JPH08314740A - Process dispatch method - Google Patents

Abstract

PURPOSE: To suppress the increase of overhead which is caused by occurrence of a prescribed event that should start a real-time process. CONSTITUTION: As shown in (C), a CPU 1 carries out a time division process A and at the same time a CPU 2 carries out a time division process B for a time longer than a prescribed carry-over boundary time T2. If such an event that should start a real-time process occurs under such conditions, the execution of the process B is discontinued and the real-time process is started. When the real-time process is over, a time dividion process C that should be carried out after the process B is carried out by the CPU 2. The shortage X of the time slice time of the process B is carried over to the next time. When the process B is started as shown in (2) of (C), the process B is carried out for a time equivalent to the sum of the shortage X and a time slice time T1 of the process B. Thereby, the process dispatch frequency can be reduced compared with the conventional system that is shown in (B).

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a process dispatch method, and more particularly to a process dispatch method for executing a process to be executed when a predetermined event occurs and a process to be executed periodically.

[0002]

2. Description of the Related Art In an information processing apparatus, when a CPU is sequentially assigned to a plurality of processes to perform multitask processing, an event occurrence wait (for example, data input) is executed in each process execution. Process that starts and executes each event when an event occurs, depending on whether the number of times that it waits for input of an interrupt signal due to completion of output of data or data output, waiting for release of specific resource, etc. is large or small. Time process) and a process that is periodically executed for a predetermined time (hereinafter referred to as a time-sharing process), and the execution priority is set for each process Switching the process to which the CPU is allocated according to the CPU usage time of the inside process, the occurrence of events, the priority of each process, etc. Dispatch system are known.

In the above method, when an event for activating a real-time process has not occurred, a plurality of time-division processes are sequentially executed as shown in FIG. 14A as an example. In FIG. 14A, time division processes A to C are shown.
An example of an operation when executing (the priority is process C> process B> process A) is shown. When the process C is executed for a predetermined time, the process B is activated (dispatched) (see FIG. 14A). ), When the process B is executed for a predetermined time, the process A is activated (in FIG. 14A), and when the process A is executed for a predetermined time, the process C is activated again (in FIG. 14A).

As is apparent from FIG. 14A, the time for allocating the CPU to each process (the execution time of each process) is set to be longer as the process priority increases. Unless any of the processes spontaneously waits for the occurrence of an event or the like, the processes are repeatedly executed in the order of process C → B → A for a time corresponding to the priority of each process.

Further, when an event for activating a real-time process occurs, the real-time process is executed as shown in FIG. 14B as an example. FIG. 14B shows an example of the operation when executing the real-time processes D to F (the priority is process F> process E> process D), and the process E is started during the execution of the process D. As soon as the event 1 to be generated occurs, the process E is started (in FIG. 14B). If the process E does not make a request to start another process or set an event, the process E is started.
When the device spontaneously shifts to the waiting state, the process (process D in FIG. 14B) that was being executed immediately before is restarted (in FIG. 14B). Similarly, when the event 2 that prompts the activation of the process E and the event 3 that prompts the activation of the process F occur consecutively (see FIG. 14 (B),
), Each time an event occurs, the running process is immediately interrupted and the corresponding process is started.

[0006] The real-time process ensures high responsiveness to the occurrence of an event, so that instead of being activated immediately when a corresponding event occurs, the real-time process uses the CPU so as not to occupy the CPU for a long time. Time is typically limited. Also in FIG. 14B,
The real-time process started by the occurrence of the event will run for a relatively short time even if there is no event that should start the other process, then voluntarily shift to the waiting state, and the other process is started. (FIG. 14 (B) and).

FIG. 14C shows an example of the operation when the time division process and the real time process are mixed. In the example shown in FIG. 14C, the time-division process C is executed for a predetermined time to activate the time-division process B (in FIG. 14C), and then the real-time process is performed after a relatively short time has elapsed. Event 4 that should activate the time process E has occurred (FIG. 14C), and the occurrence of this event 4 interrupts the execution of the time-sharing process B and immediately activates the real-time process E. The time-divisional process B is restarted when the real-time process E voluntarily shifts to the waiting state (in FIG. 14C), and the CPU usage time before execution suspension (time in FIG. 14C). And C after resuming execution
When the total of the PU usage time (time after FIG. 14C) reaches a predetermined time, the time division process C to be executed next is activated (in FIG. 14C).

