WO2025139217A1 - Computing resource processing method, device, and storage medium - Google Patents

Abstract

Provided in the embodiments of the present disclosure are a computing resource processing method, a device, and a storage medium. In the embodiments of the present disclosure, resource scheduling for physical computing resource objects are combined with dormancy control over the physical computing resource objects, sleep state switching prediction information respectively corresponding to a plurality of physical computing resource objects are maintained, and the physical computing resource objects are scheduled on the basis of the maintained sleep state switching prediction information, such that a task can be scheduled to a physical computing resource object in a relatively light sleep state or a relatively busy physical computing resource object to the greatest possible extent, which is conducive to making some physical computing resource objects be in a deep sleep state for a longer time to the greatest possible extent and making some physical computing resource objects be in a busy state to the greatest possible extent, thereby reducing the frequency of the physical computing resource objects entering and exiting the sleep state.

Description

Computing resource processing method, device and storage medium

This disclosure claims priority to the Chinese patent application filed with the China Patent Office on December 26, 2023, with application number 202311817214.3 and application name “Computing Resource Processing Method, Device and Storage Medium”, all contents of which are incorporated by reference in this disclosure.

Technical Field

The present disclosure relates to the field of cloud computing technology, and in particular to a computing resource processing method, device and storage medium.

Background Art

With the development of cloud computing technology, more and more users choose to use cloud servers as their infrastructure. Cloud servers usually use multi-core processors (Central Processing Unit, CPU), such as 4-core processors or 8-core processors, to provide users with higher computing performance.

In order to achieve higher computing performance while reducing device power consumption, the Advanced Configuration and Power Management Interface (ACPI) specification defines CPU sleep states. When the CPU is idle, entering sleep state can save power consumption and help other CPUs work at a higher frequency.

However, the prior art faces the problem that the CPU frequently enters and exits the sleep state. Since entering and exiting the sleep state itself consumes additional power consumption, frequent entry and exit of the sleep state not only fails to save power consumption, but may even cause other CPUs to be unable to obtain a higher frequency.

Summary of the invention

Various aspects of the present disclosure provide a computing resource processing method, device, and storage medium to reduce the frequency of physical computing resource objects entering and exiting a sleep state.

The embodiment of the present disclosure provides a physical machine, wherein an operating system is running on hardware resources of the physical machine, wherein the hardware resources include multiple physical computing resource objects, and the multiple physical computing resource objects support multiple sleep states with different sleep depths, and the operating system includes: a target resource scheduler and a target sleep controller; the target sleep controller is used to maintain sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, and the sleep state transfer prediction information includes the sleep state transfer prediction information of the corresponding physical computing resource object when it is awakened from any sleep state in the next sleep state. The target resource scheduler is used to select a target physical computing resource object according to the sleep state transition prediction information corresponding to each of the multiple physical computing resource objects; and schedule the target task to be scheduled to the target physical computing resource object.

The disclosed embodiment also provides a computing resource processing method, including: maintaining sleep state transfer prediction information corresponding to each of a plurality of physical computing resource objects, the sleep state transfer prediction information including probability information of the corresponding physical computing resource object entering each of the plurality of sleep states at the next sleep when awakened from any of the plurality of sleep states; selecting a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the plurality of physical computing resource objects; and scheduling a target task to be scheduled to the target physical computing resource object.

The embodiment of the present disclosure also provides a multi-core processor system, including multiple processor cores and a memory, wherein the memory is used to store program code of an operating system, and the multiple processor cores are used to run the program code of the operating system to implement the steps in the computing resource processing method.

The embodiment of the present disclosure also provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor is enabled to implement the steps in the computing resource processing method.

The embodiment of the present disclosure further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in the computing resource processing method are implemented.

In this embodiment, resource scheduling of physical computing resource objects is combined with sleep control of physical computing resource objects. By maintaining sleep state transfer prediction information corresponding to multiple physical computing resource objects, the physical computing resource objects are scheduled according to the maintained sleep state transfer prediction information. This can schedule tasks to physical computing resource objects with relatively shallow sleep states or relatively busy states as much as possible, which is beneficial for keeping some physical computing resource objects in deep sleep states for as long as possible, keeping some physical computing resource objects in busy states as much as possible, reducing the frequency of physical computing resource objects entering and exiting sleep states, and thereby reducing the frequency of predicting sleep states, thereby reducing problems caused by prediction errors.

BRIEF DESCRIPTION OF THE DRAWINGS

The drawings described herein are used to provide a further understanding of the present disclosure and constitute a part of the present disclosure. The illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation on the present disclosure. In the drawings:

FIG. 1a is a schematic diagram of a sleep state transition provided by an exemplary embodiment of the present disclosure;

FIG1b is a schematic diagram of sleep power consumption provided by an exemplary embodiment of the present disclosure;

FIG2 is a software-hardware architecture diagram of a physical machine provided by another exemplary embodiment of the present disclosure;

FIG3 is a schematic diagram of a flow chart of controlling a physical computing resource object to enter a third dormant state provided by another exemplary embodiment of the present disclosure;

FIG4 is a schematic diagram of the control logic of a menu sleep controller provided by yet another exemplary embodiment of the present disclosure;

FIG5 is a schematic diagram of a flow chart of a computing resource processing method provided by another exemplary embodiment of the present disclosure;

FIG6a is a schematic diagram of the deployment and implementation of a computing resource processing method provided by another exemplary embodiment of the present disclosure in an actual application scenario;

FIG6b is a schematic diagram of a process of updating a data structure provided by another exemplary embodiment of the present disclosure;

FIG6c is a schematic diagram of a process of selecting a sleep state provided by another exemplary embodiment of the present disclosure;

FIG6d is a schematic diagram of a process of scheduling by a target resource scheduler provided by another exemplary embodiment of the present disclosure;

FIG7 is a schematic diagram of a computing resource processing device provided by another exemplary embodiment of the present disclosure;

FIG. 8 is a schematic diagram of an electronic device provided by yet another exemplary embodiment of the present disclosure.

DETAILED DESCRIPTION

In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the technical solutions of the present disclosure will be clearly and completely described below in combination with the specific embodiments of the present disclosure and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present disclosure.

It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation portals for users to choose to authorize or refuse. In addition, the various models involved in this disclosure (including but not limited to language models or large models) are in compliance with relevant laws and standards.

Taking a multi-core processor scenario as an example, the ACPI specification can define the power states that each CPU core in the multi-core processor can be in, such as power state C0, power state C1, power state C2...power state Cn. Among them, power state C0 is the effective power state of the CPU core executing instructions, power state C1-power state Cn are the sleep states of the CPU core, and the sleep state is a low-load power state. In a multi-core processor scenario, when the CPU core is not needed to perform a task, the API (Application Programming Interface) provided by the CPU firmware can be used to control the CPU core to any sleep state. Compared with the CPU core in the C0 state, this can reduce the overall power consumption of the multi-core processor, that is, consume less power and emit less heat. Among them, from the sleep state C1-sleep state Cn, under the condition of maintaining the same sleep time, the power consumption saved in each sleep state increases successively.

As shown in the sleep state transition diagram of FIG1a, when the CPU core is in power state C0, ACPI can change the overall performance of the multi-core processor through the defined throttling logic and the sleep state transition logic defined by ACPI. For example, the multi-core processor can start the throttling logic, thereby using the sleep state transition logic to issue an entry C1 command, so that the CPU core enters the sleep state C1 from the power state C0, and when the CPU core receives the sleep interrupt instruction, it can exit the sleep state C1 and return to the power state C0. In addition, the method of entering/exiting the sleep state C2 and C3 is the same as the method of entering/exiting the sleep state C1 described above, and will not be repeated here.

It should be noted that, from sleep state C1 to sleep state Cn, the power consumption saved in each sleep state increases in sequence, which can also be understood as the process from sleep state C1 to sleep state Cn is the process of sleep state from "shallow" to "deep". When some CPU cores enter a deeper sleep state, other CPU cores can obtain a higher operating frequency. As shown in Table 1, here a multi-core processor is taken as an example. The processor has a total of 48 CPU cores. Assuming that sleep state C6 is the deepest sleep state supported by the multi-core processor, the following Table 1 shows the corresponding relationship between the number of CPU cores that are not in sleep state C6 and the operating frequency that other CPU cores can obtain. It can be seen from Table 1 that the fewer the number of CPU cores that are not in sleep state C6, the more the number of CPU cores in sleep state C6, and the higher the operating frequency that other cores can obtain.

Table 1

From the above, it can be seen that the CPU core can save power by entering the sleep state when it is idle. When it is scheduled again, it can be awakened and exit the sleep state. As shown in Figure 1b, when the CPU core enters or exits a sleep state, it will consume power. The power consumption and time required. Among them, the time required to exit the sleep state can be called the exit delay time. In addition, as the depth of the sleep state increases, the power consumption and time consumed to enter or exit the sleep state will gradually increase. It can be seen that if the appropriate sleep state cannot be selected for the CPU core, the CPU core will frequently enter and exit the sleep state. Entering and exiting the sleep state itself will consume additional power consumption, resulting in other CPU cores being unable to obtain a higher frequency. Not only will it not bring power consumption savings, but it will also cause a waste of power consumption resources and response delays. For example, in some cases, if the actual sleep time of the CPU core is shorter than the expected sleep time, a deeper sleep state is mistakenly selected for the CPU core based on the expected sleep time. Since the actual sleep time of the CPU core in this sleep state is shorter, the power consumption saved by sleep is less than the power consumption consumed by entering and exiting the sleep state, which not only wastes CPU power consumption but also causes a longer response delay. For example, in other cases, the actual sleep time of the CPU core is longer than the expected sleep time, and a shallower sleep state is incorrectly selected for the CPU core based on the expected sleep time, which will cause the CPU core to wake up from the sleep state prematurely, wasting CPU power consumption. In addition, since the CPU core fails to enter a deeper sleep state, other cores cannot obtain a higher frequency.

In the disclosed embodiment, based on the relevant contents of the above-mentioned multi-core processor, the concept of sleep state is introduced into various physical computing resource objects. Physical computing resource objects may include CPU, DPU (Data Processing Unit) or TPU (Tensor Processing Unit) and other physical computing resource objects similar to CPU. In the case of introducing sleep state into various physical computing resource objects, in the scenario where multiple physical computing resource objects work together, technical problems such as frequent entry and exit of sleep state leading to power consumption waste, response delay and other CPUs being unable to obtain higher operating frequencies will also be faced.

In view of the above problems, in the embodiments of the present disclosure, it is proposed to combine resource scheduling of physical computing resource objects with sleep control of physical computing resource objects. By maintaining sleep state transfer prediction information corresponding to each of a plurality of physical computing resource objects, the physical computing resource objects are scheduled according to the maintained sleep state transfer prediction information. This allows tasks to be scheduled to physical computing resource objects with relatively shallow sleep states or relatively busy states as much as possible, which is beneficial for keeping some physical computing resource objects in deep sleep states for as long as possible, keeping some physical computing resource objects in busy states as much as possible, reducing the frequency of physical computing resource objects entering and exiting sleep states, giving full play to the advantages of sleep states, saving power consumption, and allowing other physical computing resource objects that are not in deep sleep states to obtain higher operating frequencies as much as possible. In addition, since the frequency of physical computing resource objects entering and exiting sleep states is reduced, it means that the frequency of predicting sleep states is also reduced, which is beneficial for reducing problems caused by prediction errors and solving the above technical problems.

The technical solutions provided by various embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

FIG2 is a software-hardware architecture diagram of a physical machine provided by an exemplary embodiment of the present disclosure. As shown in FIG2 , the physical machine includes an application layer 21 , a kernel layer (ie, an operating system) 22 , and a hardware resource layer 23 .

