JP2000066910A - Task scheduling method and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To minimize the consumption of a secondary battery by dividing from an inserted task to a task of an execution sequence which will not be able to keep a deadline into another task groups and merging two groups into one group. SOLUTION: A scheduler 112 calculates the deadline of a task Tk and inserts a data structure for executing the task Tk to an appropriate position of a management queue 113. As the result of inserting the task Tk, the execution frequency (f) of tasks to be increased is calculated. When tasks in the same group are carried out with the execution frequency (f), the first task that does not keep the deadline of the task is defined as Tp. Division is made with from the Tk to the Tp as one task group. When the execution frequency of the former divided group is smaller than that of the latter group, the two groups are merged into one group to balance the execution frequencies of the former and the latter groups so that execution is performed with the same execution frequency.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a task scheduling method for minimizing secondary battery consumption in a real-time system.

[0002]

2. Description of the Related Art An example of a conventional system for power saving is a Proceed issued from IEEE in 1995.
edings of First Symposium
onOperating Systems Desi
13th of gn and Implementation
Scheduling published on pages 23 to 23
For Reduced CPU Energy in a paper by Mark Weiser.

The conventional system described in this paper is
It consists of a PU, a memory, a clock generator, a frequency control circuit, and a disk. The general program is
Stored on disk. The memory includes means for controlling the schedule of the execution of the general program. In this paper, we evaluate three proposed energy-saving scheduling algorithms by simulation using the execution trace data of a general program. 1. unbound-delay perfect-f
result (OPT): Operate at the minimum speed so as to eliminate idle time over the entire trace data. However, this is unrealistic because it requires knowledge of future operations during execution. The response time is also undesirably increased. 2. bounded-delay limited-f
uture (FUTURE): Operates at the minimum speed so as to eliminate idle time within a certain time (called a window). This is unrealistic because it requires knowledge of future operations at runtime. The response time is desirably within the window time. 3. bounded-delay limited-p
ast (PAST): Determines the CPU speed assuming that the operation in the immediately preceding window is similar to the next one. The evaluation result depends on the adjustment interval. adjustment interval
PAST, FUTURE becomes OP
Approach T. PAST is better than FUTURE (since the access cycle can be carried over to the next interval).

[0004] Japanese Patent Application Laid-Open No. 8-6681 discloses that
As a "power saving control system", in a multiprocessor system having a plurality of CPUs, a processor bus monitoring unit that detects an operation state of each CPU by monitoring a processor bus, and a system state monitor that monitors a system load state Unit, and a system state control unit that controls power consumption by each CPU based on a notification from the system state monitoring unit, and when a state where the load on a specific CPU is small due to waiting for key input or the like continues, A system in which a system state control unit switches a clock supplied to the CPU to a low frequency is disclosed.

[0005] Further, Japanese Patent Application Laid-Open No. 8-6803 discloses that
As an "information processing method and device thereof", an information processing method of multitask execution with minimum power consumption and a device thereof are disclosed. This is because the hardware resources acquired by the task and the power consumption thereof are detected, the execution priority of the task having the large total power of the acquired hardware resources is set high, and the scheduler of the operating system executes the high execution. This is a method of selecting tasks of priority and giving execution rights.

[0006] Also, Japanese Patent Application Laid-Open No. 8-50551 discloses that
As a `` real-time application task scheduling and processing system '', regarding the execution scheduling of software tasks, in particular, regarding an online task scheduler for a real-time system, check whether each task request satisfies its deadline when executing a dispatcher, As if arranged to end as close as possible to the deadline, C
It is disclosed to schedule tasks in the database based on the PU's late busy period.

[0007]

However, this prior art has the following problems.

The first problem is that no consideration is given to a real-time system, and a deadline time at which execution of a program must be completed may not be satisfied.

The reason is that the conventional technology is intended for a time-sharing type system such as UNIX, and does not take deadline time in a real-time system into consideration.

The second problem is that power saving including secondary batteries is not achieved. The reason is that the existence of the secondary battery is not considered.

Further, even in the conventional examples described in the above publications, there is no disclosure of a task scheduling method for minimizing secondary battery consumption in a real-time system.