By the way, when switching the process to which the CPU is assigned, the CPU is released from the running process to the OS (operating system), the OS switches the running process to the queue, and the process to be executed next. Is performed, and after the judged process is taken out from the queue or the like, the process is realized by a series of processing such as yielding the CPU from the OS to the process taken out from the queue. This process switching is an indispensable process when multi-tasking is performed, but it is not an essential process to be executed by the CPU like process execution, but an overhead of the information processing device. In order to improve the substantial processing capability of the information processing device, it is desirable to reduce the overhead as much as possible.

In FIGS. 14A to 14C described above,
The timing at which the process is switched is indicated by circled numbers and arrows, but as is clear from FIG. 14,
In the dispatch method described above, when an event occurs during execution of the time-sharing process and the real-time process needs to be activated (FIG. 14C), only the time-sharing process is executed and the real-time process is executed. There is a problem in that the number of process switchings increases, the overhead increases, and the substantial processing capability of the information processing apparatus decreases as compared with the case where only the information processing apparatus is executed.

Further, in connection with the above, Japanese Patent Laid-Open No. 2-156337 discloses that in order to set an appropriate priority for each process,
A technique is disclosed in which the priority of each process is determined in consideration of the degree of response request and the degree of I / O request using a fuzzy function. Further, Japanese Patent Laid-Open No. 6-12263 discloses a method of dynamically changing the priority of processes in consideration of the CPU usage rate and the I / O usage rate in order to improve the throughput of the entire device. There is. However, none of the techniques described in any of the above publications considers the above-mentioned problem, and it is impossible to avoid an increase in overhead when an event that should activate a real-time process occurs during execution of a time-sharing process.

The present invention has been made in consideration of the above facts, and an object thereof is to obtain a process dispatch method capable of suppressing an increase in overhead due to occurrence of a predetermined event for activating a real-time process.

[0012]

In order to achieve the above object, the invention according to claim 1 is to provide a plurality of time division processes to be executed periodically and a real time process to be executed when a predetermined event occurs. In each of the above, for each of the plurality of time-sharing processes, a normal execution time and a carry-forward boundary time shorter than the normal execution time are set in advance,
When the predetermined event has not occurred, each of the plurality of time-sharing processes is sequentially and repeatedly executed by the normal execution time of each time-sharing process, and the real-time process is activated when the predetermined event occurs. In order to suspend the execution of the predetermined time-division process being executed for, if the elapsed time from the start of execution of the predetermined time-division process is equal to or more than the carry-over boundary time of the predetermined time-division process, When the execution of the predetermined time-sharing process is interrupted, the real-time process is started, and when the time-sharing process is ready for execution, another time-sharing process to be executed subsequent to the predetermined time-sharing process is executed. Along with the execution, the shortage of the normal execution time of the predetermined time division process caused by the interruption of the execution is carried over to the time of executing the predetermined time division process again. It is a symptom.

According to a second aspect of the present invention, when each of a large number of processes including a real-time process to be executed when a predetermined event occurs, when a predetermined event occurs, a preset allowance is set. If it is determined whether the currently executing process is in the waiting state within the delay time, and if it is determined that the currently executing process is in the waiting state within the allowable delay time, the process is placed in the waiting state. It is characterized in that the real-time process is activated at times.

According to a third aspect of the present invention, when each of a large number of processes including a real-time process to be executed when a predetermined event occurs, when the predetermined event occurs, the current process is being executed. The process is waiting,
Alternatively, the real-time process is activated when an elapsed time after the occurrence of the predetermined event reaches a preset allowable delay time.