The hardware resource layer 23 includes various hardware resources of the physical machine. The hardware resources of the physical machine include multiple physical computers. A physical computing resource object, multiple physical computing resource objects work together, and multiple physical computing resource objects can be integrated on one chip but are not limited to this. Taking the physical computing resource object as a CPU as an example, multiple physical computing resource objects can be implemented as a multi-core CPU. In addition, the hardware resources of the physical machine also include memory, communication components, display, power components, audio components, and various external devices, which are not shown in Figure 2. In addition, Figure 2 only illustrates the physical computing resource objects, and does not limit the number of physical computing resource objects.

In this embodiment, for multiple physical computing resource objects, multiple sleep states with different sleep depths are supported. Depending on the type of physical computing resource object and the manufacturer, the number of sleep states supported by the physical computing resource object will be different. In addition, the power consumption and time consumed in entering and exiting each sleep state will also be different. For example, the physical computing resource objects provided by some manufacturers can support sleep state C1, sleep state C2 and sleep state C3, and the physical computing resource objects provided by other manufacturers can support sleep state C1, sleep state C2, sleep state C3, sleep state C4, sleep state C5 and sleep state C6, which are not limited to this. Regardless of the type of physical computing resource object, regardless of which manufacturer provides the physical computing resource object, for the multiple sleep states supported by the physical computing resource object, the sleep depth of different sleep states will be different. Optionally, for the convenience of description, the number after C can be used to represent the sleep depth of the sleep state. As the number after C increases, the sleep depth of the sleep state increases successively, and accordingly, the power consumption and time consumed in entering and exiting these sleep states will also increase successively.

In this embodiment, an operating system, i.e., a kernel layer 22, runs on the hardware resources of the physical machine, and above the operating system is an application layer 21, which includes various application programs running on the physical machine. Depending on the application scenario and implementation form of the physical machine, the application program may also be slightly different in implementation. Taking the physical machine as a mobile phone, computer and other terminal devices as an example, the application program running on the physical machine may be an e-commerce shopping application, various video applications, mailbox applications, instant messaging applications, various office applications, etc. Taking the physical machine as a conventional server, cloud server, server cluster and other server devices as an example, the application program running on the physical machine may be various databases or data warehouses, streaming computing applications, log processing applications, distributed computing, various virtual machines or containers, etc.

Relative to the application program, the operating system is a group of interrelated system software programs that are responsible for and control various operations of the physical machine, use and run hardware and software resources, and provide public services to organize user interactions. In the embodiment of the present disclosure, the operating system acts as a bridge between the application program and the hardware resources. The operating system includes various hardware resource drivers for driving various hardware resources. In addition, the operating system of the present embodiment also includes: a sleep controller for controlling the sleep state of physical computing resource objects, and a resource scheduler for scheduling physical computing resource objects. In the present embodiment, in order to allow the physical computing resource objects to reasonably enter and exit the sleep state, a new sleep controller and a new resource scheduler are provided. For the sake of ease of description and distinction, the new sleep controller and the new resource scheduler provided in the embodiment of the present disclosure are respectively referred to as the target sleep controller 11 and the target resource scheduler 12. It is explained here that the operating system may only include the target sleep controller 11 and the target resource scheduler 12, or may include other sleep controllers and their respective resource schedulers. The specific resource scheduler may be determined according to application requirements and application scenarios, and is not limited thereto. Next, the process of the target sleep controller 11 and the target resource scheduler 12 cooperating with each other to perform sleep control and resource scheduling on the physical computing resource objects will be described in detail.

In this embodiment, the target sleep controller 11 can maintain sleep state transition prediction information corresponding to each of the multiple physical computing resource objects, wherein the sleep state transition prediction information corresponding to any physical computing resource object is used to predict the state transition of the physical computing resource object in different sleep states. The sleep state transition prediction information may include probability information of the corresponding physical computing resource object entering each sleep state when it is awakened from any sleep state at the next sleep. The probability information can be implemented as a probability value or as a weight, which is not limited. For example, assuming that a physical computing resource object can support sleep state C0, sleep state C1 and sleep state C2, the sleep state transfer prediction information corresponding to the physical computing resource object may include: when the physical computing resource object is awakened from sleep state C0, the probability of entering sleep state C0 at the next sleep is 70%, the probability of entering sleep state C1 is 15%, and the probability of entering sleep state C2 is 15%; when the physical computing resource object is awakened from sleep state C1, the probability of entering sleep state C0 at the next sleep is 30%, the probability of entering sleep state C1 is 40%, and the probability of entering sleep state C2 is 20%; when the physical computing resource object is awakened from sleep state C2, the probability of entering sleep state C0 at the next sleep is 50%, the probability of entering sleep state C1 is 15%, and the probability of entering sleep state C2 is 30%.

In this embodiment, the target sleep controller 11 can disclose the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects it maintains to the target resource scheduler 12. Optionally, the target sleep controller 11 can actively provide the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects it maintains to the target resource scheduler 12, or the target resource scheduler 12 can also actively request the target sleep controller 11, and the target sleep controller 11 provides the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects it maintains to the target resource scheduler 12 according to the request. For the target resource scheduler 12, when performing task scheduling, it can select the target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, and schedule the target task to be scheduled to the target physical computing resource object. For example, the target resource scheduler 12 may select physical computing resource object D1 as the target physical computing resource object from physical computing resource object D1, physical computing resource object D2, physical computing resource object D3 and physical computing resource object D4 according to the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, and schedule the target task to the target physical computing resource object.

In the embodiment of the present disclosure, the target resource scheduler 12 performs resource scheduling according to the sleep state transition prediction information corresponding to each of the multiple physical computing resource objects, with the scheduling principle of giving priority to physical computing resource objects that are not in a sleep state, in a shallow sleep state, or have a high probability of entering a shallow sleep state. In this way, the resource scheduling of the physical computing resource objects can be combined with the sleep control of the physical computing resource objects. The sleep state transition prediction information is used to schedule physical computing resource objects, and tasks can be scheduled to physical computing resource objects with relatively shallow sleep states or relatively busy states as much as possible. This is beneficial for some physical computing resource objects to be in deep sleep states for as long as possible or have more opportunities to be in a busy state as much as possible, thereby reducing the frequency of physical computing resource objects entering and exiting the sleep state, thereby saving power consumption and allowing other physical computing resource objects to obtain a higher operating frequency as much as possible.

In some optional embodiments, as the running state of the physical computing resource object changes, such as from a sleep state to a working state, or from a working state to a sleep state, the sleep state transition prediction information of each physical computing resource object will change, so the sleep state transition prediction information of the physical computing resource object can also be updated to obtain more accurate sleep state transition prediction information, so as to perform more accurate resource scheduling based on the sleep state transition prediction information in the future. The update operation of the sleep state transition prediction information can be performed by the target sleep controller 11 during the maintenance process.

Specifically, when maintaining sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, the target sleep controller 11 can monitor the operating state of the multiple physical computing resource objects, where the operating state may include: working state and sleep state, and the sleep state may include multiple sleep states with different sleep depths; when monitoring that any physical computing resource object is awakened from a first sleep state, the actual sleep time of the physical computing resource object in the first sleep state can be obtained, wherein the first sleep state can be any sleep state, that is, when monitoring that any physical computing resource object is awakened from any sleep state, the target sleep controller 11 can obtain the actual sleep time of the physical computing resource object in any sleep state; further, according to the actual sleep time of the physical computing resource object in the first sleep state and the exit delay time of the first sleep state, the sleep state transfer prediction information corresponding to any physical computing resource object is updated.

The actual sleep time of the physical computing resource object in the first sleep state may be obtained by following steps R1-R2:

In step R1, the target sleep controller 11 can obtain the first time when the physical computing resource object enters the first sleep state, and the second time when the physical computing resource object is awakened from the first sleep state by the awakening event. The awakening event can be a timer arrival event or an interrupt event, etc. An interrupt event means that when the physical computing resource object executes a normal program, some abnormal conditions or special requests that need to be processed urgently appear in the system, and the physical computing resource object temporarily suspends the running task and turns to process the more urgent tasks that occur randomly.

Step R2: Calculate the actual sleep time of any physical computing resource object in the first sleep state according to the second time and the first time. The disclosed embodiment does not limit the implementation method of calculating the actual sleep time of any physical computing resource object in the first sleep state according to the second time and the first time. Optionally, the target sleep controller 11 may calculate the difference between the second time and the first time as the actual sleep time of any physical computing resource object in the first sleep state; or Alternatively, the target sleep controller 11 may also utilize a preset difference correction rule to correct the difference between the second time and the first time, and use the corrected difference as the actual sleep time of any physical computing resource object in the first sleep state; alternatively, the target sleep controller 11 may also perform a weighted summation of the second time and the first time according to their respective preset weights to obtain the actual sleep time of any physical computing resource object in the first sleep state.

After obtaining the actual sleep time of any physical computing resource object in the first sleep state, the target sleep controller 11 can update the sleep state transfer prediction information corresponding to any physical computing resource object according to the actual sleep time and the exit delay time of the first sleep state. Specifically, according to the different matching conditions between the actual sleep time of any physical computing resource object in the first sleep state and the exit delay time of the first sleep state, the method of updating the sleep state transfer prediction information corresponding to any physical computing resource object will also be different. In the embodiment of the present disclosure, the method of defining whether the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state is not limited, and can be flexibly defined according to application requirements. Among them, the target sleep controller 11 can determine whether the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state by any of the following methods.

Method 1: The target sleep controller 11 can determine whether the actual sleep time of any physical computing resource object in the first sleep state is the same as the exit delay time of the first sleep state. If they are the same, the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state. If they are not the same, the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state.

Method 2: The target sleep controller 11 can also calculate the error between the actual sleep time of any physical computing resource object in the first sleep state and the exit delay time of the first sleep state. If the error is within a preset error range, the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state. If the error is not within the preset error range, the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state.

Method 3: The target sleep controller 11 can determine whether the actual sleep time of any physical computing resource object in the first sleep state is within a certain multiple of the exit delay time of the first sleep state. If so, the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state. If not, the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state.

Method 4: The target sleep controller 11 may also determine whether the actual sleep time of any physical computing resource object in the first sleep state is greater than the exit delay time of the first sleep state; if so, further determine whether the difference between the actual sleep time of any physical computing resource object in the first sleep state and the exit delay time of the first sleep state is greater than the set A difference threshold is determined; if so, the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state; if any of the above judgment operations is no, the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state.

No matter how to define whether the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state, it can be divided into two situations: one is the situation that the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state, and the other is the situation that the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state. The following will introduce the two situations separately.

Case 1: When the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state, it means that the first sleep state is a relatively better sleep state for any physical computing resource object. Therefore, the probability information of any physical computing resource object entering the first sleep state again when it is awakened from the first sleep state can be relatively increased, so that it can enter the first sleep state again with a higher probability when it needs to sleep next time.

Further optionally, the target sleep controller 11 relatively increases the probability information of any physical computing resource object re-entering the first sleep state at the next sleep when it is awakened from the first sleep state, including: the probability information of any physical computing resource object re-entering the first sleep state at the next sleep when it is awakened from the first sleep state can be increased separately; or, the probability information of any physical computing resource object entering other sleep states at the next sleep when it is awakened from the first sleep state can be reduced separately, and the other sleep state refers to the sleep state other than the first sleep state; or, while increasing the probability information of any physical computing resource object re-entering the first sleep state at the next sleep when it is awakened from the first sleep state, the probability information of any physical computing resource object entering other sleep states at the next sleep when it is awakened from the first sleep state can be reduced.

