processing core of the second core cluster, if there is no idle second processing core in the first core cluster where the second core cluster is located, the request to the second core cluster The main processing core returns a core sharing request failure message. In some embodiments, in the case of receiving a core sharing request from the external host or the first processing core, if there is no idle second processing core in the first core cluster where the user is located, the external host or the first processing core is sent to the external host or the first processing core. The core returns a core share request failure message.

In some embodiments, in step S2, after receiving the core sharing request from the external host or the first processing core in the many-core system, the primary processing core of the first core cluster, according to the core sharing request, from the second core At least one second processing core is determined from at least one second processing core of the cluster, and the determined second processing core serves as a shared core between the first core cluster and the second core cluster.

In some embodiments, in step S2, after receiving the core sharing request from the external host or the first processing core in the many-core system, the main processing core of the first core cluster, according to the core sharing request, There is no available (idle) second processing core in the first core cluster, then at least one second processing core is determined from at least one second processing core in the second core cluster, and the determined second processing core is used as the first processing core. A shared core between a core cluster and a second core cluster.

For example, determining at least one second processing core from the at least one second processing core of the second core cluster as a shared core includes: sending a core sharing request to the main processing core of the second core cluster for the first In response to the core sharing request, the main processing core of the two-core cluster determines at least one second processing core from at least one second processing core in the core cluster where it is located as a shared core.

It should be noted that, in the embodiments of the present disclosure, “first” and “second” in the first core cluster and the second core cluster are only for distinguishing different core clusters, and do not refer to a particular core cluster. In the core system, all the core clusters have the same status, can receive the core sharing request as the first core cluster, and execute the core sharing method in response to the core sharing request, and each core cluster in all the core clusters can also be used as the second core. The cluster requests core sharing from other core clusters, which is not limited in this embodiment of the present disclosure.

In some embodiments, after receiving the core sharing request from the external host or the first processing core in the many-core system, the main processing core of the first core cluster, according to the core sharing request, from at least one of the first core cluster where it is located. At least one second processing core is determined from one second processing core as a shared core, and at least one second processing core is determined from at least one second processing core of the second core cluster as a shared core. In other words, a part of the second processing cores can be selected from the first core cluster, and a part of the second processing cores can be selected from the second core cluster as shared cores together.

In some embodiments, after determining the second processing core as the shared core, the main processing core of the first core cluster or the second core cluster further marks the second processing core to mark the second processing core It is shared for the management of each core in the cluster.

FIG. 3 is a flowchart of a core sharing method according to an embodiment of the present disclosure.

3, in order to facilitate management, in some embodiments, after the at least one second processing core is determined in step S2, the core sharing method further includes:

Step S3: Update the core cluster list corresponding to the shared core, where the updated core cluster list includes identification information corresponding to the first core cluster and the second core cluster respectively.

As an example, before sharing, each second processing core in each cluster prestores a core cluster list corresponding to the core cluster where the second processing core is located, and the core cluster list includes but is not limited to: the core cluster corresponding to the core cluster where the second processing core is located Identification information, information of each member in the cluster, the information of each member in the cluster includes but not limited to: the identification, address, location, etc. of each member.

As an example, after determining the shared core, the main processing core of the first core cluster adds the identification information corresponding to the second core cluster in the core cluster list of each shared core, thereby updating the core cluster list corresponding to each shared core, and updating The latter list of core clusters includes identification information corresponding to the first core cluster and the second core cluster, so that the shared core can identify that subsequent jobs belong to the first core cluster or the second core cluster, and perform corresponding processing.

In some embodiments, after determining the shared core, the main processing core of the first core cluster also feeds back information of the shared core to the second core cluster, the information including but not limited to the identification and address information of the shared core, so as to facilitate The processing cores in the second core cluster can access the shared core to obtain required data, or implement required services through the shared core.

In some embodiments, in the case of receiving the core sharing request sent by the external host or the first processing core, if the main processing core of the first core cluster sends the core sharing request to the main processing core of the second core cluster, the After determining the shared core, the main processing core of the two-core cluster also feeds back the information of the shared core to the first core cluster, the information including but not limited to the identification and address information of the shared core, so as to facilitate the processing in the first core cluster The cores can access the shared cores to obtain the required data, or implement the required services through the shared cores.

FIG. 4 is a flowchart of a core sharing method provided by an embodiment of the present disclosure.

