JP2009175850A - Multi-arithmetic processing system, multi-arithmetic processing method, and multi-arithmetic processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To suppress power consumption related with task processing in a computer. <P>SOLUTION: This multi-arithmetic processing system is provided with: a plurality of central arithmetic processing parts 10 for performing the execution processing of a plurality of preliminarily set different tasks; a task monitoring part 13 for monitoring the utilization information of each central processing arithmetic part 10 for each task; an arithmetic part selection part 110 for selecting a central arithmetic processing part 10 for preferentially performing the execution processing of each task based on the monitor result; and a task allocation part 11 for preferentially allocating the task to each selected central arithmetic processing part. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for performing load distribution among central processing units in a multiprocessor system including a plurality of central processing units.

Currently, a multiprocessor system is being developed in which data processing is performed by a plurality of central processing units (CPUs) connected in parallel.
By efficiently allocating tasks and distributing processing load to each CPU in this multiprocessor system, the CPU can be used efficiently, reducing power consumption in computer operation, and mobile with multiple CPUs There is a need for a method of reducing power consumption that enables long-term operation of equipment.
On the other hand, as the processing capacity of each CPU increases, there is a disadvantage that the noise of the cooling device for the CPU that performs high-speed processing deteriorates the surrounding environment of the computer.

  As a related technology, the upper and lower limits of CPU usage are preset for each of multiple CPUs, and one task (process) of the CPU that has a usage rate outside the range between the upper and lower limits is extracted. CPU load distribution that allocates the extracted tasks by selecting the CPU whose usage rate is between the upper limit value and the lower limit value and selecting the CPU closer to the upper limit value when the usage rate of the extracted task is added Is disclosed (Patent Document 1).

  In addition, each server device is configured so that the CPU usage rate (server usage rate) for each of the servers indicating the total of all CPU usage rates provided in each server does not exceed the preset upper limit value of each server usage rate. A related technique for performing distributed processing of a CPU processing load by assigning tasks to is disclosed (Patent Document 2).

JP 09-274608 A JP 2005-182641 A

However, in the related information disclosed in Patent Document 1 above, tasks assigned to each CPU are based on the upper limit value (upper limit usage rate) of each CPU, and the usage rate is uniformly high on all CPUs. obtain. For this reason, the power consumption in each CPU increases uniformly, and the noise accompanying the operation of the cooling device that cools each CPU may also increase.
Further, the related information disclosed in Patent Document 2 is difficult to apply to a multiprocessor device having a plurality of different CPUs, and each task assigned is assigned in the same manner as Patent Document 1 described above. There is an inconvenience that the usage rate of the CPU may be uniformly high.

[Object of invention]
It is an object of the present invention to provide a multi-operation processing system, a multi-operation processing method, and a multi-operation processing program that improve the disadvantages of the related technology and effectively suppress power consumption related to task processing in a computer. Objective.

  In order to achieve the above object, a multi-operation processing system according to the present invention is a multi-operation processing system that includes a plurality of different central operation processing units and performs execution processing of a plurality of different preset tasks, A task monitoring unit that monitors usage rate information of each central processing unit for each task, and a calculation unit selection unit that selects a central processing unit that preferentially executes the execution process of each task based on the monitoring result; The system includes a task assignment unit that preferentially assigns the task to each selected central processing unit.

  In addition, the multi-operation processing method according to the present invention includes a plurality of different central processing units, a task assignment unit that assigns a plurality of different tasks to the central processing units, and the central processing operations for the tasks. A task monitoring unit that monitors the usage rate of each unit, and a calculation unit selection unit that selects a central processing unit that preferentially assigns the tasks, and a multi-operation processing method that executes the task, CPU usage information acquisition step for acquiring usage information of each central processing operation unit for, a calculation unit selection step for selecting a central processing unit to which the task is preferentially assigned, and the selected central processing unit A task priority assignment step for preferentially assigning the task to the task, and an arithmetic processing step for performing arithmetic processing on the assigned task That.