[Object of the Invention] An object of the present invention is to provide a CPU,
In a real-time system having a program memory, a clock generator having a frequency control function, and a secondary battery, the power consumption of the secondary battery is proportional to the α-th power (α> 1) of the frequency supplied to the CPU and contributing to program execution. By taking advantage of this fact, a task scheduling method for minimizing secondary battery consumption in a real-time system is performed by keeping the deadline of the tasks constituting the real-time system and executing the task at a constant frequency as low as possible. To provide.

[0013]

According to the present invention, as a means for solving the above-mentioned problems, in a method for scheduling a plurality of tasks for performing real-time processing, tasks operating at the same frequency are grouped into one task group.
Each task in the task group is arranged in the order of deadline for each task, and the execution end time of tasks other than the last task in the task group is before the deadline of the task, and The execution end time of a task has a configuration equal to the deadline of the task, and when a new task is inserted into a task group executed until the deadline of real-time processing, the The task from the inserted task to the task in the execution order that cannot keep the deadline is divided into another task group, and the divided task is executed. If the execution frequency is lower than the execution frequency, The averaged, in order to run in the same execution frequency, to merge the two task groups to one task group, that is to provide a task scheduling method comprising.

The CPU 100, the memory 110,
In a real-time system 001 having a secondary battery 120, a clock generator 130, and a frequency control circuit 140 for controlling the frequency of a clock supplied to the CPU, the memory 110 has a plurality of tasks for performing real-time processing. A scheduler 112 for scheduling and controlling the execution of the task; and an execution management queue 113 for the task. The scheduler 112 calculates a deadline of the task Tk for processing the event in the real-time system 001 (A1). When,
A step (A2, A3) of inserting execution queue data for executing the task Tk from the deadline into an appropriate position of the execution management queue into a task group, and a task to be added as a result of the insertion of the task Tk Calculating the quantity w and the task execution frequency f (A4);
Task T (j, 1) in the task group (TGj)
To T (j, nj) at the execution frequency f, it is checked whether or not the deadline of the task can be maintained. The task Tk or the last task T in the immediately preceding task group is checked.
Step (A5) of finding the first task Tp that cannot keep the deadline from (j, m) to T (j, nj)
And the number of tasks in the task group TGj into which the task Tk is inserted is nj, the task T (j, 1)
In order to divide the task group so that the deadline of the tasks up to (j, nj) is protected, the tasks Tk or T (j, n) to Tp are divided into one task group, and necessary data is divided into the tasks. If the step (A7) for setting the group management data and the execution frequency of TGj is not smaller than the execution frequency of TGj + 1 in the adjacent task groups TGj and TGj + 1, the above steps (A5, A7) are repeated until the task T (j, nj). Steps (A6, A9) to be performed and the adjacent task group TGj
And TGj + 1, the execution frequency of TGj <TGj +
In the case of the execution frequency of 1, task groups TGj and TG
averaging the execution frequency of j + 1 and merging the two task groups into one task group to execute at the same execution frequency (A10); Investigating whether the task groups of the tasks to be executed are different (S4
1) and, if not, a step of setting the execution frequency of the next task group in the frequency control circuit 140 (S4).
2) and a step (S43) of removing the first task included in the first task group of the execution management queue from the queue and starting the execution of the task to be executed next (S43). It is also a scheduling method.

The execution management queue 113 is arranged in the order of the execution end time for each task group executed at the same frequency, and the execution start time,
Task group management data consisting of task group execution end time and task group execution frequency, and task task management data consisting of task execution start time, execution end time, deadline, and workload , The task scheduling method.

Further, in a real-time system in which the frequency of the CPU is controlled to a fixed multiple, a task scheduling method is characterized in that the method further includes a step of finding the frequency of the fixed multiple that is the closest to the calculated frequency.

[0017] A task scheduling method is characterized in that a secondary battery is used as a power supply.

Further, there is provided a recording medium on which a program for causing a computer to execute the above-described task scheduling method is recorded.