Referring to FIG. 4 , in some embodiments, in order to effectively solve the problem of conflict between the writing and writing of data of different core clusters by the shared core, after the at least one second processing core is determined in step S2, the core sharing method further includes: :

Step S4 , judging whether there is available storage space in the storage area corresponding to the shared core array, if so, go to step S5 , otherwise go to step S7 .

In some embodiments, after it is determined that there is available storage space in the storage area corresponding to the shared core array, it can be further determined whether the memory capacity of the available storage space is greater than or equal to the preset memory capacity, if greater than or equal to If the memory capacity is preset, step S5 is performed; otherwise, step S7 is performed. The storage area corresponding to the shared core array is the storage area of all shared cores in the shared core array.

The preset memory capacity may be the total memory capacity required for executing common tasks and storing task data of the common tasks; it may also be executing tasks corresponding to the first core cluster and the second core cluster, and storing The total memory capacity required for the first task data and the second task data. The present disclosure does not limit the specific value of the preset memory capacity.

Step S5: Create a first shared storage area corresponding to the first core cluster in the storage area corresponding to the shared core array.

It can be understood that creating the first shared storage area corresponding to the first core cluster in the storage area corresponding to the shared core array is to create the first shared storage area in the available storage space of the shared core array.

In step S5, in response to determining that there is available storage space in the storage area corresponding to the shared core array, a first shared storage area corresponding to the first core cluster is created in the storage area corresponding to the shared core array. In some embodiments, in response to determining that there is available storage space in the storage area corresponding to the shared core array, and determining that the memory capacity of the available storage space is greater than or equal to the preset memory capacity, the storage area corresponding to the shared core array A first shared storage area corresponding to the first core cluster is created in the area. The first shared storage area is used to store first task data corresponding to tasks of the first core cluster.

Step S6: Create a second shared storage area corresponding to the second core cluster in the storage area corresponding to the shared core array, and end the process.

It can be understood that creating the second shared storage area corresponding to the second core cluster in the storage area corresponding to the shared core array is to create the second shared storage area in the available storage space of the shared core array. The second shared storage area is used for storing second task data of the task corresponding to the second core cluster.

In some embodiments, after creating the second shared storage area corresponding to the second core cluster in the storage area corresponding to the shared core array, the method further includes: creating a general shared storage area in the storage area corresponding to the shared core array, The general shared storage area is used to carry the execution and operation process of the tasks corresponding to the first core cluster and the second core cluster respectively.

Step S7: Determine the idle processing cores among the processing cores that do not belong to any core cluster in the many-core system.

Step S8: Add at least one of the idle processing cores to the first core cluster, determine the at least one idle processing core as a shared core, add it to the shared core array, and jump to step S5.

In step S7, in response to determining that there is no available storage space in the storage area corresponding to the shared core array, determine an idle processing core among the processing cores that do not belong to any core cluster in the many-core system. In some embodiments, in response to determining that the memory capacity of the available storage space of the shared core array is less than the preset memory capacity, determining an idle processing core among the processing cores not belonging to any core cluster in the many-core system.

In some embodiments, in step S7, the main processing core of the first core cluster can monitor or obtain the status of each processing core in the many-core system, so as to determine all processing cores currently available in the many-core system, that is, idle Process the nucleus.

In some embodiments, in step S8, adding at least one idle processing core to the first core cluster where it is located may further include: creating an identifier of the idle processing core, and adding the identifier of the idle processing core, Add information such as address and location to the core cluster list corresponding to the first core cluster where you are located to update the core cluster list corresponding to the first core cluster where you are located, and at the same time send the core cluster list corresponding to the first core cluster where you are located to the idler processing core for storage.

In some embodiments, after receiving the core sharing request, if there is no idle second processing core in the first core cluster where the user is located, the process may further jump to step S7.

In some embodiments, the determined shared core array includes one shared core, then in step S5 and step S6, create a first shared storage area corresponding to the first core cluster from the storage area of the shared core, and create The second shared storage area corresponding to the second core cluster can also create a general shared storage area for carrying the operation process of the task.

In some embodiments, the determined shared core array includes multiple shared cores, wherein the first shared storage area corresponding to the first core cluster may correspond to one or more shared cores in the array, that is, the first core cluster corresponds to The first shared storage area includes one or more shared core storage areas; similarly, the second shared storage area corresponding to the second core cluster may also correspond to one or more shared cores in the array, that is, the second core cluster The corresponding second shared storage area includes storage areas of one or more shared cores; similarly, the general shared storage area may also correspond to one or more shared cores, that is, the general shared storage area includes storage areas of one or more shared cores. Area. Among them, the shared kernels corresponding to different regions are different.

