KR101463533B1 - Method of core scheduling for asymmetric multi-core processor in a mobile device and apparatus for performing the same - Google Patents

Abstract

There is provided an asymmetric multiprocessor core scheduling method and a mobile device in a mobile device that performs scheduling based on CPU usage of each application in an AMP environment of a mobile device. The scheduling method comprising the steps of: a first core having a first operating speed executing the application in response to an application execution request; a second core having a second operating speed monitoring a utilization rate of the CPU; And executing the application by the second core when the first usage rate is less than a first usage rate. Thus, power consumption can be lowered and the battery usage time of the mobile device can be increased.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention [0001] The present invention relates to an asymmetric multiprocessor core scheduling method and a mobile device in a mobile device.

The present invention relates to a method and apparatus for asymmetric multiprocessor core scheduling in a mobile device, and more particularly to a core scheduling method and a mobile device capable of reducing power consumption.

Processor performance has been rapidly improving. In order to improve the performance of the core, out-of-order and pipelining technologies have been developed and applied. However, performance improvements through these techniques are limited, because if the clock frequency is raised above a certain level, the processor consumes a lot of power. In addition, because of the severe heat problem, the performance improvement of the processor using the increase of the clock frequency has reached a limit. Thus, recent processors have been developed as multi-cores, which improves overall system performance by increasing parallelism.

Multicore processors have been mainly used for high performance PCs and servers, but recently they have started to be adopted for mobile devices such as smart phones and tablets. Although multicore consumes less power than superscalar processors, power consumption is expected to increase if mobile devices are equipped with more cores in the future.

Recently, a processor called asymmetric multi-core processor (AMP) has begun to emerge. Unlike conventional symmetric multi-core processors (SMPs) , Studies on cores with excellent parallelism are underway.

In general, high-performance cores are configured to maximize the capacity of superscalar to accommodate higher throughputs. They are concentrated on high-capacity processing and have high performance, but have a large area within the chip, a complex structure, and high power consumption. On the other hand, low-performance cores are simple and feature-less, but they are small and seek efficiency. It has a relatively small area and can be designed with much less power consumption.

AMP is a second core having the advantages of a first core with a first operating speed and a second operating speed by integrating different cores on the same chip while having the same instruction set architecture (ISA) And that the operating speed is faster than the second operating speed. In addition, since the ISA of each core is the same, execution of a thread in any core can be executed without special instruction conversion.

Korean Registered Patent No. 10 - 1065436 ("Stochastic Scheduling Method of Multicore Processor for Real-Time Parallel Operation with Uncertain Computation ", Registered on September 8, 2011 by Gyeongsang National University Industry &

An object of the present invention is to provide a first core having a low CPU usage in a second core having a second operating speed and having a first operating speed in consideration of CPU usage of each application in an AMP environment of a mobile device, Characterized in that the first operating speed is faster than the second operating speed by performing a scheduling based on the CPU usage by lowering the power consumption and increasing the battery usage time of the mobile device, And to provide a core scheduling method.

It is another object of the present invention to provide a method and apparatus for performing a task with a low CPU usage in a second core having a second operating speed and a first core having a first operating speed, Wherein the first operating speed is higher than the second operating speed. The present invention provides a mobile device capable of reducing power consumption and increasing battery usage time by performing scheduling based on CPU usage by shutting off the power supply.

According to an aspect of the present invention, there is provided a method for scheduling asymmetric multiprocessors in a mobile device, the method comprising: a first core having a first operating speed for executing the application; Wherein the first operating speed is greater than the second operating speed; monitoring a utilization of the CPU; And executing the application by the second core when the CPU usage rate is less than a preset first usage rate. The method may further include disconnecting power of the first core when the second core executes the application and a predetermined time has elapsed. In addition, the step of shutting off the power supply may be such that, when the CPU usage rate is less than a second usage rate set in advance, the power supply of the first core is shut off after a lapse of time less than when the CPU usage rate is equal to or higher than the second usage rate . And the second usage rate is smaller than the first usage rate. Here, the predetermined time may be set by a user's input. Also, the first usage rate may be determined based on the maximum frequency of the second core and the maximum frequency of the first core. The first usage rate may be determined based on a value obtained by dividing a maximum frequency of the second core by a maximum frequency of the first core. Here, when the application is terminated, the second core may perform a background task.