According to a fourth aspect of the present invention, in the second or third aspect of the invention, the plurality of processes include a plurality of time division processes to be executed periodically, and the plurality of time division processes are included. A normal execution time and a carry-over boundary time shorter than the normal execution time are defined for each of the time-division processes, and when the predetermined event has not occurred, each of the plurality of time-division processes is If there is a predetermined time-division process that is being executed and has passed the carry-over boundary time or more after the execution is started when the predetermined event occurs, the time-division process is executed sequentially and repeatedly for the normal execution time. , The execution of the predetermined time-sharing process is suspended, the real-time process is started, and when the time-sharing process is ready to be executed, it should be executed following the predetermined time-sharing process. Time and executes the dividing process, the usual shortage of execution times of the predetermined time-division processes caused by interruption of the execution, is characterized in that carried over to the time of executing said predetermined time-division process again.

According to a fifth aspect of the invention, in the first or fourth aspect of the invention, the shortage of the normal execution time caused by interrupting the execution of the predetermined time division process in the past is carried over to the predetermined time. When the time-division process is executed, the priority of the predetermined time-division process is set to be higher than that of the normal time.

According to a sixth aspect of the invention, in the invention of the first or fourth aspect, when a plurality of processing units execute a plurality of processes in parallel, at least one of the plurality of processing units is used. The carry-over boundary time of the time division process executed in the processing unit is changed according to the execution status of the time division process in another processing unit.

[0018]

In general, in order to execute a time-sharing process, especially a time-sharing process having a low priority, as required, an execution rate per a predetermined period (CP of a specific time-sharing process in a predetermined period)
It is important that the ratio of U usage time) satisfies the specified value, and the size of the execution unit (execution time per execution)
Is not that important. If the execution rate per a predetermined period satisfies the specified value, the overhead can be reduced by increasing the execution unit and decreasing the number of executions. The inventor of the present application has reached the above facts and completed the present invention.

Therefore, according to the first aspect of the invention, when the execution of the predetermined time-sharing process being executed to start the real-time process is interrupted when a predetermined event occurs, the predetermined time-sharing process is executed. If the elapsed time from the start of execution is equal to or longer than the carry-over boundary time of the predetermined time division process,
When the execution of a predetermined time-sharing process is interrupted, a real-time process is started, and when the time-sharing process becomes ready to execute, another time-sharing process that should be executed following the predetermined time-sharing process is executed. The shortage of the normal execution time of the predetermined time division process caused by the suspension of the execution is carried over to the execution of the predetermined time division process again.

According to the above, when the elapsed time from the start of execution of the time-division process is the carry-over boundary time or more, that is, when the remaining execution time of the time-division process is small, the real-time process is executed and then interrupted. Since the time-sharing process that was being performed is not restarted, unnecessary process switching (dispatching) is performed in order to restart the execution of the time-sharing process even though the execution time of the time-sharing process is short. , The number of process dispatches is less than before. As a result, it is possible to prevent the overhead from increasing due to the occurrence of a predetermined event that should activate the real-time process.

Further, in the above, when the time-division process is ready to be executed after the real-time process is activated, another time-division process to be executed is executed following the predetermined time-division process whose execution has been interrupted. Therefore, the execution rate of the time-sharing process whose execution was interrupted due to the occurrence of a predetermined event will temporarily decrease, but the shortage of the normal execution time of the time-sharing process caused by the interruption of execution When the divided process is executed again, it is carried over, and when it is executed again, it is executed for a time longer than the normal execution time. Therefore, the average value of the execution ratio of the time divided process is maintained at a constant value. It will be.

By the way, when a predetermined event for activating a real-time process occurs, conventionally, as shown in FIG. 14, it is customary to immediately interrupt the process being executed and activate the real-time process. As a result, the number of process switchings and the overhead are increased. However, since the predetermined event occurs asynchronously with the execution of the process, there is a possibility that the execution of the process being executed will be completed within a relatively short time after the occurrence of the predetermined event.

Therefore, according to the second aspect of the present invention, when a predetermined event for activating a real-time process occurs, it is determined whether or not the process currently being executed enters a standby state within a preset allowable delay time. If it is determined that the currently executing process is in the standby state within the allowable delay time, the real-time process is activated when the process is in the standby state. According to the above, even if a predetermined event occurs, execution of other processes will not be interrupted if the currently executing process enters the standby state within the preset allowable delay time. Since the number of dispatches is small, it is possible to prevent the overhead from increasing due to the occurrence of a predetermined event that should activate the real-time process.