For example, assuming that any physical computing resource object is awakened from sleep state C0 (sleep state C0 is an example of the first sleep state), before updating the sleep state transition prediction information corresponding to any physical computing resource object, if any physical computing resource object is awakened from sleep state C0, the probability of entering sleep state C0 in the next sleep state is 70%, the probability of entering sleep state C1 is 15%, and the probability of entering sleep state C2 is 15%. If the actual sleep time of any physical computing resource object in sleep state C0 matches the exit delay time of sleep state C0, the target sleep state C2 is 15%. When the controller 11 updates the sleep state transition prediction information, the probability of entering the sleep state C0 in the next sleep state when awakened from the sleep state C0 can be increased from 70% to 80%, or the probability of entering the sleep state C1 and the sleep state C2 in the next sleep state when awakened from the sleep state C0 can be reduced from 15% to 10%, or the probability of entering the sleep state C0 in the next sleep state when awakened from the sleep state C0 can be increased from 70% to 80%, and the probability of entering the sleep state C1 and the sleep state C2 can be reduced from 15% to 10%. In this Note that in the above example, the reduction ranges of the two probabilities of entering the sleep state C1 and the sleep state C2 are the same, but they can also be different, for example, the probability of entering the sleep state C1 is reduced by 10%, and the probability of entering the sleep state C2 is reduced by 5%.

Case 2: When the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state, it means that the first sleep state is not a relatively optimal sleep state for any physical computing resource object. Therefore, a sleep state whose exit delay time matches the actual sleep time can be determined and recorded as the second sleep state. The second sleep state is a relatively optimal sleep state for any physical computing resource object. Therefore, the probability information of any physical computing resource object entering the second sleep state when it is awakened from the first sleep state is relatively increased, so that it can enter the second sleep state with a higher probability when it needs to sleep next time. Among them, the second sleep state is a sleep state whose exit delay time matches the actual sleep time. That is to say, when the actual sleep time does not match the exit delay time of the first sleep state, the target sleep controller 11 can relatively increase the probability information corresponding to the sleep state whose exit delay time matches the actual sleep time, thereby updating the sleep state transition prediction information more accurately.

Specifically, in case 2, the target sleep controller 11 can relatively increase the probability information of any physical computing resource object entering the second sleep state when it is awakened from the first sleep state in the next sleep state based on the following implementation methods:

Implementation method 1: separately increase the probability information of any physical computing resource object entering the second sleep state in the next sleep state when it is awakened from the first sleep state. For example, assuming that any physical computing resource object is awakened from sleep state C0 (sleep state C0 is an example of the first sleep state), before updating the sleep state transition prediction information corresponding to any physical computing resource object, if any physical computing resource object is awakened from sleep state C0 (first sleep state), the probability of entering sleep state C0 in the next sleep state is 70%, the probability of entering sleep state C1 is 15% (second sleep state), and the probability of entering sleep state C2 is 15%, then if the actual sleep time of any physical computing resource object in sleep state C0 does not match the exit delay time of sleep state C0, the target sleep controller 11 can increase the probability of entering sleep state C1 in the next sleep state when it is awakened from sleep state C0 (first sleep state) from 15% to 70%.

Implementation method 2: increase the probability information of any physical computing resource object entering the second sleep state at the next sleep state when it is awakened from the first sleep state, and reduce the probability information of any physical computing resource object entering other sleep states at the next sleep state when it is awakened from the first sleep state, where other sleep states refer to sleep states other than the second sleep state (including the first sleep state). Continuing with the above example, if the actual sleep time of any physical computing resource object in sleep state C0 does not match the exit delay time of sleep state C0, it will enter sleep state C1 at the next sleep state when it is awakened from sleep state C0 (first sleep state). The probability of entering the sleep state C0 and C2 is increased from 15% to 70%, and the probability of entering the sleep state C0 and C2 is reduced accordingly. For example, the probability of entering the sleep state C0 in the next sleep state when awakened from the sleep state C0 (the first sleep state) can be reduced from 70% to 20%, and the probability of entering the sleep state C2 in the next sleep state when awakened from the sleep state C0 (the first sleep state) can be reduced from 15% to 10%.

Implementation method 3: separately reducing the probability information of any physical computing resource object entering the first sleep state in the next sleep state when it is awakened from the first sleep state. Using the above example, when the actual sleep time of any physical computing resource object in sleep state C0 does not match the exit delay time of sleep state C0, the target sleep controller 11 can reduce the probability of any physical computing resource object entering sleep state C0 in the next sleep state when it is awakened from sleep state C0 from 70% to 15%.

Implementation method 4: reduce the probability information of any physical computing resource object entering the first sleep state in the next sleep state when it is awakened from the first sleep state, and increase the probability information of any physical computing resource object entering other sleep states in the next sleep state when it is awakened from the first sleep state, where other sleep states refer to sleep states other than the first sleep state (including the second sleep state). Using the above example, when the actual sleep time of any physical computing resource object in sleep state C0 does not match the exit delay time of sleep state C0, the probability of entering sleep state C0 in the next sleep state when it is awakened from sleep state C0 (first sleep state) is reduced from 70% to 20%, and the probability of entering sleep states C1 and C2 is increased accordingly. For example, the probability of entering sleep state C1 in the next sleep state when it is awakened from sleep state C0 (first sleep state) can be increased from 15% to 60%, and the probability of entering sleep state C2 in the next sleep state when it is awakened from sleep state C0 (first sleep state) is increased from 15% to 20%.

Implementation method 5: separately reducing the probability information of any physical computing resource object entering other sleep states in the next sleep state when it is awakened from the first sleep state, where other sleep states refer to sleep states other than the first sleep state and the second sleep state. Using the above example, the target sleep controller 11 can reduce the probability of any physical computing resource object entering sleep state C2 in the next sleep state when it is awakened from sleep state C0 from 15% to 5%.

Through the above method, the target sleep controller can dynamically update the sleep state transfer prediction information according to the running status of the physical computing resource object, so as to maintain more accurate sleep state transfer prediction information. On the one hand, it is convenient to more accurately select the sleep state that the physical computing resource object needs to enter when sleep is needed next time. On the other hand, it can also provide a more accurate data basis for resource scheduling based on sleep state transfer prediction information.

In the above embodiments of the present disclosure, it is not limited to the specific method used by the target sleep controller to maintain the sleep state transition prediction information. In some optional embodiments, the target sleep controller can create multiple A data structure corresponding to each of the multiple physical computing resource objects is initialized to obtain the initial value of the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects; and then as the running state of the physical computing resource object changes, the sleep state transfer prediction information recorded in the data structure can be dynamically updated according to the method described above. Among them, the data structure is used to store the sleep state transfer prediction information corresponding to the corresponding physical computing resource object. Among them, the data structure refers to a data structure capable of storing information, including but not limited to: arrays, lists, linked lists, queues, etc. The disclosed embodiment will take an array as an example for illustrative explanation, and the array will be called a sleep state weight array, but the form of the data structure is not limited.

Specifically, the target sleep controller can create sleep state weight arrays corresponding to multiple physical computing resource objects during the initialization process. The sleep state weight array is a multidimensional array including multiple rows and multiple columns, as shown in Table 2. Among them, a row in the sleep state weight array represents a sleep state that the corresponding physical computing resource object may be in when it is awakened, and a column represents a sleep state that the corresponding physical computing resource object may enter when it sleeps next time. The array elements where the rows and columns intersect represent the weight of the corresponding physical computing resource object entering the sleep state represented by the column when it is awakened from the sleep state represented by the row and the next time it sleeps. The larger the weight, the higher the probability. It should be noted that before initialization, the element value of the sleep state weight array should be null.

Next, the target sleep controller 11 may initialize the sleep state weight arrays corresponding to each of the multiple physical computing resource objects to obtain the initial values of the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects. In an optional embodiment, the target sleep controller 11 may initialize the array elements with the same number of rows and columns to the first value, and initialize the array elements with different number of rows and columns to the second value, wherein the first value is greater than the second value, and the sum of the first value and the second value in the same row is the third value. Wherein, as shown in Table 2, the third value may be 100, and the sum of the first value and the second value in the same row is 100. In addition, the third value may also be any value, such as 90, 120 or 150, etc., which is not limited in this embodiment. Correspondingly, the first value and the second value may also be any value under the condition that "the first value is greater than the second value" and "the sum of the first value and the second value in the same row is the third value", which is not limited in this embodiment. In Table 2, the first value is 70 and the second value is 15 as an example for illustration. The following table shows a multidimensional array after initialization, which is only used for illustrative purposes to better describe the meanings of the rows and columns.

Table 2

That is to say, in the sleep state weight array shown in Table 2 above, the row where the sleep state C0 is located can indicate that: when the corresponding physical computing resource object is awakened from the sleep state C0, the weight of entering the sleep state C0 at the next sleep is 70%, when it is awakened from the sleep state C0, the weight of entering the sleep state C1 at the next sleep is 15%, and when it is awakened from the sleep state C0, the weight of entering the sleep state C2 at the next sleep is 15%; the row where the sleep state C1 is located can indicate that: when the corresponding physical computing resource object is awakened from the sleep state C1, the weight of entering the sleep state C0 at the next sleep is 15%. The weight is 15%, the weight of entering the sleep state C1 at the next sleep when awakened from the sleep state C1 is 70%, and the weight of entering the sleep state C2 at the next sleep when awakened from the sleep state C1 is 15%; the row where the sleep state C2 is located can indicate: the weight of the corresponding physical computing resource object entering the sleep state C0 at the next sleep when awakened from the sleep state C2 is 15%, the weight of entering the sleep state C1 at the next sleep when awakened from the sleep state C2 is 15%, and the weight of entering the sleep state C2 at the next sleep when awakened from the sleep state C2 is 70%.

On this basis, combined with the sleep state weight array and the above-listed initialization methods, the "relatively increasing the probability information of any physical computing resource object entering the first sleep state again in the next sleep state when it is awakened from the first sleep state" and "the target sleep controller relatively increasing the probability information of any physical computing resource object entering the second sleep state in the next sleep state when it is awakened from the first sleep state" mentioned in the above embodiments are exemplified for the adjustment method of the probability information. It should be noted that no matter how the multidimensional array is updated, the sum of the values in the same row of the multidimensional array after increase or decrease is still the preset third value. This will be further explained below.

Optionally, the target sleep controller 11 may relatively increase the probability information of any physical computing resource object re-entering the first sleep state when it is awakened from the first sleep state when the actual sleep time of any physical computing resource object in the first sleep state matches the exit delay time of the first sleep state. Among them, the target sleep controller 11 may relatively increase the probability information of any physical computing resource object re-entering the first sleep state when it is awakened from the first sleep state to a fourth value, and relatively reduce the probability information of any physical computing resource object entering other sleep states when it is awakened from the first sleep state to a fifth value, the fourth value may be greater than the fifth value, and the sum of the values in the same row is still the third value in the aforementioned embodiment.

Optionally, the target sleep controller 11 may relatively increase the probability information of any physical computing resource object entering the second sleep state when it is awakened from the first sleep state when the actual sleep time of any physical computing resource object in the first sleep state does not match the exit delay time of the first sleep state. Among them, the target sleep controller 11 may relatively increase the probability information of any physical computing resource object entering the second sleep state when it is awakened from the first sleep state to the sixth value, and relatively reduce the probability information of any physical computing resource object entering other sleep states when it is awakened from the first sleep state to the seventh value, the sixth value may be greater than the seventh value, and the sum of the values in the same row is still the third value in the aforementioned embodiment.

In this way, based on the aforementioned initialized data structure, the target sleep controller 11 can use the data structure to According to the structure, the sleep state transfer prediction information corresponding to any physical computing resource object is updated more quickly.

In some optional embodiments, in addition to maintaining the sleep state transfer prediction information corresponding to each physical computing resource object, the target sleep controller 11 may also monitor the task queue in any physical computing resource object. When the task queue in the physical computing resource object is empty, or when the task queue in the physical computing resource object is empty and continues for a preset time, the target sleep controller 11 may determine that the physical computing resource object needs to sleep. When it is detected that the physical computing resource object needs to sleep, the target sleep controller 11 may determine the third sleep state according to the sleep state transfer prediction information corresponding to the physical computing resource object, and control any physical computing resource object to enter the third sleep state. Among them, the third sleep state is any one of the multiple sleep states, and is the sleep state that any physical computing resource object needs to enter this time. The above process will be further explained below in conjunction with the internal architecture and working principle of the operating system shown in Figure 3.

