Task branch prediction model training method, task branch prediction method

Abstract

The present disclosure provides a training method for a task branch prediction model and a task branch prediction method. The training method includes: for each historical moment in a plurality of consecutive historical moments, obtaining the selection status information of each task node before the target task node at the historical moment, as well as the selection status information of the target task node; generating a sample feature vector of the target task node based on the selection status information of each task node before the target task node; and using the selection status information of the target task node at the moment as a sample label of the target task node; and training the task branch prediction model based on all the sample feature vectors and the sample labels. The prediction model trained according to the embodiments of the present disclosure can improve task processing efficiency.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a task branch prediction model training method, a task branch prediction method, a device, a processing core, an electronic device, and a computer-readable medium. Background Art

[0002] In related technologies, the processing cores included in a many-core chip can be divided into multiple core clusters, each of which is used to process different tasks in a computation graph, thus enabling multi-task processing of the computation graph. Each task in the computation graph is assigned task configuration information. After the task to be executed is determined, the task configuration information is loaded onto the many-core chip. This method of loading task configuration information results in low task processing efficiency during multi-tasking. Summary of the Invention

[0003] The present disclosure provides a task branch prediction model training method, a task branch prediction method, an apparatus, a processing core, an electronic device, and a computer-readable medium.

[0004] In a first aspect, the present disclosure provides a method for training a task branch prediction model, the method comprising:

[0005] For each of the plurality of consecutive historical moments, obtaining selection status information of each task node preceding the target task node at the historical moment, as well as selection status information of the target task node;

[0006] generating a sample feature vector of the target task node according to the selection state information of each task node before the target task node; and using the selection state information of the target task node at that moment as a sample label of the target task node;

[0007] The task branch prediction model is obtained by training according to all the sample feature vectors and the sample labels.

[0008] In a second aspect, the present disclosure provides a task branch prediction method, comprising:

[0009] Get the selection status information of each task node before the current task node;

[0010] Generate a feature vector corresponding to the current task node according to the selection status information of each task node;

[0011] Input the feature vector into a preset task branch prediction model to predict the next task node of the current task node;

[0012] The task branch prediction model is trained using the task branch prediction model training method provided by the embodiment of the present disclosure.

[0013] In a third aspect, the present disclosure provides a training device for a task branch prediction model, the training device for the task branch prediction model comprising:

[0014] A data acquisition module is used to obtain, for each of a plurality of consecutive historical moments, selection status information of each task node preceding a target task node at the historical moment, as well as selection status information of the target task node;

[0015] a sample data generation module, configured to generate a sample feature vector of the target task node based on the selection status information of each task before the target task node; and use the selection status information of the target task node at that moment as a sample label of the target task node;

[0016] A training module is used to train the task branch prediction model based on all the sample feature vectors and the sample labels.

[0017] In a fourth aspect, the present disclosure provides a task branch prediction device, comprising:

[0018] An acquisition module is used to obtain the selection status information of each task node before the current task node;

[0019] A feature vector generating module, configured to generate a feature vector corresponding to the current task node according to the selection status information of each task node;

[0020] The prediction module is used to input the feature vector into a preset task branch prediction model to predict the next task node of the current task node.

[0021] In a fifth aspect, the present disclosure provides a processing core, which includes the training device provided by an embodiment of the present disclosure and / or the task branch prediction device provided by an embodiment of the present disclosure.

[0022] In a sixth aspect, the present disclosure provides an electronic device comprising: a plurality of processing cores; and an on-chip network configured to exchange data between the plurality of processing cores and external data; wherein one or more instructions are stored in one or more of the processing cores, and one or more of the instructions are executed by one or more of the processing cores, so that one or more of the processing cores can execute the training method of the task branch prediction model provided by an embodiment of the present disclosure, and / or enable one or more of the processing cores to execute the task branch prediction method provided by an embodiment of the present disclosure.

[0023] In the seventh aspect, the present disclosure provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processing core, implements the training method of the task branch prediction model provided by an embodiment of the present disclosure, and / or the computer program, when executed by a processing core, implements the task branch prediction method provided by an embodiment of the present disclosure.

[0024] The present disclosure provides a training method, a task branch prediction method, an apparatus, a processing core, an electronic device, and a computer-readable medium for a task branch prediction model. The training model is obtained by training based on the historical selection paths of each task in a computational graph. The training model is used to perform task branch prediction, thereby pre-loading the task configuration information of the next task onto the chip during the execution of the current task, thereby improving task processing efficiency.

