CN108549574A - Threading scheduling management method, device, computer equipment and storage medium - Google Patents

Abstract

This application involves a kind of threading scheduling management method, device, computer equipment and storage mediums.The method includes：Selection, which is moved out, processor and moves into processor, and judgement, which is moved out, processor and moves into whether processor meets thread migration condition；Wherein, the processor of moving out includes one or more thread queues；If so, searching the thread run in the thread queue moved out on processor, the corresponding residue load capacity to be migrated of the thread found is calculated；According to remaining load capacity selection target thread to be migrated, the subject thread is dispatched to and moves into processor.The threads load amount of scheduling can be avoided excessive using this method, to the problem of scheduling occurred.

Description

Thread scheduling management method, device, computer equipment and storage medium

technical field

The present application relates to the field of computer technology, in particular to a thread scheduling management method, device, computer equipment and storage medium.

Background technique

With the development of computer technology, computer equipment can dynamically schedule threads on processors during operation, so that threads can achieve load balance on each processor and improve the execution efficiency of threads. For the threads on the processor, there is a difference in priority. In the existing thread scheduling method, the thread is only selected according to the priority of the thread for scheduling. Since different threads have different loads, it may be possible to use this thread scheduling method. As a result, the load is unbalanced after the thread is scheduled, that is, the problem of over-scheduling occurs. Therefore, how to solve the problem of overscheduling in the thread scheduling process has become a technical problem to be solved at present.

Contents of the invention

Based on this, it is necessary to provide a thread scheduling management method, device, computer equipment and storage medium capable of solving the problem of overscheduling in the thread scheduling process, aiming at the above technical problems.

A thread scheduling management method, the method comprising:

Selecting the outgoing processor and the incoming processor, and judging whether the outgoing processor and the incoming processor meet the thread migration condition; wherein, the outgoing processor includes one or more thread queues;

If yes, find the threads running in the thread queue on the outgoing processor, and calculate the remaining load to be migrated corresponding to the found threads;

A target thread is selected according to the remaining load to be migrated, and the target thread is scheduled to the transfer-in processor.

In one of the embodiments, the step of searching for the thread queue running on the outgoing processor includes: acquiring information about the processor core group where the outgoing processor is located and the processor core group where the outgoing processor is located information; when the processor that moves out and the processor that moves in are in the same processor core group, obtain the number of threads in the thread queue with the highest priority on the processor that moves out; if the number of threads is more If the number of threads is one, the thread queue is searched in order of priority from high to low; if the number of threads is one, the thread queue is searched in order of priority from low to high.

In one of the embodiments, the method further includes: when the processor core group where the outgoing processor is located is a small core group, and the processor core group where the incoming processor is located is a large core group, perform the following steps : If there is no thread executing the scheduled task running on the moving-in processor, search for the thread queue of the thread executing the scheduled task in the order of load from high to low; if there is no thread executing the scheduled task running on the moving-out processor and move in If there are threads executing scheduled tasks running on the processor, the search ends.

In one of the embodiments, the method further includes: when the processor core group where the outgoing processor is located is a large core group, and the processor core group where the incoming processor is located is a small core group, perform the following steps : If the total load of the threads executing the scheduled tasks on the outgoing processor is the largest in the large core group, search for the thread queues for executing the scheduled tasks according to the thread load in ascending order; otherwise, end the search for the thread queues for executing the scheduled tasks thread queue.

In one of the embodiments, the method further includes: acquiring storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the input and output device within a preset time; If the number of accesses to the input and output device is greater than a first threshold, record the running thread as a first type thread.

In one of the embodiments, the method further includes: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the memory within a preset time; if the If the number of memory access times is greater than a second threshold, the running thread is recorded as a second type thread.

A thread scheduling management device, the device comprising:

The selection module is used to select the outgoing processor and the incoming processor, and judge whether the outgoing processor and the incoming processor meet the thread migration condition; wherein, the outgoing processor includes one or more thread queues;

A search module, configured to search for threads running in the thread queue on the outgoing processor if the migration conditions are met, and calculate the remaining load to be migrated corresponding to the found threads;

A scheduling module, configured to select a target thread according to the remaining load to be migrated, and schedule the target thread to the transfer-in processor.

A computer device, comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, when the processor executes the computer program, the following steps are implemented: selecting a moving-out processor and a moving-in processor , judging whether the moving-out processor and the moving-in processor meet the thread migration condition; wherein, the moving-out processor includes one or more thread queues; if so, searching for threads running in the thread queue on the moving-out processor, Calculate the remaining load to be migrated corresponding to the found thread; select a target thread according to the remaining load to be migrated, and schedule the target thread to the migration-in processor.

A computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented: selecting a moving-out processor and a moving-in processor, and judging whether the moving-out processor and the moving-in processor are Satisfy the thread migration condition; wherein, the outgoing processor includes one or more thread queues; if so, search for threads running in the thread queues on the outgoing processor, and calculate the remaining load to be migrated corresponding to the found thread ; Select a target thread according to the remaining load to be migrated, and schedule the target thread to the transfer-in processor.

In the above thread scheduling management method, device, computer equipment and storage medium, the terminal calculates the remaining load to be migrated of the thread, selects the target thread according to the remaining load to be migrated, and schedules the target thread to the incoming processor, so that the load can be selected The most suitable thread is scheduled, avoiding the excessive load of the scheduled thread, resulting in the problem of over-scheduling, and improving the reliability of thread scheduling.

Description of drawings

Fig. 1 is a schematic flow diagram of a thread scheduling management method in an embodiment;

Fig. 2 is a thread state diagram in an embodiment;

Fig. 3 is a schematic flow chart of the steps of searching the thread queue running on the outgoing processor in one embodiment;

Figure 4a is a schematic diagram of the state of the running thread of an embodiment;

Fig. 4b is a schematic diagram of the thread migration state of Fig. 4a;

Fig. 5 is a structural block diagram of a thread scheduling management device in an embodiment;

Figure 6 is an internal block diagram of a computer device in one embodiment.