As shown in FIG3 , in the first step (shown in ① of FIG3 ), the ACPI specification can be used to configure the sleep state. Specifically, the physical machine can obtain a fixed ACPI description table, which defines ACPI information of various fixed hardware. From the description table, various sleep states supported by any physical computing resource object and the exit delay time corresponding to various sleep states are obtained. In the second step (shown in ② of FIG3 ), the target resource scheduler 12 can perform resource scheduling for each physical computing resource object; in the third step (shown in ③ of FIG3 ), during the resource scheduling process, if it is detected that the task queue in any physical computing resource object is empty, or the task queue in the physical computing resource object is empty and lasts for a preset time, the target sleep controller 11 can determine the third sleep state that the physical computing resource object needs to enter according to the sleep state transfer prediction information of the physical computing resource object maintained, and provide the third sleep state to the sleep control module 13, and the sleep control module 13 selects the driver module 14 corresponding to the physical computing resource object, and uses the driver module 14 to control the physical computing resource object to enter the third sleep state through the firmware corresponding to the physical computing resource object, as shown in ④ of FIG3 . It is to be noted that in FIG3 , the operating system including the target sleep controller 11 , the target resource scheduler 12 , the sleep management module 13 and the driver module 14 is illustrated as an example. This is only a schematic illustration of the internal architecture of the operating system, and the internal implementation architecture of the operating system is not limited to this.

Specifically, when the target sleep controller 11 determines the third sleep state according to the sleep state transition prediction information corresponding to the physical computing resource object, it can be implemented based on the following steps Y1 to Y4:

Step Y1, obtaining the sleep state most recently exited by any physical computing resource object as the fourth sleep state. For example, the sleep states most recently exited by the physical computing resource object are sleep state C1 and sleep state C2, respectively, and the target sleep controller 11 may use the sleep state C2 most recently exited by the physical computing resource object as the fourth sleep state.

Step Y2: Obtain, from the sleep state transition prediction information corresponding to any physical computing resource object, probability information of entering each sleep state in the next sleep state when awakened from the fourth sleep state.

Step Y3: According to the probability information of entering each sleep state in the next sleep state when awakened from the fourth sleep state, the sleep state with the largest probability information is selected as the candidate sleep state.

For example, the target sleep controller 11 obtains the probability information of entering each sleep state in the next sleep state when awakened from the fourth sleep state: sleep state C0 is 15%, sleep state C1 is 70% and sleep state C2 is 15%. Therefore, the target sleep controller 11 can select the sleep state C1 with the largest probability information as the candidate sleep state.

Step Y4: When the exit delay time of the candidate sleep state is less than the maximum tolerable exit delay time of the system, the candidate sleep state is used as the third sleep state. The maximum tolerable exit delay time of the system can be a preset fixed value, a value dynamically configured by the system according to the system operation status, or a value customized by the user, which is not limited in this embodiment.

In addition to the above-mentioned step Y4 "the exit delay time of the candidate sleep state is less than the maximum tolerable exit delay time of the system", the target sleep controller 11 may also select another sleep state whose exit delay time is less than the maximum tolerable exit delay time of the system as the third sleep state when the exit delay time of the candidate sleep state is greater than or equal to the maximum tolerable exit delay time of the system. Among them, when there is one sleep state whose exit delay time is less than the maximum tolerable exit delay time of the system, the target sleep controller 11 may select this sleep state as the third sleep state. When there are multiple sleep states whose exit delay time is less than the maximum tolerable exit delay time of the system, the target sleep controller 11 may select the sleep state whose exit delay time is less than the maximum tolerable exit delay time of the system and whose probability information is the largest as the third sleep state.

Through the above manner, the target sleep controller 11 can determine the third sleep state more accurately according to the sleep state transition prediction information corresponding to the physical computing resource object.

On the basis of the corresponding execution logic of the target sleep controller 11 introduced in the above-mentioned embodiments, the target resource scheduler 12 may also cooperate with the target sleep controller to complete the joint scheduling of the target tasks, which will be further explained below.

The target resource scheduler 12 may select a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the plurality of physical computing resource objects, and may implement the selection based on the following steps U1-U2:

Step U1, responding to a resource scheduling trigger event, generating sleep habit information corresponding to a plurality of physical computing resource objects according to sleep state transition prediction information corresponding to each of the plurality of physical computing resource objects. The resource scheduling trigger event includes: time interruption, peripheral interruption, current application actively giving up running or higher priority application needing to run, etc.

Step U2: Select a target physical computing resource object from the multiple physical computing resource objects according to the sleep habit information corresponding to each of the multiple physical computing resource objects.

Through the above manner, the target resource scheduler 12 can more accurately select a target physical computing resource object from a plurality of physical computing resource objects based on the sleep habit information corresponding to each of the plurality of physical computing resource objects.

In some optional embodiments, when executing the above step U1, the target resource scheduler 12 may generate the sleep habit information corresponding to each of the plurality of physical computing resource objects based on the following steps U11-U12:

Step U11: For any physical computing resource object, obtain the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state from the sleep state transition prediction information corresponding to the physical computing resource object.

Step U12: Generate sleep habit information corresponding to any physical computing resource object based on the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state. The target resource scheduler 12 may construct a data structure and write the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state into the data structure to obtain the sleep habit information corresponding to the physical computing resource object.

The data structure includes but is not limited to: arrays, lists, linked lists, queues, etc. The target resource scheduler 12 can also directly use the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state as the sleep habit information corresponding to the physical computing resource object.

For example, assume that there are three sleep states, namely, sleep state C0, sleep state C1, and sleep state C2. The target resource scheduler 12 can obtain from the sleep state transition prediction information corresponding to any physical computing resource object that the probability of the physical computing resource object exiting from sleep state C0 and re-entering sleep state C0 at the next sleep is 40%, the probability of exiting from sleep state C1 and re-entering sleep state C1 at the next sleep is 35%, and the probability of exiting from sleep state C2 and re-entering sleep state C2 at the next sleep is 60%. Based on this, the target resource scheduler 12 can use the above probability information as the sleep habit information corresponding to any physical computing resource object.

In some other optional embodiments, when executing the above step U1, the target resource scheduler 12 can also divide the sleep habit information into two sleep habit types, namely, shallow sleep habit and deep sleep habit, in the following manner to obtain the sleep habit information. If the sleep habit type corresponding to a physical computing resource object is shallow sleep habit, it means that the physical computing resource object is used to being in a shallow sleep state; if the sleep habit type corresponding to a physical computing resource object is deep sleep habit, it means that the physical computing resource object is used to being in a deep sleep state. This will be further explained below.

If the probability information of any physical computing resource object exiting the shallowest sleep state and entering the shallowest sleep state at the next sleep is the largest, the target resource scheduler 12 can determine that the sleep habit type corresponding to any physical computing resource is a shallow sleep habit, and generate the sleep probability of any physical computing resource object under the shallow sleep habit according to the maximum probability information. The target resource scheduler 12 can directly use the maximum probability information as the sleep probability of any physical computing resource object under the shallow sleep habit; or multiply the maximum probability information by a preset probability coefficient to obtain the sleep probability of any physical computing resource object under the shallow sleep habit. The maximum probability information can also be superimposed with a preset correction value to obtain the sleep probability of any physical computing resource object under the shallow sleep habit, which is not limited in this embodiment.

If the probability information that any physical computing resource object exits the deepest sleep state and enters the deepest sleep state at the next sleep is the largest, the sleep habit type corresponding to any physical computing resource object is determined to be a deep sleep habit, and the sleep probability of any physical computing resource under the deep sleep habit is generated according to the maximum probability information. The method of generating the sleep probability of any physical computing resource under the deep sleep habit according to the maximum probability information is the same as the method of generating the sleep probability of any physical computing resource object under the shallow sleep habit, which will not be repeated here.

In this way, the target resource scheduler 12 can not only obtain the sleep habit information more accurately, but also classify the sleep habit information more accurately into two types: shallow sleep habit and deep sleep habit.

In some optional embodiments, when the target resource scheduler 12 selects a target physical computing resource object from a plurality of physical computing resource objects, it may determine at least one candidate physical computing resource object corresponding to a shallow sleep habit according to the sleep habit information corresponding to each of the plurality of physical computing resource objects. For example, there are five physical computing resource objects, namely, physical computing resource object D1 to physical computing resource object D5, and their corresponding sleep habit information respectively corresponds to a shallow sleep habit, a shallow sleep habit, a deep sleep habit, a deep sleep habit, and a deep sleep habit. The target resource scheduler 12 may select physical computing resource object D1 and physical computing resource object D2 as candidate physical computing resource objects.

After determining at least one candidate physical computing resource object corresponding to the shallow sleep habit, the target resource scheduler 12 may select a candidate physical computing resource object whose sleep probability meets the requirements as the target physical computing resource object according to the sleep probability of the at least one candidate physical computing resource object under the shallow sleep habit.

Specifically, the target resource scheduler 12 may select a target physical computing resource object from at least one candidate physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit, combined with other auxiliary information. Among them, the other auxiliary information may include: at least one of the type of resource scheduling triggering event, the current operating state of the candidate physical computing resource object, and the current utilization rate of the candidate physical computing resource object. The following will be explained in different cases.

Optionally, taking other auxiliary information as the type of resource scheduling trigger event as an example, when the target resource scheduler 12 selects a target physical computing resource object from at least one candidate physical computing resource object, it may select the target physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under shallow sleep habits and the type of resource scheduling trigger event. The target resource scheduler 12 may first select at least one target candidate physical computing resource object from at least one candidate physical computing resource object based on the type of resource scheduling trigger event, wherein the correspondence between the type of resource scheduling trigger event and its corresponding physical computing resource object may be maintained by the target resource scheduler 12. Thereafter, the target resource scheduler 12 may select a candidate with the highest sleep probability from at least one target candidate physical computing resource object. The physical computing resource object is used as the target physical computing resource object.

Optionally, taking the current operating state of the candidate physical computing resource object as an example of other auxiliary information, the target resource scheduler 12 may select the target physical computing resource object according to the current operating state of the at least one candidate physical computing resource object and the sleep probability under the shallow sleep habit when selecting the target physical computing resource object from the at least one candidate physical computing resource object. Among them, the target resource scheduler 12 may first select at least one target candidate physical computing resource object whose operating state meets the preset state condition from the at least one candidate physical computing resource object according to the current operating state of the at least one candidate physical computing resource object, and select the candidate physical computing resource object with the highest sleep probability from the at least one target candidate physical computing resource object as the target physical computing resource object.

Optionally, taking the current utilization rate of the candidate physical computing resource object as an example of other auxiliary information, when the target resource scheduler 12 selects the target physical computing resource object from at least one candidate physical computing resource object, it can select the candidate physical computing resource object with the utilization rate that meets the requirements and the largest sleep probability as the target physical computing resource object according to the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit and the current utilization rate. Among them, the requirement corresponding to the utilization rate can be set as any condition according to actual needs, such as not more than 80%, not more than 70% or not more than 60%, etc., which is not limited in this embodiment.

In the above manner, since other auxiliary information is combined, the target resource scheduler 12 can more accurately select a target physical computing resource object from at least one candidate physical computing resource object according to the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit.

On the basis of the above embodiments, the multiple physical computing resource objects mentioned in the above embodiments may be multi-core processors. In addition, the physical machine mentioned in the above embodiments may be applicable to vertical/dedicated application scenarios. When applied to vertical/dedicated scenarios, the operating system of the physical machine may only include a target resource scheduler and a target sleep controller to implement the above embodiments.