In other words, in the case where the shared core array includes multiple shared cores, the multiple shared cores can be divided into different areas, one of which corresponds to the first core cluster and is used to store the first task data corresponding to the tasks of the first core cluster , one area corresponds to the second core cluster, and is used to store the second task data corresponding to the tasks of the second core cluster, and the other area is used as a common computing path area for the first core cluster and the second core cluster, which is used to carry the first core cluster. The execution operation process of the tasks corresponding to the cluster and the second core cluster respectively.

FIG. 5 is a schematic structural diagram of a shared core array according to an embodiment of the disclosure. Referring to FIG. 5 , in some embodiments, the shared core array includes a plurality of shared cores, and the storage area corresponding to the shared core array includes a first shared storage area M1 corresponding to the first core cluster, and a second shared storage area corresponding to the second core cluster. The area M2, the general shared storage area M0, the first shared storage area, the second shared storage area and the general shared storage area may respectively correspond to one or more shared cores. The shared core corresponding to the first shared storage area M1 may be configured to store first task data corresponding to tasks of the first core cluster, including first input data, first output data, and first task configuration information; the second shared The shared core corresponding to the storage area M2 may be configured to store the second task data corresponding to the tasks of the second core cluster, including the second input data, the second output data and the second task configuration information; the shared core corresponding to the general shared storage area M0 The core may be configured to receive and execute tasks corresponding to the first core cluster and the second core cluster respectively, that is, an operation process carrying the tasks of the first core cluster and the second core cluster.

In some embodiments, referring to FIG. 5 , the first shared storage area M1 may include a first input data area, a first output data area, and a first task configuration area, and the first input data area is configured to store the data of the first core cluster. the first input data corresponding to the task; the first output data area is configured to store the first output data corresponding to the task of the first core cluster; the first task configuration area is configured to store the first task corresponding to the task of the first core cluster configuration information.

In some embodiments, the first shared storage area M1 corresponds to multiple shared cores, the first input data area, the first output data area, and the first task configuration area may correspond to one or more shared cores, respectively, and the first input data area The corresponding shared core is configured to store the first input data, the shared core corresponding to the first output data area is configured to store the first output data, and the shared core corresponding to the first task configuration area is configured to store the first task configuration information.

In some embodiments, referring to FIG. 5 , the second shared storage area M2 may include a second input data area, a second output data area, and a second task configuration area, and the second input data area is configured to store the data of the second core cluster. the second input data corresponding to the task; the second output data area is configured to store the second output data corresponding to the task of the second core cluster; the second task configuration area is configured to store the second task corresponding to the task of the second core cluster configuration information.

In some embodiments, the second shared memory area M2 corresponds to multiple shared cores, the second input data area, the second output data area, and the second task configuration area may respectively correspond to one or more shared cores, and the second input data area The corresponding shared core is configured to store the second input data, the shared core corresponding to the second output data area is configured to store the second output data, and the shared core corresponding to the second task configuration area is configured to store the second task configuration information.

In some embodiments, the shared core array is correspondingly configured with a task list, the task list may be stored in the main processing core of the first core cluster and/or the main processing core of the second core cluster, and the shared core array may be configured according to the task list. The tasks are processed in sequence, and the task list may include, but is not limited to, computing tasks, signaling tasks, switching tasks, and the like. Among them, the computing task is, for example, a task for instructing to perform operations such as computation, calculation, etc., the signaling task is for example a task for instructing to send the required notification, request, etc. to the corresponding core cluster, and the switching task is for example, for sharing When the core switches from serving one core cluster to serving another core cluster, instructs the task of performing operations such as buffer reset and input and output address switching. After the task is processed, the corresponding task in the task list will be cleared.

FIG. 6 is a block diagram of the composition of a core sharing apparatus according to an embodiment of the present disclosure.

Referring to FIG. 6 , an embodiment of the present disclosure provides a core sharing apparatus 300 , and the core sharing apparatus 300 is applied to a many-core system, where the many-core system includes a plurality of preconfigured core clusters, and each core cluster includes at least one second core cluster. Processing cores, the core sharing apparatus 300 includes: a receiving module 301 and a sharing determination module 302 .

The receiving module 301 is configured to receive a core sharing request; the sharing determining module 302 is configured to determine at least one first core cluster from the second processing cores of the first core cluster and/or the second predetermined core cluster according to the core sharing request. Two processing cores are used as shared cores between the first core cluster and the second core cluster, and at least one shared core forms a shared core array. The shared core array is configured to receive and execute tasks corresponding to the first core cluster and the second core cluster respectively, store first task data corresponding to tasks of the first core cluster, and store tasks corresponding to the second core cluster of the second task data.