Since the time-sharing process is normally executed for a preset time, when the time-sharing process is being executed, the time-sharing process is executed based on the elapsed time from the start of execution of the time-sharing process. It is possible to determine whether or not to enter the standby state. In addition, it is difficult to detect the standby time for a real-time process in advance. However, for example, by measuring in advance the maximum time from startup to the spontaneous standby status, time sharing It is possible to determine whether or not the time-sharing process is in a standby state when the process is running.

According to the third aspect of the present invention, when a predetermined event for activating a real-time process occurs, the process currently being executed enters a standby state or after the predetermined event occurs. The real-time process is started when the elapsed time reaches the preset allowable delay time. As described above, it is difficult to detect beforehand the time when a real-time process or a time-sharing process spontaneously enters the waiting state. However, according to the above, the running process is in the waiting state within the allowable delay time. In this case, the real-time process can be executed subsequently to the process in the waiting state, and the execution of other processes can be uninterrupted. Therefore, the number of times the process is dispatched is small, and it is possible to prevent the overhead from increasing due to the occurrence of a predetermined event that should activate the real-time process.

Further, as described in claim 4, claim 2
Alternatively, also in claim 3, when a predetermined event occurs, if there is a predetermined time-sharing process that is being executed and has passed over the carry-over boundary time since the start of execution, the predetermined time-sharing process is executed. Interrupt and launch the real-time process,
When it becomes possible to execute a time-division process, another time-division process that should be executed following the predetermined time-division process is executed, and the normal execution time of the predetermined time-division process caused by interruption of execution is It is preferable to carry over the deficiency in performing the predetermined time-sharing process again.

According to the above, if there is a predetermined time-sharing process that is being executed and has passed over the carry-over boundary time since the start of execution, the real-time process is started immediately after the occurrence of a predetermined event. The responsiveness of real-time process startup to the occurrence of a predetermined event is improved. Similarly to claim 1, when a predetermined event occurs, the process is unnecessarily dispatched in order to restart the execution of the time-sharing process even though the remaining time of the time-sharing process being executed is small. Since the process is not performed and the number of times of dispatching the process is smaller than in the conventional case, it is possible to prevent the overhead from increasing due to the occurrence of a predetermined event that should activate the real-time process.

By the way, when the execution of the predetermined time-division process is interrupted and the shortfall of the normal execution time caused by the interruption is carried over, the shortfall is added to the normal execution time to execute the predetermined time-division process. While doing so, the execution of the predetermined time-sharing process may be interrupted again due to the occurrence of a predetermined event or the like. In this case, the execution rate of the predetermined time-sharing process is greatly reduced, which is not preferable.

In view of the above, as described in claim 5, the shortage of the normal execution time caused by interrupting the execution of the predetermined time-sharing process in the past is carried over to execute the predetermined time-sharing process. During this time, it is preferable that the priority of the predetermined time division process is set higher than that of the normal time. As a result, it is possible to prevent the execution rate of the specific time-division process from being significantly reduced by suspending the execution of the specific time-division process multiple times at short time intervals.

Further, in claim 1 or claim 4, when a predetermined event for executing the real-time process occurs during execution of the time-sharing process, the execution is performed when the time-sharing process is ready to be executed. The number of dispatches will be less if you carry over the lack of execution time of the time-sharing process than when you restart the time-sharing process that interrupted
The increase in overhead can be suppressed. However, whether or not to carry over the shortage of the normal execution time of the time-sharing process depends on whether the time from the start of execution of the time-sharing process to the occurrence of a predetermined event is the carry-over boundary time or more. To do.

Therefore, when a time-sharing process in which a relatively large value is set as the carry-over boundary time is being executed, when a predetermined event occurs, the time-sharing process whose execution is interrupted is normally executed. As in the conventional case, the overhead is likely to increase without carrying over the lack of time. In particular, when executing multiple processes in parallel in multiple processing units, if the process being executed in each processing unit is a time division process in which a relatively large value is set as the carry-over boundary time, It is very likely that the overhead will increase with the occurrence of a given event.