In addition, the physical machine can also be configured for application in various general scenarios. In general scenarios, the operating system mentioned in the above embodiments may also include other resource schedulers and other sleep controllers. Different sleep controllers have different sleep control logics, and different resource schedulers have different resource scheduling logics. It also allows users to flexibly choose which sleep controller and which resource scheduler to use according to application requirements.

For example, the operating system may include several other resource schedulers, several other sleep controllers, and the target sleep controller and target resource scheduler in the aforementioned embodiment. Different other resource schedulers and different other sleep controllers are independent of each other. The concept of mutual independence means that each other resource scheduler can perform resource scheduling according to its own scheduling logic, and each other sleep controller can perform sleep control according to its own sleep control logic. In other words, the resource scheduling process of other resource schedulers does not depend on the sleep control process of any other sleep controller. The scheduler and other sleep controllers are decoupled. Among them, the user can use any combination of several other resource schedulers and several other sleep controllers. Specifically, the physical machine can determine a resource scheduler and a sleep controller from several other resource schedulers and several other sleep controllers in response to the user's selection operation. The selected resource scheduler and sleep controller are configured to take effect, so as to schedule the target task based on the resource scheduler and the sleep controller. The embodiment of the present disclosure does not limit the sleep control logic of any other sleep controller, nor does it limit the resource scheduling logic of any other resource scheduler. The physical machine can also select the target resource scheduler and the target sleep controller to cooperate in the scheduling of the target task in response to the user's selection operation, that is, the target resource scheduler and the target sleep controller are configured to take effect. Relative to other resource schedulers and other sleep controllers, the target resource scheduler and the target sleep controller need to be used in combination. The two are related and will affect each other. It should be noted that when any resource scheduler in other resource schedulers and any sleep controller in other sleep controllers are configured to take effect, the target resource scheduler and the target sleep controller are configured to be invalid.

The following will take the Linux operating system as an example to exemplify the sleep control logic of other sleep controllers.

In Linux systems, multiple sleep controllers such as menu, timer events oriented (TEO), and ladder can be configured. Among them, the menu sleep controller is a sleep controller that attempts to predict the idle duration and uses the predicted value to select a sleep state for a physical computing resource object. For the detailed sleep control logic of the menu sleep controller, please refer to the description in the following embodiment. The teo sleep controller is a sleep controller that attempts to find the deepest sleep state suitable for given conditions. Specifically, the teo sleep controller monitors the idle duration of each physical computing resource object, and attempts to compare the observed idle duration value with the available sleep states, and uses this information to select the sleep state that is most likely to "match" the upcoming idle interval of the physical computing resource object. Among them, the ladder sleep controller is a sleep controller that gradually enters various sleep states from shallow to deep, that is, it will preferentially enter the shallowest sleep state, and determine whether to enter the next deeper sleep state according to the sleep duration.

In the disclosed embodiment, it is assumed that the user chooses to use other sleep controllers and other resource schedulers. Taking the menu sleep controller and the CFS (Completely Fair Scheduler) resource scheduler as an example, as shown in FIG4 , the working logic between the two is decoupled. The CFS will perform resource scheduling for the physical computing resource objects according to its own resource scheduling strategy. In this process, the menu will perform sleep control according to its own sleep control logic. The sleep control logic is as follows:

(1) The sleep controller menu can record each historical actual sleep time to predict the next sleep time.

(2) The sleep controller menu can calculate the correction coefficient based on the above historical actual sleep time. The correction coefficient can be used to predict the impact of factors such as I/O (Input/Output) on the actual sleep time so as to correct these impacts in subsequent steps. Specifically, the sleep controller can calculate the correction coefficient according to the following formula:

Correction coefficient = 1024 × actual sleep time ÷ estimated sleep time

(3) If each historical actual sleep time meets the following conditions: standard deviation <20 and variance <=400, then the sleep controller menu can take the average value of each historical actual sleep time as the next sleep time.

(4) The sleep controller menu can obtain the number of threads in the I/O wait state, where the I/O wait state refers to the state of waiting for the disk I/O request to be completed.

(5) The sleep controller menu can correct the next sleep time according to the number of threads in the I/O wait state and the correction factor.

(6) The sleep controller menu can obtain the exit delay time set by pm qos (power management quality of service) within the device and system range, and select the minimum value as the first delay tolerance latency_req. After that, the sleep controller can predict the second delay tolerance based on the following formula:

Second delay tolerance = predicted next sleep time ÷ (10 × number of threads in I/O wait state + 1) (Formula 2)

(7) The sleep controller menu may use the second delay tolerance to correct the first delay tolerance, that is, select the minimum value from the second delay tolerance and the first delay tolerance as the corrected target tolerance.

(8) The sleep controller menu can select the deepest sleep state with an exit delay time less than or equal to the target tolerance from multiple preset sleep states, and then the sleep controller can control the physical computing resource object to enter the sleep state.

It should be noted that the control logic of the menu sleep controller introduced above has many defects, such as the inability to link with the resource scheduler to reduce the number of times the physical computing resource object enters and exits the sleep state, and the frequent number of physical computing resource objects entering and exiting the sleep state leads to high power consumption. In addition, the high failure rate of predicting the sleep state also leads to waste of power. This also verifies from another perspective the technical effect produced by the linkage between the target resource scheduler and the target sleep controller in the aforementioned embodiments of the present disclosure.

In addition to the physical machines provided in the above embodiments, the embodiments of the present disclosure also provide a computing resource processing method, which will be described below in conjunction with the accompanying drawings.

FIG5 is a flow chart of a computing resource processing method provided by an exemplary embodiment of the present disclosure, which may include the steps shown in FIG5 :

Step 51: Maintain sleep state transition prediction information corresponding to each of the multiple physical computing resource objects, wherein the sleep state transition prediction information includes probability information of the corresponding physical computing resource object entering each of the multiple sleep states at the next sleep when awakened from any of the multiple sleep states.

Step 52: Select a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the plurality of physical computing resource objects.

Step 53: Schedule the target task to be scheduled to the target physical computing resource object.

It should be noted that the execution subject of the embodiment of the method may be an operating system, specifically a target resource scheduler and a target sleep controller in the operating system. The execution subjects of different steps may be different, for example, some steps may be executed by the target resource scheduler, and some steps may be executed by the target sleep controller, which is not limited in this embodiment.

In this embodiment, the resource scheduling of physical computing resource objects can be combined with the sleep control of physical computing resource objects. By maintaining the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, the physical computing resource objects are scheduled according to the maintained sleep state transfer prediction information. This can schedule tasks to physical computing resource objects with relatively shallow sleep states or relatively busy states as much as possible, which is beneficial for some physical computing resource objects to be in a deep sleep state for as long as possible, and for some physical computing resource objects to be in a busy state as much as possible, thereby reducing the frequency of physical computing resource objects entering and exiting the sleep state.

In some optional embodiments, sleep state transfer prediction information corresponding to each of a plurality of physical computing resource objects is maintained, including: when any physical computing resource object is monitored to be awakened from a first sleep state, obtaining the actual sleep time of any physical computing resource object in the first sleep state, where the first sleep state is any sleep state; and updating the sleep state transfer prediction information corresponding to any physical computing resource object according to the actual sleep time and the exit delay time of the first sleep state.

In some optional embodiments, when monitoring any physical computing resource object to be awakened from a first sleep state, the actual sleep time of any physical computing resource object in the first sleep state is obtained, including: obtaining a first time when any physical computing resource object enters the first sleep state, and a second time when it is awakened from the first sleep state by a wake-up event; and calculating the actual sleep time of any physical computing resource object in the first sleep state based on the second time and the first time.

In some optional embodiments, the sleep state transfer prediction information corresponding to any physical computing resource object is updated according to the actual sleep time and the exit delay time of the first sleep state, including: when the actual sleep time matches the exit delay time of the first sleep state, the probability information of any physical computing resource object entering the first sleep state again when it is awakened from the first sleep state at the next sleep state is relatively increased; when the actual sleep time does not match the exit delay time of the first sleep state, the probability information of any physical computing resource object entering the second sleep state when it is awakened from the first sleep state at the next sleep state is relatively increased, and the second sleep state is a sleep state whose exit delay time matches the actual sleep time.

In some optional embodiments, relatively increasing the probability information of any physical computing resource object entering the first sleep state again at the next sleep state when being awakened from the first sleep state includes: increasing the probability information of any physical computing resource object entering the first sleep state again at the next sleep state when being awakened from the first sleep state; and/or reducing the probability information of any physical computing resource object entering other sleep states at the next sleep state when being awakened from the first sleep state, where other sleep states refer to sleep states other than the first sleep state.

In some optional embodiments, relatively increasing the probability information of any physical computing resource object entering the second sleep state at the next sleep state when it is awakened from the first sleep state includes: increasing the probability information of any physical computing resource object entering the second sleep state at the next sleep state when it is awakened from the first sleep state; and/or, reducing the probability information of any physical computing resource object entering the first sleep state at the next sleep state when it is awakened from the first sleep state; and/or, reducing the probability information of any physical computing resource object entering other sleep states at the next sleep state when it is awakened from the first sleep state, and the other sleep states refer to sleep states other than the first sleep state and the second sleep state.

In some optional embodiments, the method also includes: during the initialization process, creating a data structure corresponding to each of the multiple physical computing resource objects, the data structure being used to store sleep state transfer prediction information corresponding to the corresponding physical computing resource object; initializing the data structure corresponding to each of the multiple physical computing resource objects to obtain an initial value of the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects.

In some optional embodiments, a data structure corresponding to each of the plurality of physical computing resource objects is created, including: creating a sleep state weight array corresponding to each of the plurality of physical computing resource objects, the sleep state weight array being a multidimensional array including a plurality of rows and a plurality of columns; wherein a row represents a sleep state that the corresponding physical computing resource object may be in when awakened, and a column represents a sleep state that the corresponding physical computing resource object may enter when it sleeps next time; array elements where rows and columns intersect represent the weight of the corresponding physical computing resource object entering the sleep state represented by the column when it is awakened from the sleep state represented by the row and when it sleeps next time, and the greater the weight, the higher the probability.

In some optional embodiments, data structures corresponding to multiple physical computing resource objects are initialized, including: initializing array elements with the same number of rows and columns to a first value, and initializing array elements with different number of rows and columns to a second value; wherein the first value is greater than the second value, and the sum of the first value and the second value in the same row is a third value.

In some optional embodiments, the method further includes: when it is monitored that any physical computing resource object needs to sleep, determining a third sleep state according to sleep state transfer prediction information corresponding to any physical computing resource object; and controlling any physical computing resource object to enter the third sleep state.

In some optional embodiments, the third sleep state is determined according to the sleep state transfer prediction information corresponding to any physical computing resource object, including: obtaining the sleep state most recently exited by any physical computing resource object as the fourth sleep state; obtaining the probability information of entering each sleep state at the next sleep state when awakened from the fourth sleep state from the sleep state transfer prediction information corresponding to any physical computing resource object; according to the probability information of entering each sleep state at the next sleep state when awakened from the fourth sleep state, selecting the sleep state with the largest probability information as the candidate sleep state; when the exit delay time of the candidate sleep state is less than the maximum tolerable exit delay time of the system, taking the candidate sleep state as the third sleep state.

In some optional embodiments, the method further includes: when the exit delay time of the candidate sleep state is greater than or equal to the system In the case where the maximum tolerable exit delay time of the system is greater than the maximum tolerable exit delay time of the system, another sleep state whose exit delay time is less than the maximum tolerable exit delay time of the system is selected as the third sleep state.

In some optional embodiments, another sleep state in which the exit delay time is less than the maximum tolerable exit delay time of the system is selected as the third sleep state, including: when there are multiple sleep states in which the exit delay time is less than the maximum tolerable exit delay time of the system, selecting the sleep state in which the exit delay time is less than the maximum tolerable exit delay time of the system and the probability information is the largest as the third sleep state.