Detailed ways

In order to make the purpose, technical solution and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described here are only used to explain the present application, and are not intended to limit the present application.

In one embodiment, as shown in FIG. 1 , a thread scheduling management method is provided, and the method is applied to a terminal as an example for illustration, including the following steps:

Step 102, select the outgoing processor and the incoming processor, and judge whether the outgoing processor and the incoming processor meet the thread migration condition; wherein, the outgoing processor includes one or more thread queues.

The processor of the terminal may be at least one of a central processing unit (Central Processing Unit, CPU), a graphics processing unit (Graphics Processing Unit, GPU), a video processing unit (Video Processing Unit, VPU), and the like. The processor of the terminal may be a single-core architecture or a multi-core architecture. When the multi-core architecture is used, one core in the multi-core architecture is regarded as one processor. In this embodiment, the processor of the terminal is taken as an example with a multi-core architecture, wherein the transfer-in processor and the transfer-out processor are each a core.

At any moment during the operation of the terminal, the processing threads on each processor may be in an unbalanced load. When the load is unbalanced, there will be too many threads in the ready queue waiting for processor time on some processors. Other processors run too few threads, resulting in inefficient overall thread processing. In order to improve thread processing efficiency, it is necessary to schedule threads among processors to achieve load balance among processors.

The terminal selects the migration-out processor and the migration-in processor. The terminal may randomly select two processors as the transfer-out processor and the transfer-in processor, respectively, and the terminal may also sequentially select two processors as the transfer-out processor and the transfer-in processor according to the address information of the processors. Further, the terminal judges whether the moving-out processor and the moving-in processor meet the thread migration condition. Specifically, the terminal obtains the remaining loadable capacity of the outgoing processor and the remaining loadable capacity of the incoming processor, and calculates the difference between the remaining loadable amount of the incoming processor and the remaining loadable amount of the outgoing processor. When the difference is greater than the preset value, it is determined that the thread migration condition is met. Among them, the processor has a fixed total load, the total load is the load of the maximum thread that the processor can handle, and the remaining load of the processor is the total load of the processor and the load of the threads currently running on the processor difference.

Egress processors include one or more thread queues. Multiple threads in the same thread queue have the same priority, and threads in different thread queues have different priorities. In this embodiment, a thread executing a predetermined task is recorded as a first thread, and a thread awakened by the first thread is recorded as a second thread. The terminal assigns the first priority to the first thread, and assigns the second priority to the second thread. Wherein, the first priority may be the highest priority, and the second priority may be the highest priority, or the second highest priority, or the lowest priority. In this embodiment, the first priority is the highest priority, and the second priority is the second highest priority. Further, the terminal assigns the lowest priority to threads performing other tasks. For example, the terminal assigns the priority of the thread executing the scheduled task as 1, assigns the priority of the thread awakened by the thread executing the scheduled task as 2, and assigns the priority of the thread executing other tasks as 3, wherein, 1, 2, 3 is a descending priority.

Scheduled tasks include tasks such as drawing pictures, playing sounds, and responding to feedback. The scheduled task is identified and determined in advance by judging whether there is a scheduled flag in the processor of the thread running on the terminal processor. Wherein, a scheduled task has a corresponding scheduled task identifier, and the scheduled task identifier is pre-recorded in the file. The file recording the scheduled task ID may be called a thread ID file, and the thread ID file may be in the form of a database table. The thread identification file is stored in the storage resources of the terminal, and the storage resources include but not limited to registers, caches, internal memory, and external storage devices.

Threads include multiple states, new state, ready state, running state, blocked state, and dead state. As shown in Figure 2, a thread is always in one of the states from creation, execution to termination. During the running of the thread, it may enter the blocked state due to various reasons. If the thread in the blocked state has not finished running, it will temporarily give up the processor and enter the waiting queue. A thread in a blocked state does not start running automatically, and can be resumed by waking it up. During the running process of the thread, it may also enter the death state due to the normal exit of the execution end or the execution termination due to an exception. A thread that executes a scheduled task can wake up multiple threads while it is running. The thread identification awakened by the thread executing the scheduled task is also pre-recorded in the thread identification file. Specifically, the identifier of the awakened thread is recorded in the thread identifier file corresponding to the identifier of the thread that wakes it up and executes the predetermined task.

Step 104, if yes, search for the thread queue running on the outgoing processor, and calculate the remaining load to be migrated corresponding to the searched thread.

If the moving-out processor and the moving-in processor meet the thread migration condition, the terminal searches for the thread queue running on the moving-out processor. Wherein, the thread migration condition includes that the current load to be migrated is greater than a preset value, and the current load to be migrated is the difference between the remaining loadable capacity of the incoming processor and the remaining loadable amount of the outgoing processor. In this embodiment, the terminal first searches for the thread queue with the highest priority. Further, each time the terminal finds a thread, it calculates the remaining load to be migrated corresponding to the found thread. Wherein, the remaining load to be migrated is the difference between the remaining loadable capacity of the incoming processor and the remaining loadable amount of the outgoing processor after the thread is scheduled to the incoming processor. For example, if the remaining loadable capacity of the currently migrating processor is m, the remaining loadable capacity of the migrating processor is n, and the load capacity of the thread found by the terminal is p, then the remaining load to be migrated corresponding to the thread is (m-p)-(n+p)=m-n-2p.

Step 106, selecting a target thread according to the remaining load to be migrated, and scheduling the target thread to the migration-in processor.

When the terminal searches the thread queue, it judges whether to select the found thread as the target thread. Specifically, the terminal judges whether to select the thread as the target thread according to the thread load and the corresponding remaining load to be migrated. More specifically, when the load of the thread is less than twice the remaining load to be migrated, the terminal selects the thread as the target thread. Further, the scheduler of the terminal schedules the target thread to the transfer-in processor.