[Operation] According to the present invention, the CPU frequency is calculated so that the task can be completed by the end of the deadline and to have the minimum frequency. By setting the execution start time and the execution end time and scheduling the tasks, the deadline of the tasks constituting the real-time system can be protected, and the CPU can be executed at a constant frequency as low as possible. As a result, the consumption of the secondary battery can be minimized.

The point is to utilize the point that the consumption of the secondary battery can be minimized by executing the CPU at a constant frequency as low as possible.

A plurality of tasks operating at the same frequency for real-time processing are set as one task group, and when a new task is inserted, the inserted task is included in the task group into which the new task is inserted. From to the task in the execution order that can not keep the deadline of the real-time processing is divided into another task group,
If the execution frequency of the previous task group is smaller than the execution frequency of the subsequent task group in the divided adjacent task groups, the two task groups are merged into one task group, and the previous task group is merged. By averaging the execution frequency of the group and the later task group,
By executing at the same execution frequency, it is possible to keep the deadline of the tasks constituting the real-time system and execute the task at a constant frequency as low as possible.

The power consumption of the secondary battery is determined by the power of the frequency supplied to the CPU and contributing to the execution of the program.
Since it is proportional to 1), it is possible to minimize the consumption of the secondary battery by keeping the frequency of the CPU low and constant.

As described above, according to the present invention, in a real-time system, (1) schedule a task so as to minimize power consumption; (2) observe a time at which the task must end its execution. Scheduling;

According to the present invention, a program is read from a CD-ROM, a floppy disk, or the like as a recording medium storing a computer program describing the above-described method into a memory of the computer, and the program is read by the program. The present invention can be realized by controlling the CPU.

[0025]

DESCRIPTION OF THE PREFERRED EMBODIMENTS [First Embodiment] [Description of Configuration] FIG. 1 is a block diagram showing a configuration of an embodiment of the present invention. Referring to FIG. 1, an embodiment of the task scheduling scheme of the present invention operates under program control. The real-time system 001 is a CPU
100, a memory 110, a secondary battery 120, a clock generator 130, and a frequency control circuit 140.

The memory 110 has n task means 111-i (i = 1, 1) for performing real-time processing.
2,. . n), means 112 for controlling the schedule of the execution of the task, and data 113 for processing the means 112. Means 111 as a task, means 112
Is called a scheduler, and the data 113 is called an execution management queue. In addition, the frequency control circuit 140
00 can be controlled.

The outline of the functions of these means is as follows.
The task 111-i (i = 1, 2,... N) is activated in response to an event i coming from outside the real-time system 001 and registered in the scheduler 112.
The task 111 must complete processing within a certain period of time for an event coming from outside, and the time at which this processing must be terminated is called a deadline. The present invention is characterized in that the scheduler 112 performs scheduling so that the task is executed at a constant frequency as low as possible while observing the deadline (time at which the execution must be completed) of the task.

FIG. 2 shows an example of the execution management queue 113 which is a data structure used in this method. In the task group, the tasks are executed at the same frequency and are arranged in the order of the execution end time. The data that manages each task group includes the task group's execution start time (which is equal to the first task's execution start time), task group's execution end time (which is equal to the last task's execution end time), and task It consists of the execution frequency in the group. Further, the task groups are arranged in the order of the execution end times of the task groups.

The data for managing the execution of each task includes a task execution start time, a task execution end time, a deadline, and a workload. The execution end time of the last task in the task group is equal to the deadline, but the execution end times of the other tasks are earlier than the deadline.

[Explanation of Operation] Next, the overall operation of the present embodiment will be described in detail with reference to FIGS.

FIGS. 3 to 5 are flowcharts showing the processing of the scheduler 112.

First, the processing when the event k arrives at the real-time system will be described with reference to FIGS.

The task Tk that processes this event is composed of tasks 111-1 to 111 -n included in the memory 210.
And the deadline of task Tk is calculated. The deadline is calculated by, for example, assuming that the arrival time of the event is a and the processing of the event must be completed within t hours, and the time d at which the processing of the event must be completed is calculated as d = a + t. (Step A1 in FIG. 3).

Next, a data structure for executing the task Tk is inserted from the deadline of the task Tk into an appropriate position in the execution management queue (for example, the position shown in FIG. 2) (steps A2 and A3 in FIG. 3). .