  Furthermore, the multi-processing program according to the present invention includes a plurality of different central processing units, a task assigning unit that assigns a plurality of different tasks to the central processing units, and the central processing operations for the tasks. A multi-operation processing program for performing execution processing of the task, comprising: a task monitoring unit that monitors a usage rate of a unit; and a calculation unit selection unit that selects a central processing unit that preferentially assigns the task. CPU usage information acquisition function for acquiring usage information of each central processing arithmetic unit for each task, arithmetic unit selecting function for selecting a central arithmetic processing unit to which the task is preferentially assigned, and the selected central arithmetic processing A task priority assignment function for preferentially assigning the task to a part, and an arithmetic processing function for performing arithmetic processing on the assigned task. It is characterized by to be executed by the computer.

  Since the present invention is configured and functions as described above, according to this, a plurality of different central processing units and a task monitoring unit for monitoring the usage rate of each central processing unit for each different task are selected. In addition, a task allocation unit that preferentially allocates the task to the central processing unit enables appropriate task allocation control for each central processing unit, and task processing of the central processing unit It is possible to provide an information distribution system, an information distribution method, and an information distribution program that can effectively suppress power consumption.

[Embodiment]
Next, the basic configuration content of the embodiment of the present invention will be described.

  As shown in FIG. 1, the arithmetic device 1 according to the present embodiment allocates (dispatches) a plurality of different CPUs 10 that process assigned arithmetic processing tasks and tasks created in the arithmetic device 1 to the CPU 10. ) A task scheduler 11, a task monitoring unit 13 that monitors the proportion of each task created in the computing device 1 occupying the usage rate in each CPU, and a usage rate at which appropriate arithmetic processing is performed in each CPU 10 (hereinafter “ The target usage rate storage unit 12 manages the target usage rate set as the upper limit of “appropriate usage rate”.

  Note that the arithmetic device 1 has a plurality of CPUs 10 that operate by program control, a main memory unit (memory) that holds programs for controlling the CPUs 10 and data controlled by the programs, and a permanent operation even when the power is cut off. It is assumed that the computer operates at least including a secondary storage unit for holding programs and data. Here, in the arithmetic device 1, a task, which is a unit of arithmetic processing to be performed, is generated and managed by executing a program.

This will be described in detail below.
Each CPU 10 has an arithmetic processing function for performing arithmetic processing on a task assigned to each CPU 10 by a preset method;

The task scheduler 11 includes a target usage rate acquisition function for acquiring the target usage rate of each CPU 10 from the target usage rate table 120 and a CPU usage information acquisition function for acquiring CPU usage information of each task from the CPU usage information table 130. ing.
Further, the task scheduler 11 includes priority CPU selection means 110 that selects a CPU to which a task is preferentially assigned based on the type of each CPU 10 and the acquired CPU usage information of each task.

  As a result, for example, by selecting a CPU with high processing performance as the priority CPU among the CPUs 10, when the processing capacity required for each task is different, or when the priority order of assignment is set in advance for each task Even in such a case, it is possible to assign a task requiring high function processing or a task having a high allocation priority to the selected priority CPU.

  Further, the task scheduler 11 assigns each task to the selected CPU 10 based on the CPU selection result by the priority CPU selection means 110 so that the total usage rate of the assigned tasks approximates the target usage rate. It has.

Here, the task is assigned to the CPU 10 selected preferentially. However, when a task is assigned to a CPU 10 other than the selected one, each task is assigned so that it falls within the target usage rate. May be.
Thereby, as shown in FIG. 2, the selected CPU and the other CPUs are bipolarized into a high-load CPU group and a low-load CPU group. For this reason, the amount of electricity consumed for the cooling operation of each low-load CPU can be suppressed, and the power applied to the heat generation of the CPU as a whole in the arithmetic device 1 can be suppressed.
Further, the operation of the low-load CPU 10 can be effectively suppressed by shifting to a preset power saving mode.