In this embodiment, the terminal calculates the remaining load to be migrated of the thread, selects the target thread according to the remaining load to be migrated, and schedules the target thread to the migration processor, so that the thread with the most appropriate load can be selected for scheduling, avoiding the The scheduled thread load is too large, resulting in the problem of over-scheduling, which improves the reliability of thread scheduling.

In one of the embodiments, the thread scheduling management method also includes: judging whether the thread corresponding to the currently searched thread queue has reached load balance on the outgoing processor and the incoming processor; if so, searching for the next thread queue; if If not, continue to search the current thread queue.

Since different threads have different requirements on processors, for example, threads executing scheduled tasks need to be executed efficiently to ensure the execution efficiency of scheduled tasks, and thus it is necessary to ensure that threads executing scheduled tasks always obtain better processors. The processor time occupied by different threads is also different. For example, the processor time occupied by the A thread with the execution priority of A is a, and the processor time occupied by the B thread with the execution priority of B is b. If three A threads The load is the same as the load of one A thread and four B threads. Under the premise of load balancing, three A threads are allocated to run on one processor, and one A thread and four B threads are allocated to another processor. Running, for a processor running three A threads, the time for one A thread to wait for the processor is the time for running two A threads, that is, the waiting time is 2a, while for a processor running one A thread and four B threads, A The time a thread waits for a processor is the time it takes to run four B threads, that is, the waiting time is 4b. Since the time to run two A threads may be different from the time to run four B threads, then for A threads on two processors , the waiting time for the processor is different, the execution efficiency of the A thread is also different, that is, the execution efficiency of the A thread is not balanced on each processor, so it is necessary to balance the scheduling of threads with different priorities on each processor , so that the threads of each priority are balanced on each processor.

The terminal judges whether the thread corresponding to the currently searched thread queue achieves load balance on the outgoing processor and the incoming processor. Specifically, the terminal obtains the difference between the number of threads in the thread queue on the outgoing processor and the number of threads in the thread queue on the incoming processor, and when the absolute value of the difference is less than or equal to the preset threshold, then Judging that the thread corresponding to the currently searched thread queue has reached load balance on the outgoing processor and the incoming processor, when the absolute value of the difference is greater than the preset threshold, then judge the thread corresponding to the currently searched thread queue Load balancing is not achieved on the outgoing and incoming processors.

Further, if the threads corresponding to the currently searched thread queue do not achieve load balance between the outgoing processor and the incoming processor, the terminal continues to search for the current thread queue. Specifically, the terminal sequentially searches the current thread queue according to the order in which the threads are stored in the priority queue, and performs step 104 on the found threads. If the threads corresponding to the currently searched thread queue have reached load balance between the outgoing processor and the incoming processor, the terminal searches for the next thread queue. Specifically, the terminal searches for the thread queue corresponding to the next priority, and performs step 104 .

Further, the terminal judges whether the outgoing processor and the incoming processor have reached load balance, and if so, ends the thread search, that is, ends the thread scheduling, otherwise, continues to search the thread queue. Specifically, after the terminal schedules the target thread to the transfer-in processor, it calculates the current load to be migrated, and ends the migration if the current load to be migrated is less than a preset value.

In this embodiment, if the threads corresponding to the currently searched thread queue reach load balance between the outgoing processor and the incoming processor, the terminal will search for the next thread queue for thread scheduling, otherwise continue to search for the current thread queue for thread scheduling , so that the threads of different priorities can achieve load balancing on the outgoing processor and the incoming processor, thereby effectively improving the execution efficiency of the thread and the utilization rate of the processor.

In one embodiment, the step of judging whether the moving-out processor and the moving-in processor meet the thread migration condition includes: when the moving-out processor and the moving-in processor meet one of the following conditions: 1) running on the moving-in processor Threads that execute scheduled tasks; 2) The optional threads to be migrated on the outgoing processor are threads that execute scheduled tasks, and the incoming processor is not in an idle state; then it is judged that the thread migration condition is not met; otherwise, it is judged that the thread is satisfied Migration conditions.

In this embodiment, the thread migration condition is further limited according to the thread executing the scheduled task, so that the thread executing the scheduled task obtains the optimal processor on the premise of the least migration, so as to improve the execution efficiency of the thread executing the scheduled task. Specifically, the terminal judges whether the outgoing processor and the incoming processor meet one of the following two conditions:

1) There are threads executing scheduled tasks running on the incoming processor. Further, there are scheduled task threads running on the incoming processor and the number of threads executing scheduled tasks among all processors of the terminal is not the minimum. In order to ensure that the thread executing the scheduled task can be executed efficiently, it is necessary to reduce the time it waits for the processor. When the number of threads on the processor of the thread executing the scheduled task increases, the waiting time of the thread executing the scheduled task will increase. , therefore, the number of threads executing scheduled tasks running on the incoming processor is not the least among all processors of the terminal as the condition for not satisfying the thread migration.

2) There is only one optional thread to be migrated on the outgoing processor, which is a thread that executes a predetermined task, and the incoming processor is not in an idle state. Among them, the optional thread to be migrated is the thread on the ready queue of the processor. If there is only one thread on the ready queue and it is a thread executing a predetermined task, and the migrating processor is not in an idle state, that is If the thread executing the scheduled task is scheduled to the transfer-in processor, it cannot be executed immediately. The minimum time for the thread to wait is the time it takes for the transfer-in processor to finish executing the thread that is occupying the processor time. If the thread is not scheduled Scheduling to the moving-in processor only needs to wait for the thread currently occupying the moving-out processor to be executed, while the scheduler of the terminal needs to occupy the resources of the terminal and a certain amount of processing time when scheduling the thread, which affects the thread execution efficiency. Therefore, in this case, thread scheduling may not be performed, that is, condition 2) may be regarded as a condition for not satisfying thread migration.