According to another aspect of the present invention, there is provided a mobile device comprising: a first core for executing an application in response to an application execution request; And a second core that monitors the usage rate of the CPU and executes the application when the CPU usage rate is less than a preset first usage rate. The controller may further include a controller for shutting down the power of the first core when the second core executes the application and a predetermined time has elapsed. The controller may cut off power to the first core after a lapse of a shorter time than when the CPU usage rate is equal to or higher than the second usage rate when the CPU usage rate is less than a second usage rate set in advance. Here, the second usage rate may be smaller than the first usage rate. Here, the apparatus may further include an input unit for receiving a numerical value by a user operation, wherein the predetermined time is set by a user's input through the input unit. The first usage rate may be determined based on the maximum frequency of the second core and the maximum frequency of the first core, and the first usage rate is a value obtained by dividing the maximum frequency of the second core by the maximum frequency of the first core . ≪ / RTI > In addition, the second core may perform a background task when the application is terminated.

According to the asymmetric multiprocessor core scheduling method and the mobile device in the mobile device according to the embodiment of the present invention, by using the scheduling method considering the CPU usage of each application on the assumption of the AMP environment in the mobile device, You can suggest ways to increase your phone's battery usage time. By using the proposed scheduling method, it is possible to minimize the consumption of the battery by performing the task with low CPU usage like the music player on the second core and turning off the power of the first core. Furthermore, Can be improved.

1 is a graph showing a CPU usage amount in a standby mode state.
2 is a graph showing the CPU usage when the music player is executed.
3 is a graph showing the CPU usage when the Android market application is executed.
4 is a graph showing the classification according to the characteristics of the application.
5 is a conceptual diagram of an asymmetric multiprocessor core scheduling method in a mobile device according to an embodiment of the present invention.
6 is a flow diagram of a method for asymmetric multiprocessor core scheduling in a mobile device in accordance with an embodiment of the present invention.
7 is a block diagram illustrating a configuration of a mobile device according to an embodiment of the present invention.
8 is a graph showing the total execution time of the workload and the processing time per request when the workload is executed in the first core and the second core.
9 is a graph comparing the completion time according to the core allocation of the CPU-bound workload and the I / O-bound workload in the AMP environment.
10 is a graph showing a processing time of a workload according to a combination of various cores and whether or not a workload is scheduled for each core.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail.

It should be understood, however, that the invention is not intended to be limited to the particular embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention.

In the following description of the embodiments of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present disclosure rather unclear.

The terms first, second, etc. may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The terminology used in this application is used only to describe a specific embodiment and is not intended to limit the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In the present application, the terms "comprises" or "having" and the like are used to specify that there is a feature, a number, a step, an operation, an element, a component or a combination thereof described in the specification, But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

In addition, the components shown in the embodiments of the present invention are shown independently to represent different characteristic functions, which does not mean that each component is composed of separate hardware or software constituent units. That is, each constituent unit is included in each constituent unit for convenience of explanation, and at least two constituent units of the constituent units may be combined to form one constituent unit, or one constituent unit may be divided into a plurality of constituent units to perform a function. The integrated embodiments and separate embodiments of the components are also included within the scope of the present invention, unless they depart from the essence of the present invention.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the relevant art and are to be interpreted in an ideal or overly formal sense unless explicitly defined in the present application Do not.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In order to facilitate the understanding of the present invention, the same reference numerals are used for the same constituent elements in the drawings and redundant explanations for the same constituent elements are omitted.

Further, for a first core having a first operating speed and a second core having a second operating speed, the first operating speed is less than the second operating speed.

FIG. 1 is a graph showing a CPU usage amount in a standby mode state, FIG. 2 is a graph showing a CPU usage amount when a music player is executed, and FIG. 3 is a graph showing a CPU usage amount when an Android market application is executed.

With reference to Figs. 1 to 3, classification of applications will be described below. 1 to 3 show the CPU usage for various types of applications operating on a smartphone. 1 shows the CPU usage in the standby mode. As shown in FIG. 1, the CPU usage in the standby mode state shows a very low usage rate of about 1%.

2 shows the CPU usage rate when the music player is executed. As shown in FIG. 2, the com.android.music process uses a small amount of CPU when processing a player function such as a song selection or a volume control. However, when there is no input from the user, only the media server process uses the CPU.