In some embodiments, the many-core system further includes a first processing core, and the receiving module 301 is configured to: receive a core sharing request sent by an external host or a first processing core, where the core sharing request is the external Generated by the host or the first processing core based on common tasks corresponding to the first core cluster and the second core cluster; or, receiving a core sharing request sent by the main processing core of the second core cluster, The main processing core of the second core cluster is a pre-designated second processing core in at least one second processing core of the second core cluster.

In some embodiments, the core sharing apparatus 300 further includes an update module (not shown in the figure), the update module is configured to update the corresponding to the shared core after the sharing determination module 302 determines at least one second processing core A core cluster list, where the updated core cluster list includes identification information corresponding to the first core cluster and the second core cluster respectively.

In some embodiments, the core sharing apparatus 300 further includes a shared storage area creation module (not shown in the figure), and the shared storage area creation module is configured to: after the sharing determination module 302 determines the at least one second processing core, A first shared storage area corresponding to the first core cluster is created in the storage area corresponding to the shared core array, and the first shared storage area is used to store the first task data; A second shared storage area corresponding to the second core cluster is created in the storage area corresponding to the array, and the second shared storage area is used to store the second task data.

In some embodiments, the core sharing apparatus 300 further includes a judging module (not shown in the figure), the judging module is used for judging the corresponding to the shared core array before the shared memory area creation module creates the first shared memory area. Whether there is available storage space in the storage area; in response to judging that there is available storage space in the storage area corresponding to the shared core array, executing the creation of the first core cluster in the storage area corresponding to the shared core array Corresponding steps of the first shared storage area.

In some embodiments, the core sharing apparatus 300 further includes a core expansion module (not shown in the figure), and the core expansion module is configured to respond to the judgment module for judging that there is no available storage space in the storage area corresponding to the shared core array, Determine an idle processing core among the processing cores that do not belong to any core cluster in the many-core system; determine at least one idle processing core as a shared core, and add it to the shared core array.

In addition, the core sharing apparatus 300 provided by the embodiments of the present disclosure is used to implement the above-mentioned core sharing method. For other descriptions of the core sharing apparatus 300, reference may be made to the description in the above-mentioned core sharing method, which will not be repeated here.

An embodiment of the present disclosure also provides a processing core, where the processing core includes the above-mentioned core sharing device.

7, an embodiment of the present disclosure provides an electronic device, the electronic device includes multiple processing cores 701 and an on-chip network 702, wherein the multiple processing cores 701 are all connected to the on-chip network 702, and the on-chip network 702 is used for interacting multiple data between cores and external data.

One or more instructions are stored in one or more processing cores 701, and one or more instructions are executed by one or more processing cores 701, so that one or more processing cores 701 can execute the above-mentioned core sharing method.

In addition, an embodiment of the present disclosure further provides a computer-readable medium on which a computer program is stored, wherein the computer program implements the above-mentioned core sharing method when executed by a processing core.

In addition, an embodiment of the present disclosure also provides a computer program product, which, when running on a computer, enables the computer to execute the above-mentioned core sharing method.

Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, functional modules/units in the systems, and devices can be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware implementation, the division between functional modules/units mentioned in the above description does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be composed of several physical components Components execute cooperatively. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor or microprocessor, or as hardware, or as an integrated circuit, such as an application specific integrated circuit . Such software may be distributed on computer-readable media, which may include computer storage media (or non-transitory media) and communication media (or transitory media). As known to those of ordinary skill in the art, the term computer storage media includes both volatile and nonvolatile implemented in any method or technology for storage of information, such as computer readable instructions, data structures, program modules or other data flexible, removable and non-removable media. Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disk (DVD) or other optical disk storage, magnetic cartridges, magnetic tape, magnetic disk storage or other magnetic storage devices, or may Any other medium used to store desired information and which can be accessed by a computer. In addition, communication media typically embodies computer readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and can include any information delivery media, as is well known to those of ordinary skill in the art .

Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should only be construed in a general descriptive sense and not for purposes of limitation. In some instances, it will be apparent to those skilled in the art that features, characteristics and/or elements described in connection with a particular embodiment may be used alone or in combination with other embodiments, unless expressly stated otherwise. Features and/or elements are used in combination. Accordingly, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the scope of the present disclosure as set forth in the appended claims.