The terminal judges whether the moving-out processor and the moving-in processor meet the condition 1), and the specific steps are: the terminal searches for the moving-in processor, obtains the identifier of the thread running on the moving-in processor, searches for the thread identification file according to the thread identifier, and determines Whether the obtained thread identification is recorded in the thread identification file, if so, then move into the processor and run the thread that executes the scheduled task, that is, satisfy condition 1), if not, then not satisfied.

The terminal judges whether the moving-out processor and the moving-in processor meet the condition 2), and the specific steps are: the terminal searches for the moving-out processor, obtains the thread ID on the ready queue of the moving-out processor, and searches for the thread ID file according to the thread ID , judging whether the acquired thread identifiers are recorded in the thread identifier file, and further, the terminal obtains the number of thread identifiers recorded in the thread identifier file on the ready queue of the outgoing processor. The terminal searches for threads on the incoming processor, and obtains the number of threads on the incoming processor. If the number of thread identifiers recorded in the thread identifier file on the ready queue of the outgoing processor is one, and the number of threads on the incoming processor is one or more, it is judged that condition 2) is satisfied, otherwise it is not satisfied .

If the moving-out processor and the moving-in processor meet condition 1) or condition 2), the terminal judges that the moving-out processor and the moving-in processor do not satisfy the thread migration condition, and the terminal ends the thread migration. If both condition 1) and condition 2) are not met, the terminal judges that the outgoing processor and the incoming processor meet the thread migration condition, and the terminal executes step 102 .

In this embodiment, when there is a thread executing a predetermined task running on the moving-in processor, or there is only one optional thread to be migrated on the outgoing processor and it is a thread performing a predetermined task, the moving-in processor is not in an idle state When , the migration of the thread is ended, thereby reducing the processor contention of the thread executing the scheduled task, so that the thread executing the scheduled task can be executed efficiently.

In one embodiment, as shown in FIG. 3 , the step of finding the thread queue running on the outgoing processor includes:

In step 302, information about a processor core group where the outgoing processor is located and information about a processor core group where the incoming processor is located are acquired.

For a multi-processor terminal, the processor may have a homogeneous co-structure architecture or a heterogeneous co-structure architecture. Among them, homogeneous co-construction means that multiple processors are in the same processor core group, and multiple processors in the same processor core group have the same structure and performance, that is, have the same clock frequency, that is, multiple processors Threads are equally efficient. Heterogeneous co-construction means that multiple processors are located in multiple different processor core groups. The processors in different processor core groups have different structures and performances, and their clock frequencies are also different, that is, the efficiency of processing threads is different. Therefore, it is necessary to schedule threads according to the processor core group where the processor is located, so that threads with higher priority are preferentially scheduled to processors with higher processing efficiency.

In this embodiment, the terminal obtains the processor core group information where the outgoing processor is located and the processor core group information where the incoming processor is located. Specifically, the terminal acquires the information of the outgoing processor, analyzes the information of the outgoing processor, and obtains the information of the processor core group where the outgoing processor is located. Similarly, the terminal acquires information about the processor core group where the incoming processor is located. Further, the terminal judges the processor core group where the outgoing processor is located and the processor core group where the incoming processor is located according to the information about the processor core group where the outgoing processor is located and the information about the processor core group where the incoming processor is located. Is it the same.

Step 304, when the moving-out processor and the moving-in processor are in the same processor core group, obtain the number of threads in the thread queue with the highest priority on the moving-out processor; Search the thread queue in order of priority from high to low. If the number of threads is one, search the thread queue in order of priority from low to high.

When the moving-out processor and the moving-in processor are in the same processor core group. Further, the terminal acquires the number of threads in the thread queue with the highest priority on the transfer-out processor, and when there are multiple threads, the terminal searches the thread queue in descending order of priority. Specifically, the terminal selects the thread queue with the highest priority, and performs step 104 . If the number of threads is one, the terminal searches the thread queue in order of priority from low to high, and executes step 104 .

In this embodiment, when there are multiple threads with the highest priority on the outgoing processor, the threads in the thread queue with the highest priority are searched and further scheduled to ensure that the thread with the highest priority can be transferred to the outgoing processor. To achieve load balancing on the incoming processor and to ensure its execution efficiency, and when there is only one thread with the highest priority on the outgoing processor, it will be searched from the thread queue with the lowest priority first, and then scheduled, which can ensure that the thread with the highest priority The thread with the highest level is not scheduled under the premise of load balancing between the outgoing processor and the incoming processor, so that it can be reserved for execution on the outgoing processor, thereby avoiding the reduction in thread execution efficiency caused by the scheduling process, making The thread with the highest priority always has the highest execution efficiency.

In one embodiment, the step of searching for the thread queue running on the outgoing processor includes: step 306, when the processor core group where the outgoing processor is located is a small core group, the process where the incoming processor is located When the processor core group is a large core group, perform the following steps: If there is no thread running on the scheduled task on the moving-in processor, search for the thread queue of the thread performing the scheduled task in the order of load from high to low; If there is no thread executing the scheduled task running on the processor and there is a thread executing the scheduled task running on the transfer-in processor, the search ends.

Among them, for a terminal with a heterogeneous co-architecture processor, the processor in the terminal is generally divided into a large core and a small core. Both the large core and the small core have their own fixed logical structures, including caches, execution units, and instruction-level units. and logic units such as bus interfaces. A nuclear group composed of multiple large nuclei is called a large nuclear group, and a nuclear group composed of small nuclei is called a small nuclear group. The logic unit of the large core has better performance than the small core. For example, the clock frequency of the large core is higher than that of the small core, that is, the thread execution efficiency of the large core group is higher than that of the small core group. To ensure the priority Higher threads have higher processing efficiency, and it is necessary to consider the processing efficiency of small core groups and large core groups to schedule threads.