FIG. 3 shows the CPU utilization when the Android market application is executed. As shown in FIG. 3, if the traffic to the server is large, the CPU utilization rate increases.

4 is a graph showing the classification according to the characteristics of the application.

The contents of FIGS. 1 to 3 are summarized in FIG. Most applications can be classified into two broad categories. The H group with high CPU usage rate will have a fast response time, so using the first core will be advantageous in terms of overall performance. If the L group is performed in the second core, the power consumption will be reduced while the performance is not significantly degraded as compared with that in the first core.

Since the first and second cores of the AMP use the same instruction set architecture (ISA), it is possible to execute the threads in any core without any special instruction conversion process. However, if the operating system running on the AMP hardware can not recognize it and all of them are treated as the same core, and the task is allocated according to the existing scheduling method, the advantage of the AMP can not be utilized 100%, and the overall performance is deteriorated.

Also, in terms of power consumption, when the size of the core is reduced by 1/4, the amount of power consumed is halved. Accordingly, when the applications to be executed in the second core are migrated from the first core to the second core and the power of the first core is cut off, the power loss can be reduced accordingly.

5 is a conceptual diagram of an asymmetric multiprocessor core scheduling method in a mobile device according to an embodiment of the present invention. Referring to FIG. 5, the asymmetric multiprocessor core scheduling method will be described in detail.

As shown in FIG. 5, when all the background tasks are allocated to the second core 20 of the AMP when the smartphone is in the standby mode (S510), battery consumption is reduced You will be able to extend the waiting time. Currently, dual-core smartphones reduce power consumption by lowering the frequency when core usage is reduced. However, since only the frequency is reduced, the power saving effect can not be obtained as much as the second core 20.

Referring again to FIG. 5, when the active mode is switched to the active mode (S520), the first core 10 can execute the application (S521). It is important for smartphone users to satisfy the user's QoS since they are generally not tolerant of the operating speed of the application. Therefore, if the input of the user is transmitted to the screen and the application is executed, it should be allocated to the first core 10. Therefore, it is a principle that the fore-ground task is always allocated to the first core 10.

At this time, the task of monitoring the CPU usage (S523) may be started in the second core. If it is less than the preset first CPU usage amount, Core Affinity can be set so that the application can be executed in the second core 20 (S525). In more detail, monitoring can be performed in seconds and divided into three classes, for example, according to CPU usage. For example, if the CPU usage is equal to or greater than a first CPU usage amount set in advance, it can be classified as a high class. If the CPU usage amount is less than the first CPU usage amount set in advance and is equal to or larger than the second CPU usage amount Medium class, and if the CPU usage is less than the second CPU usage, it can be classified as a low class.

Here, the value obtained by dividing the maximum frequency of the second core 20 by the maximum frequency of the first core 10 may be determined as a percentage and set as a threshold value. The threshold value may be the first CPU usage amount. If the first core is twice as fast as the second core, the threshold value can be 50%.

The threshold value can be used as a criterion for turning on and off the first core 10. For example, if the CPU usage is 100%, it can be classified as a 'high' class with a threshold value higher than a threshold value, a 40% 'medium' class as a threshold value and a 'low' have. The basic algorithm can set the CPU affinity value so that the applications of the medium class and the low class can operate in the second core 20.

After a certain time, it is determined that there is no input by the user and the first core 10 is turned off. That is, when the second core 20 executes the application and the predetermined time has elapsed, the first core 10 is not used any longer, so that the power of the first core 10 can be shut off (S527) .

Here, since the probability of not using the first core 10 is higher than that of the low class application, the time until the first core 10 is turned off can be shortened. For example, if the medium class is classified into the medium class, the power of the first core 10 may be shut off after 30 seconds have elapsed. In addition, if the CPU usage is less than the second CPU usage, it can be classified as a low class, and power of the first core 10 can be shut off after 15 seconds, for example.

The time such as 30 seconds or 15 seconds can be set by the user like the screen timeout of the smartphone. This is because users have different preferences. Thus, if there is no application using the first core 10, the power of the first core 10 can be cut off after waiting for a predetermined time.

6 is a flow diagram of a method for asymmetric multiprocessor core scheduling in a mobile device in accordance with an embodiment of the present invention.