In some optional embodiments, a target physical computing resource object is selected based on sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, including: responding to a resource scheduling trigger event, generating sleep habit information corresponding to each of the multiple physical computing resource objects based on the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects; and selecting a target physical computing resource object from the multiple physical computing resource objects based on the sleep habit information corresponding to each of the multiple physical computing resource objects.

In some optional embodiments, sleep habit information corresponding to each of the multiple physical computing resource objects is generated based on sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, including: for any physical computing resource object, obtaining probability information that any physical computing resource object exits from any sleep state and re-enters the sleep state at the next sleep state from the sleep state transfer prediction information corresponding to any physical computing resource object; generating sleep habit information corresponding to any physical computing resource object based on probability information that any physical computing resource object exits from any sleep state and re-enters the sleep state at the next sleep state.

In some optional embodiments, sleep habit information corresponding to any physical computing resource object is generated based on probability information of any physical computing resource object exiting from any sleep state and entering the sleep state again at the next sleep, including: if the probability information of any physical computing resource object exiting from the shallowest sleep state and entering the shallowest sleep state at the next sleep is the largest, determining that the sleep habit type corresponding to any physical computing resource is a shallow sleep habit, and generating the sleep probability of any physical computing resource object under the shallow sleep habit based on the largest probability information; if the probability information of any physical computing resource object exiting from the deepest sleep state and entering the deepest sleep state at the next sleep is the largest, determining that the sleep habit type corresponding to any physical computing resource object is a deep sleep habit, and generating the sleep probability of any physical computing resource under the deep sleep habit based on the largest probability information.

In some optional embodiments, a target physical computing resource object is selected from a plurality of physical computing resource objects according to sleep habit information corresponding to each of the plurality of physical computing resource objects, including: determining at least one candidate physical computing resource object corresponding to a shallow sleep habit according to the sleep habit information corresponding to each of the plurality of physical computing resource objects; and selecting a candidate physical computing resource object whose sleep probability meets the requirements as a target physical computing resource object according to the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit.

In some optional embodiments, the at least one candidate physical computing resource object is selected according to the sleep state of the at least one candidate physical computing resource object in the shallow sleep state. The method comprises: selecting a target physical computing resource object from at least one candidate physical computing resource object according to the sleep probability of at least one candidate physical computing resource object in a shallow sleep state and in combination with other auxiliary information; wherein the other auxiliary information comprises at least one of the type of resource scheduling triggering event, the current operating state of the candidate physical computing resource object, and the current utilization rate of the candidate physical computing resource object.

In some optional embodiments, a target physical computing resource object is selected from at least one candidate physical computing resource object based on the sleep probability of at least one candidate physical computing resource object in a shallow sleep habit in combination with other auxiliary information, including: based on the sleep probability of at least one candidate physical computing resource object in a shallow sleep habit and the current utilization rate, selecting a candidate physical computing resource object whose utilization rate meets the requirements and whose sleep probability is the highest as the target physical computing resource object.

The detailed implementation and beneficial effects of each step in the method of this embodiment have been described in detail in the aforementioned embodiments and will not be elaborated here.

The following takes a multi-core CPU and the CPU cores it contains as an example, and describes in detail the implementation methods of the embodiments of the method shown in FIG. 5 in combination with FIG. 6a, FIG. 6b, FIG. 6c and FIG. 6d.

Among them, Figure 6a is a schematic diagram of the deployment and implementation of the computing resource processing method provided by an embodiment of the present disclosure in an actual application scenario.

Among them, the target sleep controller can be initialized and create a data structure for each CPU core (i.e., the physical computing resource object in the previous text) during initialization, such as an array in the format of weight[][], wherein a row represents a sleep state that the corresponding physical computing resource object may be in when it is awakened, and a column represents a sleep state that the corresponding physical computing resource object may enter during the next sleep. The array elements where the rows and columns intersect represent the weight of the corresponding physical computing resource object entering the sleep state represented by the column during the next sleep when it is awakened from the sleep state represented by the row. The larger the weight, the higher the probability.

The target sleep controller can initialize the data structure. Specifically, the target sleep controller can initialize the elements of the intersection of the row and column to 70 when the array i and j subscripts are the same. The target sleep controller can initialize these elements to (100-70)/2=15 when the array i and j subscripts are different. It should be noted that, regardless of the initialization or subsequent element update process, the elements of each row should always satisfy the condition of adding up to 100. Figure 6a shows an example of the initialized data structure, taking three sleep states C0-C2 as an example, and taking 8 CPU cores as an example. The 8 CPU cores shown in Figure 6a are core 0 to core 7.

After initialization, the target sleep controller can also monitor the status of each CPU core and update the data structure according to the status monitoring result. The update process will be described below in conjunction with FIG. 6 b.

As shown in FIG6b , the target resource scheduler may start sleeping when the task queue in the physical computing resource object is empty. The target sleep controller can determine the actual sleep time according to the time of the last sleep state entry and the time of being awakened by the wake-up event. The wake-up event can be a timer arrival event or an interrupt event. After that, the target sleep controller can determine the sleep state that best matches the actual duration of the sleep state.

When the target sleep controller determines that the actual sleep time matches the last predicted sleep state, the weight of the sleep state can be increased by 20 and the other weights can be reduced by 10; when the target sleep controller determines that the actual sleep time does not match the last predicted sleep state, the weight of the sleep state that best matches the actual duration of the sleep can be increased by 15 and the weight of the last predicted sleep state can be reduced by 15.

As shown in FIG. 6 b , in this way, the target sleep controller can update the data structure more accurately.

Based on the above content, the target sleep controller can more accurately select a sleep state from multiple sleep states based on the data structure. This will be further described below in conjunction with FIG. 6c.

As shown in FIG6c, for a CPU core that needs to enter a sleep state again after being awakened, the target sleep controller can obtain the sleep state that it has exited most recently (i.e., "obtain the last sleep state of the last exit" in FIG6b), and use the sleep state as a row, and determine the sleep state with the highest weight in the multiple column elements corresponding to the row as the state to be entered in the next sleep state. Afterwards, the target sleep controller can determine whether the exit delay time of the sleep state is less than the maximum tolerable exit delay event of the system. If not, the target sleep controller continues to select a sleep state at a shallower level and continues to determine whether the exit delay time of the selected sleep state is less than the maximum tolerable exit delay time of the system; if yes, the sleep state selection can be completed.

Based on the corresponding execution logic of the target sleep controller described above, the target resource scheduler can also cooperate with the target sleep controller to complete the joint scheduling of the target task, which will be further explained below in conjunction with FIG. 6d.

As shown in FIG6d , the target resource scheduler can respond to events such as time interrupts, peripheral interrupts, the current application actively giving up running, or higher priority applications needing to run, and read the data structure (weight[][] array) of each CPU core. For any CPU core, the target resource scheduler can determine whether the deepest sleep state of the CPU core has the highest weight. If so, the target resource scheduler can continue to search for the CPU core in the shallowest sleep state and with the highest weight, and schedule the target task on the CPU core in the deepest sleep state to the CPU core in the shallowest sleep state and with the highest weight, and then run the task at the head of the task queue on the CPU core in the shallowest sleep state and with the highest weight. In this way, tasks can be scheduled as much as possible to physical computing resource objects with relatively shallow sleep states or relatively busy states, which is conducive to keeping some physical computing resource objects in deep sleep states for as long as possible, keeping some physical computing resource objects in busy states as much as possible, and reducing the frequency of physical computing resource objects entering and exiting sleep states.

It should be noted that some of the processes described in the above embodiments and the accompanying drawings include steps that are performed in a specific order. There are multiple operations in this article, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 51, 52, etc., are only used to distinguish different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., and do not represent the order of precedence, nor do they limit the "first" and "second" to be different types.

The present disclosure embodiment also provides a computing resource processing device, as shown in Figure 7, the device includes: a maintenance module 701, which is used to maintain sleep state transfer prediction information corresponding to each of a plurality of physical computing resource objects, wherein the sleep state transfer prediction information includes probability information of the corresponding physical computing resource object entering each of the plurality of sleep states at the next sleep when it is awakened from any of the plurality of sleep states; a selection module 702, which is used to select a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the plurality of physical computing resource objects; and a scheduling module 703, which is used to schedule the target task to be scheduled to the target physical computing resource object.

In some optional embodiments, when maintaining sleep state transfer prediction information corresponding to each of a plurality of physical computing resource objects, the maintenance module 701 is specifically used to: when monitoring that any physical computing resource object is awakened from a first sleep state, obtain the actual sleep time of any physical computing resource object in the first sleep state, where the first sleep state is any sleep state; and update the sleep state transfer prediction information corresponding to any physical computing resource object according to the actual sleep time and the exit delay time of the first sleep state.

In some optional embodiments, when the maintenance module 701 monitors that any physical computing resource object is awakened from a first sleep state, it obtains the actual sleep time of any physical computing resource object in the first sleep state, and is specifically used to: obtain the first time when any physical computing resource object enters the first sleep state, and the second time when it is awakened from the first sleep state by the wake-up event; calculate the actual sleep time of any physical computing resource object in the first sleep state based on the second time and the first time.

In some optional embodiments, when the maintenance module 701 updates the sleep state transfer prediction information corresponding to any physical computing resource object according to the actual sleep time and the exit delay time of the first sleep state, it is specifically used to: when the actual sleep time matches the exit delay time of the first sleep state, relatively increase the probability information that any physical computing resource object will enter the first sleep state again when it is awakened from the first sleep state in the next sleep; when the actual sleep time does not match the exit delay time of the first sleep state, relatively increase the probability information that any physical computing resource object will enter the second sleep state when it is awakened from the first sleep state in the next sleep, the second sleep state being a sleep state whose exit delay time matches the actual sleep time.

In some optional embodiments, the maintenance module 701 is specifically used to: increase the probability information of any physical computing resource object entering the first sleep state again when it is awakened from the first sleep state when it is awakened from the first sleep state; state; and/or, reducing the probability information of any physical computing resource object entering other sleep states when it is awakened from the first sleep state at the next sleep state, where the other sleep states refer to sleep states other than the first sleep state.

In some optional embodiments, when the maintenance module 701 relatively increases the probability information that any physical computing resource object enters the second sleep state at the next sleep state when it is awakened from the first sleep state, it is specifically used to: increase the probability information that any physical computing resource object enters the second sleep state at the next sleep state when it is awakened from the first sleep state; and/or, reduce the probability information that any physical computing resource object enters the first sleep state at the next sleep state when it is awakened from the first sleep state; and/or, reduce the probability information that any physical computing resource object enters other sleep states at the next sleep state when it is awakened from the first sleep state, and other sleep states refer to sleep states other than the first sleep state and the second sleep state.

In some optional embodiments, the maintenance module 701 is also used to: during the initialization process, create a data structure corresponding to each of the multiple physical computing resource objects, the data structure being used to store sleep state transfer prediction information corresponding to the corresponding physical computing resource object; initialize the data structure corresponding to each of the multiple physical computing resource objects to obtain the initial value of the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects.

In some optional embodiments, when the maintenance module 701 creates a data structure corresponding to each of the multiple physical computing resource objects, it is specifically used to: create a sleep state weight array corresponding to each of the multiple physical computing resource objects, where the sleep state weight array is a multidimensional array including multiple rows and multiple columns; wherein a row represents a sleep state that the corresponding physical computing resource object may be in when it is awakened, and a column represents a sleep state that the corresponding physical computing resource object may enter when it sleeps next time; the array elements where rows and columns intersect represent the weight of the corresponding physical computing resource object entering the sleep state represented by the column when it is awakened from the sleep state represented by the row and when it sleeps next time, and the larger the weight, the higher the probability.

In some optional embodiments, when the maintenance module 701 initializes the data structures corresponding to multiple physical computing resource objects, it is specifically used to: initialize array elements with the same number of rows and columns to a first value, and initialize array elements with different number of rows and columns to a second value; wherein the first value is greater than the second value, and the sum of the first value and the second value in the same row is a third value.