In this embodiment, when the processor core group where the outgoing processor is located is a small core group, and the processor core group where the incoming processor is located is a large core group, that is, the thread processing efficiency of the outgoing processor is lower than that of the incoming processor. When processing thread processing efficiency of the processor, further, the terminal judges whether there is a thread executing a predetermined task on the incoming processor. Specifically, the terminal acquires the threads on the incoming processor, searches for the thread identification file, determines whether the identification of the thread on the incoming processor is recorded in the thread identification file, and obtains the number of threads executing scheduled tasks on the incoming processor. . Further, if there is no thread executing the scheduled task on the moving-in processor, the terminal searches for the thread queue of the thread executing the scheduled task on the moving-out processor in descending order of load. Specifically, the terminal acquires the scheduled task execution threads and their load on the outgoing processor, sorts the threads for executing the scheduled tasks according to the load, and further, the terminal searches for the scheduled task threads in descending order of the load. For the thread queue of the task thread, go to step 104. If there is no thread executing the scheduled task running on the outgoing processor and there is a thread executing the scheduled task running on the incoming processor, the search ends.

In this embodiment, when a thread that executes a scheduled task on the incoming processor, the thread that performs the scheduled task with the largest load on the outgoing processor is preferentially scheduled to the incoming processor, so that the thread can be scheduled with the least amount of scheduling. The thread that executes the scheduled task is load-balanced on the outgoing processor and the incoming processor, and when there is a thread that executes the scheduled task on the incoming processor and there is no thread that executes the scheduled task on the outgoing processor, the search ends. That is, thread scheduling between the outgoing processor and the incoming processor is terminated to ensure that the thread executing the scheduled task can be executed on the incoming processor with maximum efficiency, thereby further improving the efficiency of executing the scheduled task.

In one embodiment, the step of finding the thread queue running on the outgoing processor includes: step 308, when the processor core group where the outgoing processor is located is a large core group, the process where the incoming processor is located When the processor core group is a small core group, perform the following steps: If the total load of the threads executing scheduled tasks on the outgoing processor is the largest in the large core group, search for the execution schedule according to the order of thread load from small to large. The task's thread queue; otherwise, end looking up the thread queue for executing the scheduled task.

In this embodiment, when the processor core group where the outgoing processor is located is a large core group, and the processor core group where the incoming processor is located is a small core group, the thread processing efficiency of the outgoing processor is higher than that of the incoming processor. When processing thread processing efficiency of the processor, further, the terminal judges whether there is a thread executing a predetermined task on the incoming processor. Specifically, the terminal acquires the threads executing the scheduled tasks on the outgoing processor and the loads of each thread, and sums the loads of the threads executing the scheduled tasks to obtain the total load of the threads executing the scheduled tasks. Further, the terminal Obtain the total load of threads executing scheduled tasks on other processors in the terminal, and determine whether the total load of threads executing scheduled tasks on the outgoing processor is among the total loads of threads executing scheduled tasks on all processors in the terminal If it is the largest, then the terminal searches the thread queue for executing the scheduled task according to the order of thread load from small to large. Otherwise, the terminal finishes searching the thread queue for executing the predetermined task, further, the terminal searches the thread queue in order of priority from low to high, and executes step 104 .

In this embodiment, when the total load of the threads executing the scheduled tasks on the outgoing processor is the largest, the thread with the largest load for executing the scheduled tasks is selected and dispatched to the incoming processor; otherwise, the thread executing the scheduled tasks is not scheduled, The thread that executes the scheduled task can be reserved on the outgoing processor with higher execution efficiency to the maximum extent, ensuring that the scheduled task can be executed more efficiently.

In one embodiment, the thread scheduling management method further includes: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the input and output device within a preset time; if accessing the input and output If the number of times of the device is greater than the first threshold, the running thread is recorded as the first type thread.

There are input/output (Input/Output, I/O) intensive tasks in the tasks performed by threads. Tasks involving network and disk I/O are all I/O-intensive tasks. This type of task is characterized by low processor consumption, because the speed of I/O is far lower than the speed of the processor and memory. Most of the time is spent waiting for I/O operations to complete. For this type of task thread, it is necessary to identify it for further targeted management, so as to improve the operating speed of the terminal.

When the thread is in the running state, the original data needed for the thread to run comes from the memory. During thread running, some data may be frequently read, and these data will be stored in registers and caches. When the thread finishes running, these cached data should be written back to memory when appropriate. The storage resources used by threads in the running state may include registers, caches, memory, and external storage devices. The information of the storage resource may include the address space of the storage resource used by the thread.

The terminal can access the storage resource of the thread according to the thread identification recorded in the thread identification file, and obtain the information of the storage resource used by the running thread.

The storage resource information of the thread includes the calculation data when the thread is running, including the data of each access to the I/O device, specifically including the address information of the I/O device when each access to the I/O device, and each access to the I/O device The access time of the device. By accessing the storage resource of the thread, the terminal intercepts the time period during which one or more threads are continuously executed in the storage resource information record, that is, the preset time, and counts the addresses of the I/O devices contained in the storage resource information within the preset time The number of messages, which is the number of times the thread accesses the I/O device when it runs within the preset time. Wherein, the types of the address information of the I/O device may be one or more. The preset time includes only the period of time that the thread is running continuously.

In this embodiment, the terminal compares the counted number of times the running thread accesses the I/O device with the first threshold, and when the number of accesses to the I/O device is greater than the first threshold, the terminal adds the first threshold to the thread identification file. Type thread, and record the running thread ID corresponding to it. Wherein, the first threshold is a preset constant greater than 0, and those skilled in the art can set the value according to specific terminal standards. The first type of thread is a thread that performs I/O-intensive tasks.

In a multi-processor terminal, there may be multiple threads running at the same time. At this time, the terminal will access the storage resources of all running threads, respectively obtain the storage resource information corresponding to the running threads, and count each running thread. The number of times the running thread accesses the input and output device within the preset time is compared with the first threshold, and the first type of thread among all the running threads is identified and recorded.