Next, as a result of the insertion of the task Tk, the work amount w of the task and the execution frequency f of the task, which increase, are calculated (step A4 in FIG. 3).

Tasks T (j, 1) to T in the same group
When executing up to (j, nj) at the execution frequency f, it is checked whether or not the deadline of the task can be maintained, and Tk or the last task T (j, m) of the immediately preceding task group is checked.
Tp is the first task that cannot keep the deadline among tasks from T to T (j, nj) (step A in FIG. 4).
5).

Now, it is assumed that the task group TGj is inserted between the i-th task T (j, i) and the task T (j, i + 1). Let n be the number of tasks in task group TGj
Assuming that j, the task group is divided so that the deadline of the tasks from T (j, 1) to T (j, nj) is protected, so that one task from Tk or T (j, m) to Tp is performed. The data is divided into groups, and necessary data is set in the task group management data (step A7 in FIG. 4).

Adjacent task groups TGj and TGj +
1, if the execution frequency of TGj is not smaller than the execution frequency of TGj + 1 (A8), the above steps (A5, A5
7) is performed up to task T (j, nj) (A6, A9).

Adjacent task groups TGj and TGj +
1, when the execution frequency of TGj <the execution frequency of TGj + 1 (A8), the task groups TGj and TGj
The execution frequency of +1 is averaged, and the execution frequencies of the two task groups are averaged and merged into one task group to execute at the same execution frequency (A10).

Next, the processing in the case where the execution of the currently executing task has been completed will be described with reference to FIG.

The scheduler means 112 first checks whether the task group of the task to be executed next is different (S41), and if it is different, sets the execution frequency of the next task group in the frequency control circuit 140 (S42). Next, the first task included in the first task group in the execution management queue is removed from the queue, preparations are made to start execution as the next task to be executed, and the execution of the task is started (S43).

[Explanation of Specific Example] In the task group, the tasks are executed at the same frequency. Also, they are arranged in the order of the execution end time. The data that manages each task group includes the task group's execution start time (which is equal to the first task's execution start time), task group's execution end time (which is equal to the last task's execution end time), and task It consists of the execution frequency in the group.

As an example, assume that there is one task group 1 as shown in FIG. 6, and that the task group 1 includes tasks 1, 2, and 3.

The execution start time of task 1 is 100
Seconds, the end time is 110 seconds, the deadline is 113 seconds, and the workload is 250. The execution start time of task 2 is 110 seconds, the end time is 115 seconds, and the deadline is 116.
Seconds, the workload is 125, the execution start time of task 3 is 115 seconds, the end time is 130 seconds, and the deadline is 1
Assume that the work amount is 375 for 30 seconds.

The execution start time of task group 1 is 10
0 seconds, execution end time is 130 seconds, execution frequency is 25 MH
z.

The sum of the work amounts is 750. Further, the task groups are arranged in the order of the execution end times of the task groups. The data for managing the execution of each task includes a task execution start time, a task execution end time, a deadline, and a workload.

The execution end time of the last task in the task group is equal to the deadline, but the execution end time of the other tasks is earlier than the deadline.

FIGS. 3 to 5 are flowcharts showing the processing of the means 112.

Next, a more specific operation of the present embodiment will be described in detail with reference to the configuration diagram of FIG. 1, the flowcharts of FIGS. 3 to 5, and the data structures of FIGS. The numerical values in the text are simplified for explanation.

Now, the processing when event 4 arrives at the real-time system at 90 seconds will be described with reference to FIG. Assume that this event requires a response within 25 seconds. Task 4 that processes this event is found, and the deadline of task 4 is calculated as 90, 25 = 115 seconds (step A1 in FIG. 3).

The work amount of task 4 is assumed to be 100. Next, a data structure for executing the task 4 is inserted at an appropriate position in the execution management queue from the deadline of the task 4.
Now, it is assumed that a task group 1 is inserted between task 1 and task 2 (step A2, step A3).