[0025] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of 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 the accompanying drawings:

[0027] Figure 1 A schematic diagram of a task calculation graph involved in an embodiment of the present disclosure;

[0028] Figure 2 A flowchart of a method for training a task branch prediction model provided in an embodiment of the present disclosure;

[0029] Figure 3 A flowchart of a task branch prediction method provided by an embodiment of the present disclosure;

[0030] Figure 4 A flowchart of a task branch prediction method provided by an embodiment of the present disclosure;

[0031] Figure 5 A flowchart of some steps of a task branch prediction method provided by an embodiment of the present disclosure;

[0032] Figure 6 A block diagram of a training device for a task branch prediction model provided by an embodiment of the present disclosure;

[0033] Figure 7 A block diagram of a task branch prediction device provided by an embodiment of the present disclosure;

[0034] Figure 8 A block diagram of the composition of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0035] In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the following exemplary embodiments of the present disclosure are described in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely

[0036] The embodiments described herein are exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0037] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0038] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0039] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded. Similar words such as "connected" or "connected" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0040] 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 commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless expressly defined as such herein.

[0041] The task branch prediction model and task branch prediction method provided by the embodiments of the present disclosure are used to predict task branch nodes in tasks with branches in a computation graph.

[0042] Figure 1 This is a schematic diagram of the task calculation graph involved in the embodiment of the present disclosure, refer to Figure 1, where P, A1, A2, A3, B1…G2 represent task nodes. Tasks with branches include: A, B, C, E, F, and G. Task A’s task branch nodes include A1, A2, and A3; task B’s task branch nodes include B1 and B2; task C’s task branch nodes include C1 and C2; task E’s task branch nodes include E1 and E2; task F’s task branch nodes include F1 and F2; and task G’s task branch nodes include G1 and G2.

[0043] Figure 2 A flowchart of a training method for a task branch prediction model provided in an embodiment of the present disclosure. The training method can be executed by an electronic device such as a terminal device or a server. The terminal device can be an in-vehicle device, a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, an in-vehicle device, a wearable device, etc. The training method can be implemented by a processor calling computer-readable program instructions stored in a memory. Alternatively, the training method can be executed by a server.

[0044] Reference Figure 1 and Figure 2 The training method of the task branch prediction model provided by the embodiment of the present disclosure includes:

[0045] Step S101: For each historical moment in a plurality of consecutive historical moments, obtain the selection status information of each task node before the target task node at the historical moment, as well as the selection status information of the target task node.

[0046] In the embodiments of the present disclosure, a target task node refers to a task branch node in a task with branches. For example, the target task node may be task branch node B1 of task B, or task branch node E2 of task E, etc. The task nodes preceding the target task node refer to the task nodes traversed from the start task node to the target task node.

[0047] The selection state information of the task node represents the branch selection result of the task. For example, when performing task A, if task branch A1 is selected, the selection state information of task node A1 represents the selection of task branch A1 among the three task branches A1, A2, and A3. In some embodiments, the selection state information of the task node can be represented by one-hot encoding. One-hot encoding uses an N-bit state register to encode N states. Each state has its own independent register bit, and at any time, only one of them is valid. For example, task A has 3 task branches. A 3-bit state register is used to identify the task branch selection result of task A. When task node A1 is selected, the one-hot encoding corresponding to task node A1 is 100. When task node A2 is selected, the one-hot encoding corresponding to task node A2 is 010. When task node A3 is selected, the one-hot encoding corresponding to task node A3 is 001. It should be noted that if a task has no branches, its corresponding one-hot encoding is 1. If some tasks are not traversed, their corresponding one-hot encoding is all zero.

[0048] In some embodiments, the selection status information of the target task node at multiple consecutive historical moments and the corresponding selection status information of each task node before the target task node can be obtained by querying the task execution log of the computation graph on the many-core chip.

[0049] Step S102 : generating a sample feature vector of the target task node based on the selection status information of each task node before the target task node; and using the selection status information of the target task node at that moment as a sample label of the target task node.

[0050] The selection status information corresponding to the target task node has a corresponding relationship with the selection status information of each task node traversed from the starting task node to the target task node. Therefore, a sample can be constructed based on the selection status information corresponding to the target task node and the selection status information of each task node traversed from the starting task node to the target task node at each historical moment in multiple consecutive historical moments.

[0051] In the disclosed embodiment, the sample includes a sample feature vector and a sample label. The sample feature vector is generated based on the selection status information of each task node before the target task node. The sample feature vector can represent the execution path information of each task node before the target task node.