As shown in FIG. 6, an asymmetric multiprocessor core scheduling method in a mobile device according to an exemplary embodiment of the present invention includes: first, in response to an application execution request, a first core executes the application; The usage rate of the CPU can be monitored (S610). As described above, since users are not generally tolerant of the operation speed of the application, the first core can cause the first core to execute the application in order to satisfy the QoS of the user. The second core can monitor the CPU usage.

 As a result of the monitoring, if the CPU usage rate is less than a preset first usage rate, the application can be executed by the second core (S620). That is, in the case where the executed application performs an operation that does not cause a delay of the speed without using the first core, the second core can execute the application, thereby further reducing power consumption.

Here, the first usage rate may be determined based on the maximum frequency of the second core and the maximum frequency of the first core. More specifically, the first usage rate may be a value obtained by dividing a maximum frequency of the second core by a maximum frequency of the first core. The maximum frequency of the cores is related to the maximum performance that each core can produce. Thus, based on the maximum frequency of each of the cores, it is possible to determine which one of the first core or the second core performs a specific application.

If the first core is twice as fast as the second core, 50% can be the threshold value. Therefore, when the CPU usage exceeds 50%, it is preferable that the first core continuously executes a specific application. When the CPU usage is less than 50%, it is preferable that the second core executes the specific application.

The second core may shut down the power of the first core after a predetermined time elapses after executing the application (S630). That is, when the second core executes the application and the predetermined time has elapsed, the first core is no longer used, so that power consumption of the first core is cut off to minimize power consumption.

Here, if the CPU usage rate is less than the preset second usage rate, the power of the first core may be shut off after a lapse of time less than when the CPU usage rate is equal to or higher than the second usage rate. The second usage rate may be less than the first usage rate. For example, when the first usage rate is higher than the first usage rate, the user is classified as a 'high' class. If the first usage rate is higher than the first usage rate, , And if it is lower than the second usage rate, it can be classified into a 'low' class. In case of the 'medium' class or the 'low' class, the second core may execute the application, and in the case of the 'low' class, the power of the first core may be cut off after a shorter time.

Here, the predetermined time may be set by a user's input. That is, it can be set by the user like the screen timeout of the smartphone. This is because users have different preferences.

Thereafter, when the application is terminated, the mobile device enters a standby mode, and the second core can perform a background task (S640). When the mobile device is in the standby mode, the second core performs the background task, which can drastically reduce battery consumption.

7 is a block diagram illustrating a configuration of a mobile device according to an embodiment of the present invention.

7, a mobile device 700 according to an embodiment of the present invention may include a first core 710, a second core 720, a controller 730, and an input 740 .

The first core 710 may execute the application in response to an application execution request.

The second core 720 monitors the CPU usage rate and can execute the application when the CPU usage rate is less than a preset first usage rate. Here, the first usage rate may be determined based on the maximum frequency of the second core 720 and the maximum frequency of the first core 710. More specifically, the first utilization rate may be determined based on a value obtained by dividing a maximum frequency of the second core 720 by a maximum frequency of the first core 710. [ In addition, the second core 720 may perform a background task when the application ends.

The control unit 730 can turn off the power of the first core 710 after a predetermined time elapses after the second core 720 executes the application. The controller 730 can shut off the power of the first core 720 after a lapse of a shorter time than when the CPU usage rate is equal to or higher than the second usage rate when the CPU usage rate is less than a second usage rate set in advance. And the second usage rate is smaller than the first usage rate.

The input unit 740 can receive a numerical value by the user's operation and the predetermined time can be set by the user's input through the input unit 740. [

Experimental Example

Since there is no commercially available mobile AMP, the experimental environment is configured to operate as a first core and a second core by using DVFS (dynamic voltage and frequency scaling) method. I used two AMD Opteron 6134 CPUs, and I set the clock frequency of the first core to 2.3 GHz, the clock frequency of the second core to 1.1 GHz, and the first core to achieve about twice the performance of the second core. The workload used in the experiments was SHA-1, one of the micro benchmarks, and File I / O of sysbench. They are CPU-bound and I / O-bound, respectively.

FIG. 8 shows the total execution time of the workload and the processing time per request by performing each workload on the first core (2.3 GHz) and the second core (1.1 GHz). I / O-bound workloads represent similar processing times in the first and second cores. 11.44 in the first core, and 11.25 in the second core, and there is no significant difference in the ending time of the workload. On the other hand, CPU-bound workloads have a large difference in request processing time per unit time. It can be seen that the faster the work is done on the first core, the same workload for the first core is 1.55 and the second core is 3.24.