Since the number of tasks in the task group 1 is 3, the task group is divided so that the deadline of the tasks from the task 1 to the task 3 is protected (steps A4 to A7). In step A4, as a result of the task 4 being inserted, the work amount w of the task and the execution frequency f of the task, which increase, are calculated (step A4). That is,
The frequency f is assumed to be the workload w = 100, the frequency f = the workload (750 + 100) / the execution time (130-100)
= 28.33 MHz.

When the tasks from task 1 to task 3 are executed at the execution frequency f = 28.33 MHz, it is checked whether the deadline of the task can be maintained.

The deadline of task 1 is 100 (second)
+ 250 / 28.33 = 108.82 sec; Task 4 deadline is 108.82 (sec) + 100 / 28.3
3 = 112.29 seconds; Task 2 ted line is 11
2.29 (sec) + 125 / 28.33 = 116.70
Seconds; Task 3 deadline is 116.70 (seconds) +
375 / 28.33 = 130 seconds.

The first task that cannot keep the deadline among tasks 1 to 4 is task 2 (step A5).

Therefore, as shown in FIG.
4 and 2 are divided into one task group 1 ', and necessary data is set in the task group management data (step A7 in FIG. 4).

As shown in FIG. 7, the information set for the newly created task group 1 'is calculated as follows. The execution start time of task group 1 'is 100 seconds, the execution end time is 116 seconds (= deadline of task 2),
The execution frequency is: work amount (250 + 100 + 125) / execution time (116-100) = 29.69 MHz.
The sum of the workload is 475.

From the above, the execution start time of task 1 is 100 seconds, and the end time is 100 + 250 / 29.69 =
108.42 seconds, the execution start time of task 4 is 10
8.42 seconds, end time is 108.42 + 100/29.
69 = 111.79 seconds, the execution start time of task 2 is 111.79 seconds, and the end time is 111.79 + 125 /
29.69 = 116 seconds, and tasks 1, 4, and 2 end before the deadline of each task.

When the execution frequency of 2 ′ is higher than the execution frequency of 1 ′ in the adjacent task groups 1 ′ and 2 ′ as a result of the execution of step A4, the execution frequencies of 1 ′ and 2 ′ are changed. Averaging is performed so as to execute at the same execution frequency. For this, the task groups are merged (step A10).

In this example, since the frequency of the task group 1 'is 29.69 MHz and the frequency of the task group 2' is 25 MHz (shown in FIG. 6), step A10 is not performed, and it is unnecessary to return from step A9 to A5. It is.

The remaining task 3 is newly set as a task group 2 '. The execution start time of task group 2 ′ is 116 seconds, the execution end time is 130 seconds (= deadline of task 3), and the execution frequency is 250/14 = 17.8.
It becomes 6. The sum of the workload is 250.

Next, the processing in the case where the execution of the currently executing task has been completed will be described with reference to FIG. Now, it is assumed that the task group is changed and processing is performed from the beginning of the task group 1 ′. The scheduler means 112 first
It is checked whether the task group of the next task to be executed is different. If the task group is different, the execution frequency of the next task group is set to 31.67 MHz in the frequency control circuit 140. Next, the first task included in the first task group in the execution management queue is removed from the queue, and preparations are made to start execution as the next task to be executed, and the execution of the task is started.

[Embodiment 2] In the frequency control circuit in the real-time system, the frequency of the CPU is 1/2 times,
1/4 times. . The process of the scheduler of the real-time system in the case where the control can be performed only at a fixed number of times as described above will be described.

This embodiment differs from FIGS. 3 and 4 of the first embodiment only in that a function F is included in the setting of the execution frequency in steps A7 and A10 in FIG. Is exactly the same as in the first embodiment.

That is, in the present embodiment, only the following parts are different. (A7 '):: TGj' execution frequency: = F ((TGj 'execution end time-TG
: (A10 '):: TGj execution frequency: = F ((TGj execution end time-TGj execution start time) / TGj work); : function F (x) is a function for obtaining the fixed multiple of the frequency of the nearest CPU above the value of the argument x, the fixed multiple of the frequency with the CPU _{is, f 1, f 2, ...} f n (f ₁ <f ₂ <
... <When a f _n), the arguments x of the function F (x), is a function that returns the _{f i-1 <x <f} i (2 ≦ i <n) and becomes f _i.