[0052] In some embodiments, when the selection status information of a task node is represented using a one-hot encoding, the process of generating a sample feature vector includes: combining the one-hot encodings corresponding to the selection status information of each task node preceding the target task node into a one-hot encoding sequence according to the order in which each task node was executed, and using the one-hot encoding sequence as the sample feature vector. Correspondingly, using the one-hot encoding corresponding to the selection status information of the target task node at that historical moment as a sample label for the target task node.

[0053] Step S103: Train a task branch prediction model based on all sample feature vectors and sample labels.

[0054] After the processing of step S102, each moment in the continuous multiple historical moments corresponds to a sample feature vector and a sample label. All the sample feature vectors and corresponding sample labels are trained to obtain a task branch prediction model, which can predict the task branch to be executed after the current task node. Figure 1 For example, if the current task node is D, the task branch prediction model can predict whether task E1 or task E2 will be executed after task D is completed.

[0055] In some embodiments, the task branch prediction model is trained based on a neural network model. Further, in some embodiments, the neural network model is a spiking neural network model. It should be understood that those skilled in the art can set the type of neural network model according to actual circumstances, such as a spiking neural network model (SNN), an artificial neural network model (ANN), or a hybrid neural network model of SNN and ANN, etc., and this disclosure does not limit this.

[0056] In some embodiments, the neural network model structure includes: a neuron layer, a temporal integration layer, at least one fully connected layer, and at least one selection layer. Furthermore, in some embodiments, the selection layer is based on a winner-takes-all criterion. The neuron layer acts as a predictor, predicting which branch to select based on input data. Specifically, network weights can be obtained through gradient descent training.

[0057] In some embodiments, a task branch prediction model is obtained by training a network based on real-time, correct branch selection results as labels. The trained network can be used for branch prediction. It should be understood that those skilled in the art can set the labels used to train the task branch prediction model based on actual circumstances, and this disclosure does not limit this.

[0058] It should be noted that the task branch prediction model provided by the embodiment of the present disclosure can also predict the task branches of multiple tasks at the same time. Figure 1, the branch nodes of task F and task G can be predicted at the same time. At this time, the corresponding neural network model includes at least one fully connected layer and at least one selection layer.

[0059] The training method of the task branch prediction model provided in the embodiment of the present disclosure is based on the historical selection path of each task in the computation graph to obtain a training model. The training model can be used for prediction, so that during the execution of the current task, the task configuration information of the next task can be loaded onto the chip in advance, thereby improving the task processing efficiency.

[0060] Figure 3 A flowchart of a task branch prediction method provided by an embodiment of the present disclosure.

[0061] Reference Figure 3 The task branch prediction method provided by the embodiment of the present disclosure includes:

[0062] Step S201: Acquire selection status information of each task node before the current task node.

[0063] The current task node refers to the computation graph (such as Figure 1 The node corresponding to the task currently being executed in the calculation graph shown in the figure is the node that corresponds to the task currently being executed. The task nodes before the current task node refer to the nodes that have been traversed from the starting node to the current task node. Figure 1 For example, if the current task node is task node D, then the task nodes before the current task node may include P, A1, B1, or P, A1, B2, or P, A2, C1, or P, A2, C2, or P, A3. Correspondingly, the selection status information of the task nodes before the current task node may include: 1(P), 100(A1), 10(B1), 00(C), or 1(P), 100(A1), 01(B2), 00(C), or 1(P), 010(A2), 00(B), 10(C1), or 1(P), 010(A2), 00(B), 01(C2), or 1(P), 001(A3), 00(B), 00(C).

[0064] Step S202: Generate a feature vector corresponding to the current task node according to the selection status information of each task node.

[0065] In some embodiments, when the selection state information of the task node is represented by one-hot encoding, the selection state information of the task node includes: the one-hot encoding corresponding to the selection state of the task node.

[0066] Figure 5 This is a flowchart of some steps of a task branch prediction method provided by an embodiment of the present disclosure. Figure 5 In some embodiments, step S202 includes:

[0067] Step S202a: The one-hot encoding corresponding to the selection status information of each task node before the current task node is combined into a one-hot encoding sequence according to the execution order of each task node, and the one-hot encoding sequence is used as the feature vector of the current task node.

[0068] For example, if the selection status information of each task node before the current task node includes: 1(P), 100(A1), 10(B1), 00(C), then the generated feature vector of the current task node is 11001000.

[0069] Step S203: Input the feature vector into a preset task branch prediction model to predict the next task node of the current task node.