In addition, the task scheduler 11 performs the assignment when the target assignment execution determination function for determining whether or not the target assignment function has been executed for all the CPUs 10 and the target assignment function for all the CPUs 10 are executed. A target unassigned task determination function is provided for determining whether or not an unassigned task (hereinafter referred to as “unassigned task”) remains.
Here, the task scheduler 11 assigns an unallocated task to each CPU 10 so that the CPU usage rate is 100% or less for each CPU 10 when unallocated tasks remain. Is provided.
At this time, the task scheduler 11 assigns an unassigned task to all the CPUs 10 including the selected CPU 10 and other CPUs 10.

In addition, the task scheduler 11 includes an upper limit allocation execution determination function for determining whether or not the usage rate upper limit allocation function for all CPUs has been executed, and the usage rate upper limit allocation function is performed for all CPUs 10. In addition, an upper limit unassigned task determination function for determining whether or not an unassigned task remains is provided.
Also in this case, the task scheduler 11 assigns unassigned tasks to all the CPUs 10 including the selected CPU 10 and other CPUs 10.

  Further, the task scheduler 11 has a predetermined time standby function for executing the target usage rate acquisition function after waiting for a preset specified time after all tasks are assigned to the CPU 10.

  The target usage rate storage unit 12 is based on the processing capacity, the type of each CPU 10 based on the processing application, and the upper limit of the usage rate at which each CPU performs an appropriate operation based on the characteristics of the cooling amount for each CPU 10 set in advance in the arithmetic unit. As a target usage rate storage function for storing a preset target usage rate in the target usage rate table 120.

Here, it is desirable to set the target usage rate of the CPU as high as possible as the processing capability of each CPU 10.
However, as shown in FIGS. 3 and 4, the power consumption and noise of a preset cooling device that performs the cooling operation for each CPU 10 increase as the CPU processing operation and the power consumption increase. Thus, the target usage rate is calculated based on a balance between the CPU usage rate, the power consumption of the cooling device, and the noise.

The task monitoring unit 13 calculates the ratio of the usage rate in each CPU 10 (hereinafter referred to as “CPU usage information”) when each task generated in the arithmetic device 1 is arithmetically processed. CPU usage information calculation management function for storing the obtained CPU usage information in the CPU usage information table 130 is provided.
The task monitoring unit 13 constantly monitors each generated task and has a function of updating the contents of the CPU usage information table 130.

[Description of Operation of First Embodiment]
Next, the overall operation of this embodiment will be described.

  First, a target usage rate preset as the upper limit of the usage rate of each central processing unit is acquired (target usage rate acquisition step), and usage rate information of each central processing processing unit for each task is acquired ( CPU usage information acquisition step), selecting a central processing unit that preferentially assigns the task (calculating unit selection step), so that the task approximates the target usage rate for the selected central processing unit (Usage approximate allocation step), preferentially assign the task to the selected central processing unit (task priority allocation step), and perform arithmetic processing on the assigned task (arithmetic processing step) .

  Here, with respect to the target usage rate acquisition step, CPU usage information acquisition step, calculation unit selection step, task priority allocation step, usage rate approximate allocation step, and calculation processing step, the execution contents are programmed and executed by a computer. You may comprise as follows.

  Next, the overall operation of this embodiment will be described in detail based on the flowchart of FIG.

First, the task scheduler 11 acquires the target usage rate of each CPU 10 from the target usage rate table 120 (step S101), and acquires the CPU usage information of each task from the task monitoring unit 13 (step S102).
Next, the priority CPU selection means 110 of the task scheduler 11 selects a CPU to which a task is preferentially assigned (step S103). Based on the selection result, the task scheduler 11 assigns tasks so that the total usage rate of each task of each CPU 10 falls within the acquired target usage rate (step S104).

  Here, the task scheduler 11 determines whether or not task allocation has been performed for all CPUs (step S105: target allocation execution determination function). For the CPU 10, the time scheduler 11 assigns tasks to the CPU 10 to which no task assignment has been made so that each CPU usage rate is equal to or less than the target usage rate (step S104).