Then, by inputting the frequency calculated in step A4 as the argument x to the function F (X), the closest multiple of the calculated frequency that is equal to or higher than the calculated frequency can be obtained.

The present invention is also a recording medium such as a CD-ROM in which a computer program describing the above-described method is stored. By installing the program from the recording medium into a computer, the present invention can be easily implemented. It can be implemented.

The text of each step in the flowcharts of FIGS. 3 and 4 will be described below. (A1): Find task Tk corresponding to event k,
Calculate the deadline time of the task; (A2): Referring to the data structure of the execution management queue, the task group that satisfies the task group execution start time <task Tk deadline ≦ task group execution end time
Find TGj; (A3): Find T (j, i) that is deadline of task T (j, i) <deadline of Tk <T (j, i + 1), m: = i (A4): Calculate the sum W of the workload from the task T (j, 1) to the task T (j, nj); the frequency f required for the execution of the workload W:
= W / (execution end time of task T (j, nj)-task T (j, 1)
(A5): The first task that cannot keep the deadline when trying to execute T (j, nj) from task T (j, m) at execution frequency f is T (j, p). (A6): p ≦ nj? (A7): Divide task T (j, p) from task T (j, m) as task group TGj ′; Execution start time of TGj ′: = execution start time of T (j, m); TGj ′ Execution end time: = Debt line of T (j, p); Work amount of TGj ′: = sum of work amounts from T (j, m) to T (j, p); Execution frequency of TGj ′: = (TGj 'execution end time-TGj
'Execution start time of') / Work of TGj '; Tasks belonging to TGj' are newly added to T (j ', i) (i = 1, ...
nj ') and renumber; T (j', i) (i = 1,... nj ') execution start time: = T (j
', i-1) execution end time; T (j', i) (i = 1,... nj ') execution end time: = T (j
', i) execution start time + work of T (j', i) / TGj '
(A8): Is the execution frequency of the task group TGj <the execution frequency of the task group TGj + 1? (A9): m: = p + 1; (A10): task group TGj and task group TGj +
1 is merged into the task group TGj; TGj execution start time: = TGj execution start time; TGj execution end time: = TGj + 1 execution end time; TGj work amount: = TGj work amount + TGj + 1 TGj execution frequency: = (TGj execution end time−TGj execution start time) / TGj work amount; new tasks belonging to TGj are T (j, i) (i = 1,... Nj) T (j, i) (i = 1,... Nj) execution start time: = T (j, i-1) execution end time; T (j, i) (i = 1, ... Execution end time of nj): = execution start time of T (j, i) + work load of Tj ′, i / execution frequency of TGj;

[0069]

The first effect is that, in a real-time system having a secondary battery, when a task is executed, the consumption of the secondary battery can be minimized.

The reason is that, according to the present scheduling method, a plurality of tasks operating at the same frequency for real-time processing are made into one task group, and when a new task is inserted, the new task is inserted. In the divided task group, the tasks in the execution order that cannot keep the deadline of the real-time processing are divided into different task groups, and the divided task groups are divided into adjacent task groups and the execution frequency of the previous task group is changed to the later task group. If the execution frequency is smaller than the execution frequency, the two task groups are merged into one task group,
By averaging the execution frequency of the previous task group and the subsequent task group and executing at the same execution frequency,
This is because the task can be executed at the same low frequency as much as possible. Rechargeable battery consumption is CPU
When the task is executed as described above at the frequency α (1), the consumption of the secondary battery is minimized.

The second effect is that the execution of the task can be completed by the deadline time at which the execution of the task must be completed by this scheduling method.

The reason is that task scheduling is performed so that execution is completed by the deadline time when execution is performed at the same frequency as much as possible.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a first exemplary embodiment of the present invention.

FIG. 2 is a configuration diagram of an execution management queue according to the first embodiment of the present invention.

FIG. 3 is a flowchart illustrating an operation of the scheduler 112 according to the first embodiment of the present invention.

FIG. 4 is a flowchart illustrating an operation of the scheduler 112 according to the first embodiment of the present invention.