[0070] Combine Figure 1 For example, if the current task node is task node D, then the feature vector generated by the selection status information of each task node before task node D is input into the task branch prediction model to predict whether task E1 or task E2 will be executed next.

[0071] The task branch prediction method provided by the embodiments of the present disclosure performs prediction based on a pre-trained task branch prediction model, thereby making the predicted task branches more accurate.

[0072] Figure 4 A flowchart of a task branch prediction method provided by an embodiment of the present disclosure.

[0073] Reference Figure 4 In some embodiments, after predicting the next task node of the current task node, the method further includes:

[0074] Step S204: Load the task configuration information of the next task node onto the many-core chip.

[0075] After predicting the next task node to be executed after the current processing task node, the task configuration information of the next task node is loaded onto the many-core chip, so that the many-core chip can determine the task node to be executed in advance, thereby improving task processing efficiency.

[0076] In some embodiments, the task configuration information may be loaded into the core cluster corresponding to the next task node on the many-core chip.

[0077] Figure 6 A block diagram of a training device for a task branch prediction model provided in an embodiment of the present disclosure.

[0078] Reference Figure 6The training device of the task branch prediction model includes: a data acquisition module 11, a sample data generation module 12 and a training module 13.

[0079] Among them, the data acquisition module 11 is used to obtain the selection status information of each task node before the target task node at each historical moment in multiple consecutive historical moments, as well as the selection status information of the target task node; the sample data generation module 12 is used to generate a sample feature vector of the current task node based on the selection status information of each task before the current task node; and, use the selection status information of the current task node at this moment as a sample label of the current task node; the training module 13 is used to train a task branch prediction model based on all sample feature vectors and sample labels.

[0080] In some embodiments, the task branch prediction model performs prediction based on a neural network model. Further, in some embodiments, the neural network model is a spiking neural network model.

[0081] In some embodiments, the structure of the neural network model includes: a neuron layer, a temporal integration layer, at least one fully connected layer, and at least one selection layer. Further, in some embodiments, the selection criteria of the selection layer include: a winner-takes-all criterion.

[0082] The training device for the task branch prediction model provided in the embodiment of the present disclosure is used to implement the training method for the task branch prediction model. For the specific functional modules included in the training device for the task branch prediction model and the interactions between the functional modules, please refer to the description of the corresponding method steps in the above embodiment, which will not be repeated here.

[0083] Figure 7 A block diagram of a task branch prediction device provided in an embodiment of the present disclosure.

[0084] Reference Figure 7 The task branch prediction device provided by the embodiment of the present disclosure includes: an acquisition module 21, a feature vector generation module 22 and a prediction module 23.

[0085] Among them, the acquisition module 21 is used to obtain the selection status information of each task node before the current task node; the feature vector generation module 22 is used to generate the feature vector corresponding to the current task node based on the selection status information of each task node; the prediction module 23 is used to input the feature vector into the preset task branch prediction model to predict the next task node of the current task node.

[0086] The task branch prediction device provided in the embodiment of the present disclosure is used to implement the task branch prediction method provided in the embodiment of the present disclosure. For the specific functional modules included in the task branch prediction device and the interactions between the functional modules, please refer to the description of the corresponding method steps in the above embodiment, which will not be repeated here.

[0087] The embodiments of the present disclosure also provide a processing core, including the training device provided by the embodiments of the present disclosure, and / or including the task branch prediction device provided by the embodiments of the present disclosure.

[0088] Figure 8 A block diagram of the composition of an electronic device provided in an embodiment of the present disclosure.

[0089] Reference Figure 8 An embodiment of the present disclosure provides an electronic device, which includes multiple processing cores 801 and an on-chip network 802, wherein the multiple processing cores 801 are connected to the on-chip network 802, and the on-chip network 802 is used to exchange data between the multiple processing cores and external data.

[0090] Among them, one or more instructions are stored in one or more processing cores 801, and one or more instructions are executed by one or more processing cores 801, so that one or more processing cores 801 can execute the above-mentioned task branch prediction model training method, or enable one or more of the processing cores to execute the above-mentioned task branch prediction method.

[0091] In addition, an embodiment of the present disclosure also provides a computer-readable medium on which a computer program is stored, wherein the computer program implements the above-mentioned task branch prediction model training method when executed by the processing core, and / or the computer program implements the above-mentioned task branch prediction method when executed by the processing core.

[0092] It will be appreciated by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In hardware implementations, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium). As is well known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable, and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media include, but are not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, it is well known to those skilled in the art that communication media generally 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 may include any information delivery media.