  When task assignment to all the CPUs 10 is completed, the task scheduler 11 determines whether or not unassigned tasks remain later (step S106: unassigned task determination function), and unassigned tasks remain. If so, the time scheduler 11 assigns the remaining unassigned tasks to the CPUs 10 such that the CPU usage rate for each CPU 10 is 100% or less (step S107). In this case, task allocation may be performed exceeding the target usage rate.

  Here, the task scheduler 11 determines whether or not the usage rate upper limit allocation function for all CPUs has been executed (step S108), and when the equal task allocation process is performed for all CPUs 10, the task scheduler 11 is not allocated. It is determined whether or not the remaining task remains (step S109). If there is an unassigned task remaining, the task scheduler 11 applies to all the CPUs 10 including the selected CPU 10 and the other CPUs 10. An unassigned task is assigned (step S110).

  Next, each CPU 10 performs arithmetic processing on a task assigned to each CPU 10 by a preset method (step S111), and the task scheduler 11 waits for a preset specified time (step S112). Returning to step S101, the subsequent operations are repeated.

As described above, in this embodiment, the CPU target usage rate is set according to the CPU type and the CPU cooling characteristics of the arithmetic unit, the CPU usage rate required for task processing is brought close to the target usage rate, and the remaining By performing task scheduling so that the CPU usage rate is low, it is possible to achieve both low power consumption and low noise in the cooling device.

In this embodiment, since the usage rate of each CPU can be controlled so that the power efficiency is in a desirable state, the power efficiency is improved without changing the upper limit performance of the system (computing device 1). Can be reduced.
Furthermore, in this embodiment, the noise of the cooling device set for cooling the CPU can be suppressed to an appropriate level by keeping the CPU usage rate within the setting range.

  As described above, in the present embodiment, among a plurality of different CPUs 10 provided in the arithmetic unit 1, a high load CPU that operates near the target usage rate of each CPU 10 and other low load CPUs 10 are set. As a result, the CPU load is increased due to the processing load, and the power consumption of the entire CPU in the arithmetic unit 1 can be effectively suppressed, particularly when the CPU processing capacity is high and the speed is high.

In this embodiment, as shown in FIG. 6, in addition to the CPU type and the characteristics of the cooling device, it is registered in the target usage rate table 120 according to the purpose of use of the arithmetic device 1 and the target type of processing by the CPU 10. The internal environment setting unit 121 for changing the value may be provided in the target usage rate storage unit 12. In this case, other configurations are the same as those in the above-described embodiment.
As a result, a target usage rate table for efficiently operating the CPU 10 can be set according to the processing target of the arithmetic device 1, the amount of data to be used, and the processing amount.

Furthermore, in the present embodiment, as shown in FIG. 6, for example, the target is set according to the external environment setting of the computing device 1 such as whether the power supply is an external supply such as AC (Alternating Current) or an internal supply such as a battery. The external environment setting means 122 for changing the value registered in the usage rate table 120 may be provided in the target usage rate storage unit 12. In this case, other configurations are the same as those in the above-described embodiment.
Thereby, according to the power supply source, for example, the value registered in the target usage rate table 120 is changed, for example, the target usage rate of each CPU 10 is halved. Can be operated.

  The present invention can be applied to a mobile information processing apparatus that includes a plurality of central processing units and in which a multiprocessor system is introduced.

It is a schematic block diagram which shows one Embodiment of the information delivery system by this invention. It is explanatory drawing which shows the task allocation condition in the information delivery system disclosed in FIG. It is explanatory drawing which shows the relationship between CPU power consumption and CPU utilization in the information delivery system disclosed in FIG. It is explanatory drawing which shows the relationship between the power consumption and noise of a cooling device in the information delivery system disclosed in FIG. 1, and CPU usage rate. It is a flowchart which shows the whole operation | movement in the information delivery system disclosed by FIG. It is a schematic block diagram which shows one Embodiment of the information delivery system by this invention.