Claims

A core sharing method based on a many-core system, the many-core system comprising a plurality of pre-configured core clusters, each of the core clusters comprising at least one second processing core, the method comprising:

receive a core sharing request;

According to the core sharing request, at least one second processing core is determined from the second processing cores of the first core cluster and/or the predetermined second core cluster, as the first core cluster and the second core cluster. Shared cores between two core clusters, at least one shared core forms a shared core array;

wherein the shared core array is configured to receive and execute tasks corresponding to the first core cluster and the second core cluster respectively, store first task data corresponding to the tasks of the first core cluster, and Second task data corresponding to the tasks of the second core cluster is stored.
The core sharing method according to claim 1, wherein the many-core system further comprises a first processing core, and the receiving a core sharing request comprises:

Receive a core sharing request sent by an external host or a first processing core, where the core sharing request is based on a common task corresponding to the first core cluster and the second core cluster by the external host or the first processing core or receive a core sharing request sent by the main processing core of the second core cluster, where the main processing core of the second core cluster is a pre-designated one of at least one second processing core of the second core cluster The second processing core.
The core sharing method according to claim 1, wherein after the at least one second processing core is determined, the method further comprises:

A core cluster list corresponding to the shared core is updated, and the updated core cluster list includes identification information corresponding to the first core cluster and the second core cluster respectively.
The core sharing method according to claim 1, wherein after the at least one second processing core is determined, the method further comprises:

creating a first shared storage area corresponding to the first core cluster in a storage area corresponding to the shared core array, where the first shared storage area is used to store the first task data;

A second shared storage area corresponding to the second core cluster is created in the storage area corresponding to the shared core array, where the second shared storage area is used for storing the second task data.
The core sharing method according to claim 4, wherein before creating a first shared storage area corresponding to the first core cluster in the storage area corresponding to the shared core array, the method further comprises:

judging whether there is available storage space in the storage area corresponding to the shared core array;

In response to determining that there is available storage space in the storage area corresponding to the shared core array, a step of creating a first shared storage area corresponding to the first core cluster in the storage area corresponding to the shared core array is performed .
The core sharing method according to claim 5, wherein after judging whether there is available storage space in the storage area corresponding to the shared core array, the method further comprises:

In response to judging that there is no available storage space in the storage area corresponding to the shared core array, determining an idle processing core in the processing cores that do not belong to any core cluster in the many-core system;

At least one of the idle processing cores is determined as the shared core, and added to the shared core array.
The core sharing method according to any one of claims 4-6, wherein the number of the shared cores is multiple, the storage area corresponding to the shared core array further includes a general shared storage area, and the first shared core The storage area, the second shared storage area and the general shared storage area respectively correspond to one or more shared cores;

Wherein, the shared core corresponding to the first shared storage area is configured to store the first task data;

The shared core corresponding to the second shared storage area is configured to store the second task data;

The shared cores corresponding to the common shared storage area are configured to receive and execute tasks corresponding to the first core cluster and the second core cluster, respectively.
The core sharing method according to claim 7, wherein the first task data includes first input data, first output data and first task configuration information, and the first shared storage area includes a first input data area, a first output data area and a first task configuration area;

Wherein, the shared core in the first input data area is configured to store the first input data; the shared core in the first output data area is configured to store the first output data; the first The shared core in the task configuration area is configured to store the first task configuration information.
A core sharing device, the core sharing device is applied to a many-core system, the many-core system includes a plurality of pre-configured core clusters, each core cluster includes at least one second processing core, and the core sharing device includes:

a receiving module for receiving a core sharing request;

A sharing determination module, configured to determine at least one second processing core from the second processing cores of the first core cluster and/or the predetermined second core cluster according to the core sharing request, as the first core cluster shared cores between the core cluster and the second core cluster, at least one shared core forms a shared core array;

wherein the shared core array is configured to receive and execute tasks corresponding to the first core cluster and the second core cluster respectively, store first task data corresponding to the tasks of the first core cluster, and Second task data corresponding to the tasks of the second core cluster is stored.
An electronic device comprising:

multiple processing cores; and

an on-chip network configured to exchange data and external data among the plurality of processing cores;

One or more of said processing cores store one or more instructions, and one or more of said instructions are executed by one or more of said processing cores to enable one or more of said processing cores to execute claim 1 The core sharing method of any one of -8.
A computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processing core, implements the core sharing method according to any one of claims 1-8.
A computer program product, when running on a computer, causes the computer to execute the core sharing method according to any one of claims 1-8.