Further, the terminal sets priorities for threads on the processor. Specifically, the terminal sets thread priorities to four levels, namely 1, 2, 3, and 4, wherein 1 is the highest priority, and 2, 3, and 4 are successively lowered. The terminal sets the priority of the thread executing the scheduled task as 1, sets the priority of the first type of thread as 2, and sets the priority of other threads as 3, and sets the priority of other threads except the thread executing the scheduled task and the first type of thread The priority is set to 4. The terminal's thread scheduler schedules threads according to priority.

As shown in Figure 4a, thread A is executed on the first processor (CPU0), wherein thread A is a thread for performing a predetermined task, and the priority of thread A is 1, and thread A is executed on the second processor (CPU1). There are thread B1 and thread B2 being executed, and the priorities of thread B1 and thread B2 have been set to 3. Among them, thread B2 is a thread awakened by thread A, and thread B1 is not a thread awakened by thread A. As shown in Figure 4b, when the performance of CPU0 is better than that of CPU1, and the terminal needs to schedule threads from CPU1 to CPU0, The terminal temporarily sets the priority of thread B2 to 2, and the thread scheduler selects thread B2 with a higher priority on CPU1, and schedules thread B2 to CPU0, so that thread B2 can preferentially obtain the first processing of thread A that wakes it up. device, so as to ensure that the scheduled tasks executed by thread A are executed efficiently.

Through the thread scheduling management method of this embodiment, the terminal counts the number of times the running thread accesses the I/O device within the preset time, compares it with the first threshold, identifies and records the I/O-intensive tasks in the thread Threads, which set the priority of threads that perform I/O-intensive tasks higher than other threads, so that threads that perform I/O-intensive tasks can obtain better processors than other threads, thereby further improving the execution of scheduled tasks efficiency.

In one embodiment, the thread scheduling management method further includes: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the memory within a preset time; if the number of accesses to the memory is greater than second threshold, the running thread is recorded as the second type of thread.

There are also computationally intensive tasks performed by threads. Computation-intensive tasks are characterized by a large number of calculations that consume processors, such as calculating pi, decoding high-definition videos, and so on. For threads that execute computation-intensive tasks, their operating efficiency is very dependent on the probability of obtaining a processor. Therefore, identifying computation-intensive tasks for further targeted management is beneficial to improving the operating speed of the terminal.

When a thread issues instructions to the processor to request operations, these instructions and data are temporarily stored in the memory and sent to the processor when the processor is idle, that is, each time the thread issues an instruction to the processor, it will access the memory to temporarily store instructions and related data. data. Therefore, by counting the number of times a thread accesses memory within a certain period of time, the number of times the thread requests the processor within a certain period of time can be obtained, so as to determine whether it is a thread performing a computationally intensive task.

In this embodiment, the terminal accesses the storage resource of the thread according to the thread identifier recorded in the file, and obtains information about the storage resource used by the running thread.

The storage resource information of the thread includes the operation data when the thread is running. The operation data includes the data of each memory access operation, specifically including the address information of the memory each time the memory is accessed. Each piece of memory address information in the memory corresponds to an access time. . By accessing the storage resource of the thread, the terminal intercepts the time period during which one or more threads are continuously executed in the storage resource information record, that is, the preset time, and counts the number of address information of the memory contained in the storage resource information within the preset time , the number is the number of times the thread accesses the memory within the running preset time. Wherein, the types of the address information of the memory may be one or more.

The terminal compares the counted number of times the running thread accesses the memory with the second threshold, and when the number of accesses to the memory is greater than the second threshold, adds the second type of thread in the thread identification file, and marks the running thread as record corresponding to it. Wherein, the second threshold is a preset constant greater than 0, and those skilled in the art can set the value according to specific terminal standards. The second type of thread is a thread that performs computationally intensive tasks.

In a multi-processor terminal, there may be multiple running threads at the same time. At this time, the terminal will access the storage resources of all running threads, respectively obtain the storage resource information corresponding to the running threads, and count each The number of times the running threads access the memory within the preset time is compared with the second threshold, and the second type of thread among all the running threads is identified and recorded.

Further, the terminal sets priorities for threads on the processor. Specifically, the terminal sets thread priorities to four levels, that is, 1, 2, 3, and 4, wherein 1 is the highest priority, and 2, 3, and 4 are successively lowered. The terminal sets the priority of the thread executing the scheduled task to 1, sets the priority of the second type thread to 2, and sets the priority of other threads to 3, and sets the priority of other threads except the thread executing the scheduled task and the second type thread The priority is set to 4. The terminal's thread scheduler schedules threads according to priority.

Through the thread scheduling management method of this embodiment, the terminal counts the number of times the running thread accesses the memory within a preset time, compares it with the second threshold, identifies and records the calculation-intensive task thread in the thread, and executes the calculation. The priority of the threads of intensive tasks is set higher than that of other threads, so that the threads executing calculation-intensive tasks can obtain better processors than other threads, thereby further improving the execution efficiency of scheduled tasks.

It should be understood that although the various steps in the flowcharts of FIG. 1 and FIG. 3 are shown sequentially as indicated by the arrows, these steps are not necessarily executed sequentially in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least some of the steps in FIG. 1 and FIG. 3 may include multiple sub-steps or multiple stages, these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times, these sub-steps or The execution order of the stages is not necessarily performed sequentially, but may be executed alternately or alternately with at least a part of other steps or substeps of other steps or stages.

In one embodiment, as shown in FIG. 5 , a thread scheduling management device is provided, including: a selection module 510, a search module 520 and a scheduling module 530, wherein:

The selection module 510 is configured to select a transfer-out processor and a transfer-in processor, and determine whether the transfer-out processor and the transfer-in processor meet the thread migration condition; wherein, the transfer-out processor includes one or more thread queues.

The search module 520 is configured to search for threads running in the thread queue on the outgoing processor if the migration condition is satisfied, and calculate the remaining load to be migrated corresponding to the found threads.