FIG. 5 is a flowchart illustrating an operation of a scheduler according to the first and second embodiments of the present invention.

FIG. 6 is a queue data structure for explaining the operation of the scheduler according to the first embodiment of the present invention.

FIG. 7 is a queue data structure for explaining the operation of the scheduler according to the first embodiment of the present invention.

[Explanation of symbols]

 001 Real-time system 100 CPU 110 Memory 120 Secondary battery 140 Frequency control circuit 130 Clock generator

Claims (6)

[Claims] 1. A scheduling method for a plurality of tasks for performing real-time processing, wherein tasks operating at the same frequency are set as one task group, and the tasks in the task group are arranged in a deadline order for each task; The execution end time of tasks other than the last task in the task group is before the deadline of the task, and the execution end time of the last task in the task group has a configuration equal to the deadline of the task. When inserting a new task into a task group that is executed before the deadline of real-time processing, in the task group into which the new task is inserted, the tasks from the inserted task to the tasks in the execution order in which the deadline cannot be protected can be performed. Divided into another task group, the divided task group In the step, when the execution frequency of the previous task group is lower than the execution frequency of the subsequent task group, the execution frequency of the previous task group and the execution frequency of the subsequent task group are averaged to execute the same execution frequency. A task scheduling method, wherein two task groups are merged into one task group. 2. A CPU (100) and a memory (110).
, A secondary battery (120) and a clock generator (1)
30) and in the real-time system (001) having a frequency control circuit (140) for controlling the frequency of a clock supplied to the CPU, the memory (110) has a plurality of tasks for performing real-time processing; And a task execution management queue (113). The scheduler (112) includes a deadline of a task Tk that processes an event in the real-time system (001). (A2), inserting the execution queue data for executing the task Tk from the deadline at an appropriate position in the execution management queue into the task group (A2, A3), As a result of the insertion, the workload w of the task and the execution frequency f of the task are increased. A calculating step (A4); and a task T (j, 1) in the task group (TGj).
To T (j, nj) at the execution frequency f, it is checked whether or not the deadline of the task can be maintained. The task Tk or the last task T in the immediately preceding task group is checked.
Step (A5) of finding the first task Tp that cannot keep the deadline from (j, m) to T (j, nj)
When the number of tasks in the task group TGj into which the task Tk is inserted is nj, the task T (j, 1)
In order to divide the task group so that the deadline of the tasks up to (j, nj) is protected, the tasks Tk or T (j, n) to Tp are divided into one task group, and necessary data is divided into the tasks. In the step (A7) of setting the group management data, in the adjacent task groups TGj and TGj + 1,
If the execution frequency of TGj is not smaller than the execution frequency of TGj + 1, the above-described steps (A5, A7) are performed in the task T (j, n).
j) to steps (A6, A9), and adjacent task groups TGj and TGj + 1,
When the execution frequency of TGj <the execution frequency of TGj + 1, a step of averaging the execution frequencies of the task groups TGj and TGj + 1 and merging the two task groups into one task group in order to execute at the same execution frequency (A10). ), And when the execution of the task being executed is completed, a step (S41) of checking whether the task group of the next task to be executed is different. If the task group is different, the execution frequency of the next task group is controlled. A step (S42) of setting the circuit 140 and a step (S43) of removing the head task included in the head task group of the execution management queue from the queue and starting the execution of the task to be executed next. A task scheduling method comprising: 3. The execution management queue (113) is arranged in the order of execution end time for each task group executed at the same frequency, and includes an execution start time of the task group, an execution end time of the task group, and a task group. The management data of the task group consisting of the execution frequency and the task execution start time, execution end time, deadline,
3. The task scheduling method according to claim 2, further comprising management data of each task composed of a workload. 4. The task scheduling according to claim 2, wherein in a real-time system in which the frequency of the CPU is controlled to be a fixed multiple, a step of obtaining the closest fixed multiple that is equal to or higher than the calculated frequency. Method. 5. The task scheduling method according to claim 1, wherein a secondary battery is used as a power supply. 6. A recording medium on which a program for causing a computer to execute the method according to claim 1 is recorded.