In some optional embodiments, the selection module 702 is also used to: when it is monitored that any physical computing resource object needs to sleep, determine the third sleep state according to the sleep state transfer prediction information corresponding to any physical computing resource object; and control any physical computing resource object to enter the third sleep state.

In some optional embodiments, when the selection module 702 determines the third sleep state according to the sleep state transition prediction information corresponding to any physical computing resource object, it is specifically used to: obtain the sleep state most recently exited by any physical computing resource object as the fourth sleep state; obtain the probability information of entering each sleep state at the next sleep state when awakened from the fourth sleep state from the sleep state transition prediction information corresponding to any physical computing resource object; select the probability information according to the probability information of entering each sleep state at the next sleep state when awakened from the fourth sleep state The maximum sleep state is used as a candidate sleep state; when the exit delay time of the candidate sleep state is less than the maximum tolerable exit delay time of the system, the candidate sleep state is used as the third sleep state.

In some optional embodiments, the selection module 702 is further used to: when the exit delay time of the candidate sleep state is greater than or equal to the maximum tolerable exit delay time of the system, select another sleep state whose exit delay time is less than the maximum tolerable exit delay time of the system as the third sleep state.

In some optional embodiments, when the selection module 702 selects another sleep state in which the exit delay time is less than the maximum tolerable exit delay time of the system as the third sleep state, it is specifically used to: when there are multiple sleep states in which the exit delay time is less than the maximum tolerable exit delay time of the system, select the sleep state in which the exit delay time is less than the maximum tolerable exit delay time of the system and the probability information is the largest as the third sleep state.

In some optional embodiments, when the selection module 702 selects a target physical computing resource object based on the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, it is specifically used to: respond to a resource scheduling trigger event, generate sleep habit information corresponding to each of the multiple physical computing resource objects based on the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects; and select a target physical computing resource object from the multiple physical computing resource objects based on the sleep habit information corresponding to each of the multiple physical computing resource objects.

In some optional embodiments, when the selection module 702 generates sleep habit information corresponding to multiple physical computing resource objects based on sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, it is specifically used to: for any physical computing resource object, obtain the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep from the sleep state transfer prediction information corresponding to any physical computing resource object; generate the sleep habit information corresponding to any physical computing resource object based on the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep.

In some optional embodiments, when the selection module 702 generates the sleep habit information corresponding to any physical computing resource object based on the probability information of any physical computing resource object exiting from any sleep state and entering the sleep state again at the next sleep, it is specifically used to: if the probability information of any physical computing resource object exiting from the shallowest sleep state and entering the shallowest sleep state at the next sleep is the largest, determine that the sleep habit type corresponding to any physical computing resource is a shallow sleep habit, and generate the sleep probability of any physical computing resource object under the shallow sleep habit based on the largest probability information; if the probability information of any physical computing resource object exiting from the deepest sleep state and entering the deepest sleep state at the next sleep is the largest, determine that the sleep habit type corresponding to any physical computing resource object is a deep sleep habit, and generate the sleep probability of any physical computing resource under the deep sleep habit based on the largest probability information.

In some optional embodiments, when the selection module 702 selects a target physical computing resource object from a plurality of physical computing resource objects according to the sleep habit information corresponding to each of the plurality of physical computing resource objects, the selection module 702 is specifically configured to: determine at least one candidate corresponding to a shallow sleep habit according to the sleep habit information corresponding to each of the plurality of physical computing resource objects; According to the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit, select the candidate physical computing resource object whose sleep probability meets the requirements as the target physical computing resource object.

In some optional embodiments, when the selection module 702 selects a candidate physical computing resource object whose sleep probability meets the requirements as the target physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit, it is specifically used to: select the target physical computing resource object from at least one candidate physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit in combination with other auxiliary information; wherein the other auxiliary information includes at least one of the type of resource scheduling trigger event, the current operating status of the candidate physical computing resource object, and the current utilization rate of the candidate physical computing resource object.

In some optional embodiments, when the selection module 702 selects a target physical computing resource object from at least one candidate physical computing resource object based on the sleep probability of at least one candidate physical computing resource object in a shallow sleep habit and other auxiliary information, it is specifically used to: select a candidate physical computing resource object whose utilization meets the requirements and has the highest sleep probability as the target physical computing resource object based on the sleep probability of at least one candidate physical computing resource object in a shallow sleep habit and the current utilization rate.

In this embodiment, resource scheduling of physical computing resource objects is combined with sleep control of physical computing resource objects. By maintaining sleep state transfer prediction information corresponding to multiple physical computing resource objects, the physical computing resource objects are scheduled according to the maintained sleep state transfer prediction information. This can schedule tasks to physical computing resource objects with relatively shallow sleep states or relatively busy states as much as possible, which is beneficial for keeping some physical computing resource objects in a deep sleep state for as long as possible, keeping some physical computing resource objects in a busy state as much as possible, and reducing the frequency of physical computing resource objects entering and exiting the sleep state.

FIG8 is a schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present disclosure. As shown in FIG8 , the device includes: a memory 801 and a processor 802. The electronic device includes but is not limited to the physical machine in the aforementioned embodiment, and this embodiment does not make any limitation.

The memory 801 is used to store computer programs and can be configured to store various other data to support operations on the electronic device. Examples of such data include instructions for any application or method used to operate on the electronic device.

In some embodiments, the processor 802 is coupled to the memory 801 and is used to execute the computer program in the memory 801, so as to: maintain sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, wherein the sleep state transfer prediction information includes probability information of the corresponding physical computing resource object entering each of the multiple sleep states when it is awakened from any of the multiple sleep states at the next sleep; select a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects; and schedule the target task to be scheduled to the target physical computing resource object.

In some optional embodiments, the processor 802 maintains the sleep state corresponding to each of the plurality of physical computing resource objects. When the state transfer prediction information is obtained, it is specifically used to: when monitoring that any physical computing resource object is awakened from the first sleep state, obtain the actual sleep time of any physical computing resource object in the first sleep state, the first sleep state being any sleep state; according to the actual sleep time and the exit delay time of the first sleep state, update the sleep state transfer prediction information corresponding to any physical computing resource object.

In some optional embodiments, when the processor 802 monitors that any physical computing resource object is awakened from a first sleep state, it obtains the actual sleep time of any physical computing resource object in the first sleep state, and is specifically used to: obtain a first time when any physical computing resource object enters the first sleep state, and a second time when it is awakened from the first sleep state by a wake-up event; and calculate the actual sleep time of any physical computing resource object in the first sleep state based on the second time and the first time.

In some optional embodiments, when the processor 802 updates the sleep state transfer prediction information corresponding to any physical computing resource object based on the actual sleep time and the exit delay time of the first sleep state, it is specifically used to: when the actual sleep time matches the exit delay time of the first sleep state, relatively increase the probability information that any physical computing resource object will enter the first sleep state again when it is awakened from the first sleep state in the next sleep; when the actual sleep time does not match the exit delay time of the first sleep state, relatively increase the probability information that any physical computing resource object will enter the second sleep state when it is awakened from the first sleep state in the next sleep, the second sleep state being a sleep state whose exit delay time matches the actual sleep time.

In some optional embodiments, when the processor 802 relatively increases the probability information that any physical computing resource object will re-enter the first sleep state at the next sleep state when it is awakened from the first sleep state, it is specifically used to: increase the probability information that any physical computing resource object will re-enter the first sleep state at the next sleep state when it is awakened from the first sleep state; and/or reduce the probability information that any physical computing resource object will enter other sleep states at the next sleep state when it is awakened from the first sleep state, and other sleep states refer to sleep states other than the first sleep state.

In some optional embodiments, when the processor 802 relatively increases the probability information that any physical computing resource object enters the second sleep state at the next sleep state when it is awakened from the first sleep state, it is specifically used to: increase the probability information that any physical computing resource object enters the second sleep state at the next sleep state when it is awakened from the first sleep state; and/or, reduce the probability information that any physical computing resource object enters the first sleep state at the next sleep state when it is awakened from the first sleep state; and/or, reduce the probability information that any physical computing resource object enters other sleep states at the next sleep state when it is awakened from the first sleep state, and other sleep states refer to sleep states other than the first sleep state and the second sleep state.

In some optional embodiments, the processor 802 is further used to: during the initialization process, create a data structure corresponding to each of the plurality of physical computing resource objects, the data structure being used to store sleep state transition prediction information corresponding to the corresponding physical computing resource object; initialize the data structure corresponding to each of the plurality of physical computing resource objects to obtain the plurality of physical The initial value of the sleep state transition prediction information corresponding to each resource object is calculated.

In some optional embodiments, when the processor 802 creates a data structure corresponding to each of the multiple physical computing resource objects, it is specifically used to: create a sleep state weight array corresponding to each of the multiple physical computing resource objects, where the sleep state weight array is a multidimensional array including multiple rows and multiple columns; wherein a row represents a sleep state that the corresponding physical computing resource object may be in when it is awakened, and a column represents a sleep state that the corresponding physical computing resource object may enter when it sleeps next time; the array elements where rows and columns intersect represent the weight of the corresponding physical computing resource object entering the sleep state represented by the column when it is awakened from the sleep state represented by the row and when it sleeps next time, and the larger the weight, the higher the probability.

In some optional embodiments, when the processor 802 initializes the data structures corresponding to each of the multiple physical computing resource objects, it is specifically used to: initialize array elements with the same number of rows and columns to a first value, and initialize array elements with different number of rows and columns to a second value; wherein the first value is greater than the second value, and the sum of the first value and the second value in the same row is a third value.

In some optional embodiments, the processor 802 is further used to: when it is detected that any physical computing resource object needs to sleep, determine the third sleep state according to the sleep state transfer prediction information corresponding to any physical computing resource object; and control any physical computing resource object to enter the third sleep state.

In some optional embodiments, when the processor 802 determines the third sleep state based on the sleep state transfer prediction information corresponding to any physical computing resource object, it is specifically used to: obtain the sleep state most recently exited by any physical computing resource object as the fourth sleep state; obtain the probability information of entering each sleep state at the next sleep state when awakened from the fourth sleep state from the sleep state transfer prediction information corresponding to any physical computing resource object; select the sleep state with the largest probability information as the candidate sleep state based on the probability information of entering each sleep state at the next sleep state when awakened from the fourth sleep state; and select the candidate sleep state as the third sleep state when the exit delay time of the candidate sleep state is less than the maximum tolerable exit delay time of the system.

In some optional embodiments, the processor 802 is further used to: when the exit delay time of the candidate sleep state is greater than or equal to the maximum tolerable exit delay time of the system, select another sleep state whose exit delay time is less than the maximum tolerable exit delay time of the system as the third sleep state.

In some optional embodiments, when the processor 802 selects another sleep state in which the exit delay time is less than the maximum tolerable exit delay time of the system as the third sleep state, it is specifically used to: when there are multiple sleep states in which the exit delay time is less than the maximum tolerable exit delay time of the system, select the sleep state in which the exit delay time is less than the maximum tolerable exit delay time of the system and the probability information is the largest as the third sleep state.

In some optional embodiments, when the processor 802 selects a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, it is specifically used to: respond to a resource scheduling trigger event, generate a plurality of physical computing resource objects according to the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects; The target physical computing resource object is selected from the multiple physical computing resource objects according to the sleep habit information corresponding to each of the multiple physical computing resource objects.

In some optional embodiments, when the processor 802 generates sleep habit information corresponding to multiple physical computing resource objects based on sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects, it is specifically used to: for any physical computing resource object, obtain probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state from the sleep state transfer prediction information corresponding to any physical computing resource object; generate sleep habit information corresponding to any physical computing resource object based on the probability information of any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state.