[0093] Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should be interpreted only in a general illustrative sense and not for purposes of limitation. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly indicated, features, characteristics, and / or elements described in conjunction with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in conjunction with other embodiments. Therefore, it will be understood by those skilled in the art that various changes in form and detail may be made without departing from the scope of the present disclosure as set forth in the appended claims.

Claims

1. A method for training a task branch prediction model, comprising: For each of the plurality of consecutive historical moments, obtaining selection status information of each task node preceding the target task node at the historical moment, as well as selection status information of the target task node; generating a sample feature vector of the target task node according to the selection state information of each task node before the target task node; and using the selection state information of the target task node at that moment as a sample label of the target task node; The task branch prediction model is obtained by training according to all the sample feature vectors and the sample labels; The task branch prediction model is used to predict the next task node of the current task node when the multi-core chip executes the current task node, so that the task configuration information of the next task node is loaded into the core cluster corresponding to the next task node in the multi-core chip.

2. The training method of the task branch prediction model according to claim 1, wherein: The task branch prediction model is obtained by training based on a neural network model.

3. The training method of the task branch prediction model according to claim 1 or 2, wherein: The task branch prediction model is obtained by network training based on real-time correct branch selection results as labels.

4. The training method of the task branch prediction model according to claim 2, wherein: The neural network model is a pulse neural network model.

5. The training method of the task branch prediction model according to claim 4, wherein: The structure of the neural network model includes: a neuron layer, a time integration layer, at least one fully connected layer and at least one selection layer.

6. The training method of the task branch prediction model according to claim 5, wherein: The selection criteria of the selection layer include: winner-takes-all criteria.

7. A task branch prediction method, comprising: Get the selection status information of each task node before the current task node; Generate a feature vector corresponding to the current task node according to the selection status information of each task node; Input the feature vector into a preset task branch prediction model to predict the next task node of the current task node; The task branch prediction model is trained using the task branch prediction model training method according to any one of claims 1 to 5; After predicting the next task node of the current task node, the method further includes: The task configuration information of the next task node is loaded into the core cluster corresponding to the next task node in the many-core chip.

8. The task branch prediction method according to claim 7, wherein: The selection status information of the task node includes: the one-hot encoding corresponding to the selection status of the task node; Generating the feature vector corresponding to the current task node according to the selection status information of each task node includes: The one-hot encoding corresponding to the selection status information of each task node before the current task node is combined into a one-hot encoding sequence according to the execution order of each task node, and the one-hot encoding sequence is used as the feature vector of the current task node.

9. A training device for a task branch prediction model, comprising: A data acquisition module is used to obtain, for each of a plurality of consecutive historical moments, selection status information of each task node preceding a target task node at the historical moment, as well as selection status information of the target task node; a sample data generation module, configured to generate a sample feature vector of the target task node based on the selection status information of each task before the target task node; and use the selection status information of the target task node at that moment as a sample label of the target task node; A training module, configured to obtain the task branch prediction model through training based on all the sample feature vectors and the sample labels; The task branch prediction model is used to predict the next task node of the current task node when the multi-core chip executes the current task node, and load the task configuration information of the next task node into the core cluster corresponding to the next task node in the multi-core chip.

10. A task branch prediction device, comprising: An acquisition module is used to obtain the selection status information of each task node before the current task node; A feature vector generating module, configured to generate a feature vector corresponding to the current task node according to the selection status information of each task node; A prediction module is used to input the feature vector into a preset task branch prediction model to predict the next task node of the current task node; After the prediction module predicts the next task node of the current task node, the device is further configured to: The task configuration information of the next task node is loaded into the core cluster corresponding to the next task node in the many-core chip.

11. A processing core comprising the training device according to claim 9 and / or the task branch prediction device according to claim 10.

12. An electronic device comprising: Multiple processing cores; as well as an on-chip network configured to exchange data between the plurality of processing cores and external data; One or more instructions are stored in one or more of the processing cores, and one or more of the instructions are executed by one or more of the processing cores, so that one or more of the processing cores can execute the training method of the task branch prediction model described in any one of claims 1-6, and / or enable one or more of the processing cores to execute the task branch prediction method described in any one of claims 7-8.

13. A computer-readable medium having a computer program stored thereon, wherein: When the computer program is executed by the processing core, the computer program implements the training method of the task branch prediction model according to any one of claims 1 to 6, and / or when the computer program is executed by the processing core, the computer program implements the task branch prediction method according to any one of claims 7 to 8.

Images