Thus, it can be seen that CPU-bound workloads are suitable for using high-performance cores, but workloads that do not use a small amount of cores are not significantly different in performance from the second cores. Since there is no OS capable of recognizing the core of the AMP environment at present, it is important to allocate workloads with different core usage to the first core and the second core.

Figure 9 compares the completion time according to which core the CPU-bound workload and the I / O-bound workload are allocated in an AMP environment with one first core and one second core. In the BEST case, we can see that CPU-bound workloads are assigned to the first core and I / O-bound workloads are assigned to the second core, respectively, thus reducing the processing time of the entire workload as compared to WORST. Reducing the total processing time means that you can reduce the time you spend on the core and reduce battery consumption.

10 is a graph showing a processing time of a workload according to a combination of various cores and whether or not a workload is scheduled for each core. Whether or not the algorithm proposed in the present invention works as well as the BEST case of FIG. 9 can be confirmed with reference to FIG. We have created various AMP environments by changing the number of cores. Manufacturers' primary goal is to create chips that can handle fore-ground tasks as soon as possible. Therefore, we have examined the processing time of the workload and how well the workload is scheduled for each core while increasing the number of the first cores.

The numbers written in front of the first core FC and the second core SC in FIG. 10 indicate the number of cores. Thus, the first AMP environment is composed of two first cores and two second cores. The result of performing the same workload on each AMP environment, the CPU-bound workload was increased by increasing the number of threads to match the number of primary cores.

It can be seen that the CPU-bound workload is well-scheduled in the first core and the total processing time is shortened as the number of threads increases. In addition, since the power is cut off when the first core is not used, it can be seen that performance is improved while consuming almost the same power in each AMP environment of FIG. 10 by reducing the time that the first core is turned on.

In addition, in the future, applications will be able to maximize the efficiency of multi-cores if they are created with parallelism in mind.

10: first core 20: second core
700: mobile device 710: first core
720: second core 730: control unit
740:

Claims (16)

The first core having the first operating speed executes the application in response to the execution request of the application, and the second core having the second operating speed monitors the utilization of the CPU; And
And executing the application by the second core when the CPU usage rate is less than a preset first usage rate,
Wherein the first operating rate is faster than the second operating rate and the first utilization rate is determined based on the maximum frequency of the second core and the maximum frequency of the first core. Way.
The method according to claim 1,
And shutting down the power of the first core when the second core executes the application and a predetermined time has elapsed.
3. The method of claim 2,
When the CPU usage rate is less than a second usage rate that is set in advance, the power of the first core is shut off after a lapse of less time than when the CPU usage rate is equal to or higher than the second usage rate. Way.
The method of claim 3,
Wherein the second usage rate is less than the first usage rate.
3. The method of claim 2,
Wherein the asynchronous multiprocessor core scheduling method is set by a user's input.
delete 2. The method of claim 1,
Wherein the maximum frequency is determined based on a value obtained by dividing a maximum frequency of the second core by a maximum frequency of the first core.
The method according to claim 1,
And when the application is terminated, performing a background task by the second core. ≪ Desc / Clms Page number 19 >
A first core for executing the application in response to an execution request of the application; And
And a second core for monitoring a usage rate of the CPU and executing the application when the CPU usage rate is less than a first usage rate that is set in advance and the first usage rate is a maximum frequency of the second core and a maximum frequency of the first core Gt;
10. The method of claim 9,
Further comprising a controller for shutting down the power of the first core when the second core executes the application and a predetermined time has elapsed.
11. The apparatus of claim 10, wherein the control unit
And when the CPU usage rate is less than a second usage rate that is set in advance, the first core is powered off after a lapse of a shorter time than when the CPU usage rate is equal to or higher than the second usage rate.
12. The method of claim 11,
Wherein the second usage rate is less than the first usage rate.
11. The method of claim 10,
Further comprising an input section for receiving a numerical value by an operation of a user,
Wherein the predetermined time is set by a user's input through the input unit.
delete 10. The method of claim 9, wherein the first utilization rate
Wherein the maximum frequency of the second core is determined based on a value obtained by dividing a maximum frequency of the second core by a maximum frequency of the first core.
10. The method of claim 9, wherein the second core
And performs a background task when the application is terminated.

Images