The scheduling module 530 is configured to select a target thread according to the remaining load to be migrated, and schedule the target thread to the migrating processor.

In one embodiment, the selection module is also used to meet one of the following conditions when the outgoing processor and the incoming processor meet one of the following conditions: the incoming processor runs threads that perform predetermined tasks; If there is only one thread and it is a thread executing a predetermined task, and the transfer-in processor is not in an idle state; then it is judged that the thread migration condition is not satisfied; otherwise, it is judged that the thread migration condition is satisfied.

In one embodiment, the search module is also used to obtain the processor core group information where the outgoing processor is located and the processor core group information where the incoming processor is located; when the outgoing processor and the incoming processor are in the same processor When the core group is used, obtain the number of threads in the thread queue with the highest priority on the outgoing processor; if there are multiple threads, search the thread queue in order of priority from high to low; if the number of threads is one , search the thread queue in order of priority from low to high.

In one embodiment, the search module is also used to perform the following steps when the processor core group where the outgoing processor is located is a small core group and the processor core group where the incoming processor is located is a large core group: If there is no thread executing the scheduled task running on the processor, search for the thread queue of the thread executing the scheduled task in the order of the load from high to low; If there are threads executing scheduled tasks, the search ends.

In one embodiment, the search module is also used to perform the following steps when the processor core group where the outgoing processor is located is a large core group and the processor core group where the incoming processor is located is a small core group: If the total load of threads executing scheduled tasks on the processor is the largest in the large core group, search for thread queues for executing scheduled tasks according to the order of thread loads from small to large; otherwise, end the search for thread queues for executing scheduled tasks.

In one embodiment, the thread scheduling management device further includes: a classification module, configured to acquire storage resource information corresponding to the running thread; use the storage resource information to count the number of times the running thread accesses the input and output device within a preset time ; If the number of accesses to the input and output devices is greater than the first threshold, record the running thread as the first type thread.

In one embodiment, the classification module is also used to obtain the storage resource information corresponding to the running thread; use the storage resource information to count the number of times the running thread accesses the memory within a preset time; if the number of accesses to the memory is greater than the second threshold, the running thread is recorded as a second type thread.

For the specific limitations of the thread scheduling management device, please refer to the above definition of the thread scheduling management method, which will not be repeated here. Each module in the above-mentioned thread scheduling management device can be fully or partially realized by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, and can also be stored in the memory of the computer device in the form of software, so that the processor can invoke and execute the corresponding operations of the above-mentioned modules.

In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure may be as shown in FIG. 6 . The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Wherein, the processor of the computer device is used to provide calculation and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a thread scheduling management method is realized. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a button, a trackball or a touch pad provided on the casing of the computer device , and can also be an external keyboard, touchpad, or mouse.

Those skilled in the art can understand that the structure shown in FIG. 6 is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation on the computer equipment to which the solution of this application is applied. The specific computer equipment can be More or fewer components than shown in the figures may be included, or some components may be combined, or have a different arrangement of components.

In one embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored on the memory and operable on the processor. When the processor executes the computer program, the following steps are implemented: selecting the outgoing processor and Move-in processor, judge whether the move-out processor and the move-in processor meet the thread migration condition; wherein, the move-out processor includes one or more thread queues; if so, find the threads running in the thread queue on the move-out processor Thread, calculate the remaining load to be migrated corresponding to the found thread; select the target thread according to the remaining load to be migrated, and schedule the target thread to the migration processor.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: when the moving-out processor and the moving-in processor meet one of the following conditions: the moving-in processor runs a thread for performing a predetermined task; the moving-out processor If there is only one optional thread to be migrated and it is a thread that executes a predetermined task, and the transfer-in processor is not in an idle state; then it is judged that the thread migration condition is not satisfied; otherwise, it is judged that the thread migration condition is satisfied.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining the processor core group information where the outgoing processor is located and the processor core group information where the incoming processor is located; When the processor is in the same processor core group, obtain the number of threads in the thread queue with the highest priority on the outgoing processor; if there are multiple threads, search the thread queue in order of priority from high to low; If the number of threads is one, search the thread queue in order of priority from low to high.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: when the processor core group where the outgoing processor is located is a small core group, and the processor core group where the incoming processor is located is a large core group, execute The following steps: if there is no thread executing the scheduled task on the moving-in processor, search for the thread queue of the thread executing the scheduled task in the order of load from high to low; if there is no thread executing the scheduled task running on the moving-out processor and If there are threads executing scheduled tasks running on the incoming processor, the search ends.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: when the processor core group where the outgoing processor is located is a large core group, and the processor core group where the incoming processor is located is a small core group, execute The following steps: if the total load of the thread that performs the scheduled task on the outgoing processor is the largest in the large core group, then search for the thread queue that performs the scheduled task according to the load of the thread from small to large; otherwise, end the search and execute the scheduled task A thread queue for tasks.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the input and output device within a preset time; If the number of accesses to the input and output devices is greater than the first threshold, the running thread is recorded as the first type thread.

In one embodiment, when the processor executes the computer program, the following steps are also implemented: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the memory within a preset time; If the memory count is greater than the second threshold, the running thread is recorded as the second type thread.