In some optional embodiments, when the processor 802 generates the sleep habit information corresponding to any physical computing resource object based on the probability information of any physical computing resource object exiting from any sleep state and entering the sleep state again at the next sleep, it is specifically used to: if the probability information of any physical computing resource object exiting from the shallowest sleep state and entering the shallowest sleep state at the next sleep is the largest, determine that the sleep habit type corresponding to any physical computing resource is a shallow sleep habit, and generate the sleep probability of any physical computing resource object under the shallow sleep habit based on the largest probability information; if the probability information of any physical computing resource object exiting from the deepest sleep state and entering the deepest sleep state at the next sleep is the largest, determine that the sleep habit type corresponding to any physical computing resource object is a deep sleep habit, and generate the sleep probability of any physical computing resource under the deep sleep habit based on the largest probability information.

In some optional embodiments, when the processor 802 selects a target physical computing resource object from a plurality of physical computing resource objects based on the sleep habit information corresponding to each of the plurality of physical computing resource objects, it is specifically used to: determine at least one candidate physical computing resource object corresponding to a shallow sleep habit based on the sleep habit information corresponding to each of the plurality of physical computing resource objects; and select a candidate physical computing resource object whose sleep probability meets the requirements as the target physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit.

In some optional embodiments, when the processor 802 selects a candidate physical computing resource object whose sleep probability meets the requirements as a target physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit, it is specifically used to: select the target physical computing resource object from at least one candidate physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit in combination with other auxiliary information; wherein the other auxiliary information includes at least one of the type of resource scheduling trigger event, the current operating status of the candidate physical computing resource object, and the current utilization rate of the candidate physical computing resource object.

In some optional embodiments, the processor 802 selects a target physical computing resource object from at least one candidate physical computing resource object based on the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit and in combination with other auxiliary information, specifically for: selecting a candidate physical computing resource object whose utilization rate meets the requirements and whose sleep probability is the largest according to the sleep probability of at least one candidate physical computing resource object under the shallow sleep habit and the current utilization rate; As the target physical compute resource object.

Further, as shown in Figure 8, the electronic device also includes: other components such as a communication component 803, a display 804 and a power supply component 805. Figure 8 only schematically shows some components, which does not mean that the electronic device only includes the components shown in Figure 8. In addition, the components in the dotted box in Figure 8 are optional components, not mandatory components, and may depend on the product form of the electronic device. The electronic device of this embodiment can be implemented as a terminal device such as a desktop computer, a laptop computer, a tablet computer, a smart phone or an IOT device, or it can be a server-side device such as a conventional server, a cloud server or a server array. If the electronic device of this embodiment is implemented as a terminal device such as a desktop computer, a laptop computer, a tablet computer, a smart phone, etc., it may include the components in the dotted box in Figure 8; if the electronic device of this embodiment is implemented as a server-side device such as a conventional server, a cloud server or a server array, it may not include the components in the dotted box in Figure 8.

Accordingly, an embodiment of the present disclosure further provides a computer-readable storage medium storing a computer program, and when the computer program is executed, each step that can be executed by an electronic device in the method embodiment shown in FIG. 5 above can be implemented.

Correspondingly, an embodiment of the present disclosure also provides a multi-core processor system, including multiple processor cores and a memory, wherein the memory is used to store the program code of the operating system, and the multiple processor cores are used to run the program code of the operating system to implement the steps in the method embodiment shown in Figure 5 above.

In this embodiment, resource scheduling of physical computing resource objects is combined with sleep control of physical computing resource objects. By maintaining sleep state transfer prediction information corresponding to multiple physical computing resource objects, the physical computing resource objects are scheduled according to the maintained sleep state transfer prediction information. This can schedule tasks to physical computing resource objects with relatively shallow sleep states or relatively busy states as much as possible, which is beneficial for keeping some physical computing resource objects in a deep sleep state for as long as possible, keeping some physical computing resource objects in a busy state as much as possible, and reducing the frequency of physical computing resource objects entering and exiting the sleep state.

The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination of them, such as static random-access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, disk or optical disk.

The above-mentioned communication component is configured to facilitate wired or wireless communication between the device where the communication component is located and other devices. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G/LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component also includes a near field communication (NFC) module to facilitate short-range communication. For example, in NFC The module can be implemented based on Radio Frequency Identification (RFID) technology, Infrared Data Association (IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.

The above-mentioned display includes a screen, and the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.

The power supply assembly provides power to various components of the device where the power supply assembly is located. The power supply assembly may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply assembly is located.

The above-mentioned audio component can be configured to output and/or input audio signals. For example, the audio component includes a microphone (Microphone, MIC), and when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in a memory or sent via a communication component. In some embodiments, the audio component also includes a speaker for outputting an audio signal.

It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, compact disc read-only memory (CD-ROM), optical storage, etc.) containing computer-usable program code.

The present disclosure is described with reference to the flowchart and/or block diagram of the method, device (system), and computer program product according to the embodiment of the present disclosure. It should be understood that each process and/or box in the flowchart and/or block diagram, as well as the combination of the process and/or box in the flowchart and/or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a product including an instruction device, and the instruction device is implemented in one or more processes of the flowchart and/or one or more blocks of the block diagram. The function specified in the box.

These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), input/output interfaces, network interfaces, and memory.

Memory may include non-permanent storage in a computer-readable medium, in the form of random access memory (RAM) and/or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, Phase-change Random Access Memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read-Only Memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Flash memory or other memory technology, Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer readable media does not include transitory media such as modulated data signals and carrier waves.

It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

The above are only embodiments of the present disclosure and are not intended to limit the present disclosure. For those skilled in the art, the present disclosure may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the scope of the claims of the present disclosure.

Claims (16)

A computing resource processing method, comprising: Maintaining sleep state transition prediction information corresponding to each of the plurality of physical computing resource objects, wherein the sleep state transition prediction information includes probability information of the corresponding physical computing resource object entering each of the plurality of sleep states when the corresponding physical computing resource object is awakened from any of the plurality of sleep states during the next sleep state; Selecting a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the plurality of physical computing resource objects; The target task to be scheduled is scheduled onto the target physical computing resource object. The method according to claim 1, wherein maintaining sleep state transition prediction information corresponding to each of the plurality of physical computing resource objects comprises: When any physical computing resource object is monitored to be awakened from a first sleep state, obtaining an actual sleep time of the any physical computing resource object in the first sleep state, where the first sleep state is any sleep state; The sleep state transition prediction information corresponding to any one of the physical computing resource objects is updated according to the actual sleep time and the exit delay time of the first sleep state. The method according to claim 2, wherein updating the sleep state transition prediction information corresponding to any one of the physical computing resource objects according to the actual sleep time and the exit delay time of the first sleep state comprises: In a case where the actual sleep time matches the exit delay time of the first sleep state, relatively increasing the probability information of any physical computing resource object re-entering the first sleep state in the next sleep state when being awakened from the first sleep state; When the actual sleep time does not match the exit delay time of the first sleep state, the probability information of any physical computing resource object entering a second sleep state at the next sleep when awakened from the first sleep state is relatively increased, and the second sleep state is a sleep state whose exit delay time matches the actual sleep time. The method according to any one of claims 1 to 3, further comprising: During the initialization process, a data structure corresponding to each of the plurality of physical computing resource objects is created, wherein the data structure is used to store sleep state transfer prediction information corresponding to the corresponding physical computing resource object; Initialize the data structures corresponding to the plurality of physical computing resource objects to obtain initial values of the sleep state transition prediction information corresponding to the plurality of physical computing resource objects. The method according to claim 4, wherein creating a data structure corresponding to each of the plurality of physical computing resource objects comprises: Creating a sleep state weight array corresponding to each of the plurality of physical computing resource objects, wherein the sleep state weight array is a multidimensional array including a plurality of rows and a plurality of columns; Wherein, a row represents a possible sleep state of the corresponding physical computing resource object when it is awakened, and a column represents a possible sleep state of the corresponding physical computing resource object when it sleeps next time; The array elements where rows and columns intersect represent the weight of the corresponding physical computing resource object entering the sleep state represented by the column when it is awakened from the sleep state represented by the row and then sleeps for the next time. The greater the weight, the higher the probability. The method according to claim 5, wherein initializing the data structures corresponding to the plurality of physical computing resource objects respectively comprises: Initialize array elements with the same number of rows and columns to a first value, and initialize array elements with different number of rows and columns to a second value; The first value is greater than the second value, and the sum of the first value and the second value in the same row is the third value. The method according to any one of claims 1 to 6, further comprising: When it is detected that any physical computing resource object needs to sleep, determining a third sleep state according to sleep state transition prediction information corresponding to the any physical computing resource object; Control any one of the physical computing resource objects to enter the third sleep state. The method according to claim 7, wherein determining the third sleep state according to the sleep state transition prediction information corresponding to any one of the physical computing resource objects comprises: Obtaining a sleep state most recently exited by any of the physical computing resource objects as a fourth sleep state; acquiring, from the sleep state transition prediction information corresponding to any one of the physical computing resource objects, probability information of entering each sleep state in the next sleep state when awakened from the fourth sleep state; According to the probability information of entering each sleep state in the next sleep state when awakened from the fourth sleep state, selecting the sleep state with the largest probability information as the candidate sleep state; In a case where the exit delay time of the candidate sleep state is less than the maximum tolerable exit delay time of the system, the candidate sleep state is used as the third sleep state. The method according to any one of claims 1 to 8, wherein selecting a target physical computing resource object according to sleep state transfer prediction information corresponding to each of the plurality of physical computing resource objects comprises: In response to a resource scheduling trigger event, generating sleep habit information corresponding to each of the plurality of physical computing resource objects according to sleep state transition prediction information corresponding to each of the plurality of physical computing resource objects; A target physical computing resource object is selected from the multiple physical computing resource objects according to the sleep habit information corresponding to each of the multiple physical computing resource objects. The method according to claim 9, wherein the plurality of physical computing resource objects respectively correspond to The sleep state transition prediction information is used to generate sleep habit information corresponding to each of the plurality of physical computing resource objects, including: For any physical computing resource object, obtaining, from the sleep state transition prediction information corresponding to the any physical computing resource object, probability information of the any physical computing resource object exiting from any sleep state and re-entering the sleep state at the next sleep state; The sleep habit information corresponding to the any physical computing resource object is generated according to the probability information that the any physical computing resource object exits from any sleep state and re-enters the sleep state at the next sleep state. A physical machine, wherein an operating system is run on hardware resources of the physical machine, the hardware resources include a plurality of physical computing resource objects, the plurality of physical computing resource objects support a plurality of sleep states with different sleep depths, and the operating system includes: a target resource scheduler and a target sleep controller; The target sleep controller is used to maintain sleep state transition prediction information corresponding to each of the plurality of physical computing resource objects, wherein the sleep state transition prediction information includes probability information of the corresponding physical computing resource object entering each sleep state when it is awakened from any sleep state at the next sleep state; The target resource scheduler is used to select a target physical computing resource object according to the sleep state transfer prediction information corresponding to each of the multiple physical computing resource objects; and schedule the target task to be scheduled to the target physical computing resource object. According to the physical machine according to claim 11, wherein the operating system further includes other resource schedulers and other sleep controllers, different sleep controllers have different sleep control logics, and different resource schedulers have different resource scheduling logics. The physical machine according to claim 12, wherein when any resource scheduler among the other resource schedulers and any sleep controller among the other sleep controllers are configured to be effective, the target resource scheduler and the target sleep controller are configured to be invalid; Among them, any resource scheduler among the other resource schedulers and any sleep controller among the other sleep controllers are independent of each other; and the target resource scheduler and the target sleep controller are associated with each other. A multi-core processor system, comprising a plurality of processor cores and a memory, wherein the memory is used to store program code of an operating system, and the plurality of processor cores are used to run the program code of the operating system, so as to implement the steps in the method described in any one of claims 1 to 10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is enabled to implement the steps in the method according to any one of claims 1 to 10. A computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps in the method according to any one of claims 1 to 10 are implemented.