Explanation of symbols

1 Processing unit 10 Central processing unit (CPU)
DESCRIPTION OF SYMBOLS 11 Task scheduler 12 Target usage rate memory | storage part 13 Task monitoring part 110 CPU selection means 120 Target usage rate table 121 Internal environment setting means 122 External environment setting means 130 CPU usage information table

Claims (11)

A multi-processing system that includes a plurality of different central processing units and performs execution processing of different tasks set in advance,
A task monitoring unit that monitors usage rate information of each central processing unit for each task, and a calculation unit selection unit that selects a central processing unit that preferentially executes the execution process of each task based on the monitoring result; A multi-operation processing system comprising: a task assignment unit that preferentially assigns the task to each selected central processing unit. In the multi-operation processing system according to claim 1,
Each of the central processing units has a target usage rate storage unit that stores a target usage rate that is set in advance as an upper limit of the usage rate at which proper operation is performed, and the task allocation unit is configured to select each of the selected central processing units. A multi-operation processing system comprising a usage rate approximate allocation function that allocates the task so as to approximate the target usage rate. In the multi-operation processing system according to claim 2,
A target usage rate calculation setting unit that calculates and sets the target usage rate based on the type of the central processing unit and the amount of heat generated by the central processing unit is provided in the target usage rate storage unit. Multi arithmetic processing system. In the multi-operation processing system according to claim 3,
The multi-operation processing system, wherein the target usage rate calculation setting means includes a task type setting function for setting the target usage rate based on a processing amount of the task assigned to the central processing unit. In the multi-operation processing system according to claim 3,
The target usage rate calculation setting means includes a task type setting function for setting the target usage rate based on a preset type of a power supply source that supplies power to the central processing unit. Multi-processing system. In the multi-operation processing system according to claim 3,
When the task is allocated to all the central processing units and there is an unallocated task that is not allocated, the task allocation unit is configured so that the usage rate of each central processing unit does not exceed 100%. A multi-operation processing system comprising a task equal assignment function for assigning the unassigned task to each central processing section. In the multi-operation processing system according to claim 6,
A power saving mode setting unit that sets the operation of each central processing unit below a preset usage rate to a power saving mode when all the tasks are assigned is provided in the task assignment unit. Multi arithmetic processing system. A plurality of different central processing units, a task assignment unit that allocates a plurality of different tasks to each central processing unit, a task monitoring unit that monitors the usage rate of each central processing unit for each task, In a multi-operation processing method that includes a calculation unit selection unit that selects a central processing unit that preferentially assigns the task, and performs the execution process of the task,
CPU usage information acquisition step for acquiring usage rate information of each central processing arithmetic unit for each task, arithmetic unit selection step for selecting a central arithmetic processing unit to which the task is preferentially assigned, and the selected central arithmetic operation A multi-operation processing method, comprising: a task priority assignment step for preferentially assigning the task to a processing unit; and an arithmetic processing step for performing arithmetic processing on the assigned task. In the multi-operation processing method according to claim 8,
A target usage rate acquisition step for acquiring a target usage rate preset as an upper limit of the usage rate of each central processing unit is provided before the CPU usage information acquisition step, and for the selected central processing unit A multi-calculation processing method comprising: a usage rate approximate allocation step for allocating the task so as to approximate the target usage rate before the task priority allocation step. A plurality of different central processing units, a task assignment unit that allocates a plurality of different tasks to each central processing unit, a task monitoring unit that monitors the usage rate of each central processing unit for each task, In a multi-operation program for performing execution processing of the task, comprising a calculation unit selection unit that selects a central processing unit that preferentially assigns the task,
CPU usage information acquisition function for acquiring usage information of each central processing arithmetic unit for each task; arithmetic unit selecting function for selecting a central arithmetic processing unit to which the task is preferentially assigned; and the selected central arithmetic A multi-operation processing program that causes a computer to execute a task priority assignment function that preferentially assigns the task to a processing unit and an arithmetic processing function that performs arithmetic processing on the assigned task. In the multi-operation processing program according to claim 10,
A target usage rate acquisition function for acquiring a target usage rate preset as an upper limit of the usage rate of each central processing unit, and the task approximates the target usage rate for the selected central processing unit A multi-operation processing program that causes the computer to execute a usage rate approximate assignment function to be assigned in this manner.

Images