In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented: selecting a moving-out processor and a moving-in processor, determining whether the moving-out processor and whether the migrating-in processor meets the thread migration condition; wherein, the migrating-out processor includes one or more thread queues; if so, find the threads running in the thread queue on the migrating-out processor, and calculate the remaining thread corresponding to the found thread The amount of load to be migrated; the target thread is selected according to the remaining load to be migrated, and the target thread is scheduled to the migration processor.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: when the moving-out processor and the moving-in processor meet one of the following conditions: the moving-in processor runs a thread for performing a predetermined task; the moving-out processing If there is only one optional thread to be migrated on the processor and it is a thread that executes a predetermined task, and the migrating processor is not in an idle state; then it is judged that the thread migration condition is not satisfied; otherwise, it is judged that the thread migration condition is satisfied.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining information about the processor core group where the outgoing processor is located and information about the processor core group where the incoming processor is located; When the incoming processor is in the same processor core group, obtain the number of threads in the thread queue with the highest priority on the outgoing processor; if there are multiple threads, search the thread queue in order of priority from high to low ; If the number of threads is one, search the thread queue in order of priority from low to high.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: when the processor core group where the outgoing processor is located is a small core group, and the processor core group where the incoming processor is located is a large core group, Perform the following steps: If there is no thread executing the scheduled task running on the moving-in processor, search for the thread queue of the thread executing the scheduled task in the order of load from high to low; if there is no thread executing the scheduled task running on the moving-out processor And if there are threads executing scheduled tasks running on the incoming processor, the search ends.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: when the processor core group where the outgoing processor is located is a large core group, and the processor core group where the incoming processor is located is a small core group, Execute the following steps: if the total load of the threads executing the scheduled tasks on the outgoing processor is the largest in the large core group, then search for the thread queues that execute the scheduled tasks in order of the thread loads from small to large; otherwise, end the search execution Thread queue for scheduled tasks.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the input and output device within a preset time ; If the number of accesses to the input and output devices is greater than the first threshold, record the running thread as the first type thread.

In one embodiment, when the computer program is executed by the processor, the following steps are also implemented: obtaining the storage resource information corresponding to the running thread; using the storage resource information to count the number of times the running thread accesses the memory within a preset time; if If the number of memory access times is greater than the second threshold, the running thread is recorded as the second type thread.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented through computer programs to instruct related hardware, and the computer programs can be stored in a non-volatile computer-readable memory In the medium, when the computer program is executed, it may include the processes of the embodiments of the above-mentioned methods. Wherein, any references to memory, storage, database or other media used in the various embodiments provided in the present application may include non-volatile and/or volatile memory. Nonvolatile memory can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Chain Synchlink DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

The technical features of the above embodiments can be combined arbitrarily. To make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be It is considered to be within the range described in this specification.

The above-mentioned embodiments only represent several implementation modes of the present application, and the description thereof is relatively specific and detailed, but it should not be construed as limiting the scope of the patent for the invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the scope of protection of the patent application should be based on the appended claims.

Claims (10)

1. A thread scheduling management method, said method comprising: Selecting the outgoing processor and the incoming processor, and judging whether the outgoing processor and the incoming processor meet the thread migration condition; wherein, the outgoing processor includes one or more thread queues; If yes, find the threads running in the thread queue on the outgoing processor, and calculate the remaining load to be migrated corresponding to the found threads; A target thread is selected according to the remaining load to be migrated, and the target thread is scheduled to the transfer-in processor. 2. The thread scheduling management method according to claim 1, wherein the step of judging whether the moving-out processor and the moving-in processor satisfy the thread migration condition comprises: When the outgoing processor and incoming processor meet one of the following conditions: There are threads executing scheduled tasks running on the incoming processor; There is only one optional thread to be migrated on the outgoing processor and it is a thread that executes a scheduled task, and the incoming processor is not idle; Then it is judged that the thread migration condition is not satisfied; otherwise, it is judged that the thread migration condition is satisfied. 3. The thread scheduling management method according to claim 1, wherein the step of searching and running in the thread queue on the outgoing processor comprises: Acquiring the processor core group information where the outgoing processor is located and the processor core group information where the incoming processor is located; When the moving-out processor and the moving-in processor are in the same processor core group, obtain the number of threads in the thread queue with the highest priority on the moving-out processor; If the number of threads is multiple, search the thread queue in order of priority from high to low; If the number of threads is one, the thread queue is searched in order of priority from low to high. 4. The thread scheduling management method according to claim 3, wherein the method further comprises: When the processor core group where the moving-out processor is located is a small core group, and the processor core group where the moving-in processor is located is a large core group, perform the following steps: If there is no thread executing the scheduled task running on the incoming processor, search for the thread queue of the thread executing the scheduled task in the order of load from high to low; If there is no thread executing the scheduled task running on the outgoing processor and there is a thread executing the scheduled task running on the incoming processor, the search ends. 5. The thread scheduling management method according to claim 3, wherein the method further comprises: When the processor core group where the moving-out processor is located is a large core group, and the processor core group where the moving-in processor is located is a small core group, perform the following steps: If the total load of the threads executing scheduled tasks on the outgoing processor is the largest in the large core group, then search the thread queue for executing scheduled tasks according to the order of thread loads from small to large; otherwise, end the search for threads that execute scheduled tasks queue. 6. The method according to claim 1, wherein the method further comprises: Obtain the storage resource information corresponding to the running thread; Using the storage resource information to count the number of times the running thread accesses the input and output device within a preset time; If the number of accesses to the input and output device is greater than a first threshold, record the running thread as a first type thread. 7. The method according to claim 1, wherein the method further comprises: Obtain the storage resource information corresponding to the running thread; Using the storage resource information to count the number of times the running thread accesses the memory within a preset time; If the number of memory access times is greater than a second threshold, record the running thread as a second type thread. 8. A thread scheduling management device, characterized in that the device comprises: The selection module is used to select the outgoing processor and the incoming processor, and judge whether the outgoing processor and the incoming processor meet the thread migration condition; wherein, the outgoing processor includes one or more thread queues; A search module, configured to search for threads running in the thread queue on the outgoing processor if the migration conditions are met, and calculate the remaining load to be migrated corresponding to the found threads; A scheduling module, configured to select a target thread according to the remaining load to be migrated, and schedule the target thread to the transfer-in processor. 9. A computer device, comprising a memory, a processor, and a computer program stored on the memory and operable on the processor, characterized in that, when the processor executes the computer program, any one of claims 1 to 7 is realized. A step of said method. 10. A computer-readable storage medium, on which a computer program is stored, characterized in that, when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are realized.