In view of the above situation, claim 6
As described above, when the plurality of processing units execute the plurality of processes in parallel, the carry-over boundary time of the time-sharing process executed by at least one of the plurality of processing units is It is preferable to change according to the execution status of the time division process in the processing unit. As a result, for example, in a situation in which a plurality of time division processes each having a relatively large carry-over boundary time set as described above are executed in parallel by a plurality of processing units, for example, at least one time division process is executed. For example, by changing the carry-over boundary time to a small value, it is possible to avoid an increase in overhead due to occurrence of a predetermined event. Therefore, it is possible to avoid an increase in overhead due to occurrence of a predetermined event regardless of the execution status of the time division process.

[0033]

Embodiments of the present invention will be described below in detail with reference to the drawings. In addition, although the following description will be given using numerical values that do not hinder the present invention, the present invention is not limited to the numerical values described below.

FIG. 1 shows a schematic block diagram of an information processing apparatus 10 to which the present invention can be applied. The information processing device 10 includes n CPUs 1, CPU 2, ..., CPUn, and each CPU is connected to a bus 14. Bus 1
Various peripheral devices (not shown) (for example, keyboard, display, printer, communication control device, etc.) are connected to 4. A shared memory 16 that functions as a main storage device of each CPU is connected to the bus 14. Various information including the process dispatch control program 18 and the system database 20 is stored in advance in the storage area of the shared memory 16.

The information processing apparatus 10 according to the present embodiment has a CP
U1, CPU2, ..., CPUn operate in parallel with each other,
Moreover, it is a so-called symmetric multiprocessor system in which the division of roles of each CPU is not defined like the master-slave system. Therefore, it is indefinite which CPU each of the multiple processes to be executed by the information processing apparatus 10 is executed, and the above-described process dispatch control program 18 switches the processes executed in each CPU. If executed.

As shown in FIG. 2, the system database 20 is configured to include various information such as the time division class queue 20A, the real time class queue 20B, the activation delay table 20C and the running process table 20D. .

A large number of processes to be executed by the information processing apparatus 10 are a group of processes (time-division processes) to be executed periodically (time-division class), or processes to be executed when a predetermined event occurs ( Each of them is pre-divided into groups (real-time classes) (real-time processes), and execution priorities are given in advance. The shared memory 16 is provided with a task control block (hereinafter referred to as “TCB”) for setting various information necessary for executing each process corresponding to each process. The time division class queue 20A is linked with the TCBs of the time division processes in the execution waiting state, and the real time class queue 20B is linked with the TCBs of the real time processes in descending order of priority. .

As shown in FIG. 3, the TCB includes process information (1) composed of information such as process name, size, class (whether a time division process or a real time process), and is used when the process is executed. Process information (2) including mapping information and the like indicating the position of a predetermined area on the shared memory 16, and information such as process priority determined in advance for each process are set in advance. Further, regarding the TCB of the time division process, a time slice time T1 (corresponding to the normal execution time of the present invention) and a carry-over boundary time T2 are also set.

Further, the TCB is provided with fields for setting process information (3), a carry-over processing flag, and carry-over time T3 (corresponding to the shortage of the normal execution time of the present invention) or remaining execution time Ts. There is. The field of the process information (3) is an area for saving the information held in the register of the CPU, the program counter, or the like when the execution of the process is first ended or interrupted. The carry-over process flag, the carry-over time T3, and the remaining execution time Ts will be described later, but are set only in the TCB of the time division process.

The activation delay table 20C is shown in Table 1 below.
As also shown, the permissible delay time from the occurrence of an event that should activate the real-time process to the activation of the corresponding real-time process is determined for each process priority and each CPU. In addition, in this embodiment, as is clear from Table 1, the process priority is divided into 160 types of levels, and the levels 1 to 160 are assigned in order from the lowest level to distinguish each level.

[0041]

[Table 1]

Further, as shown in Table 2 below, the running process table 20D is a table for registering the information of the process currently being executed by each CPU, and each time a new process is started by each CPU. The information such as "class" and "priority" is rewritten in. If the process to be activated is a time-division process, information such as "time slice time" and "carry-on boundary time" is also set. The “CPU usage time” represents the CPU usage time of the process being executed, and is updated by the CPU consumption time measurement / holding unit 32 described later so that the value gradually increases with the passage of time. The value is reset each time a new process is started in each CPU.

[0043]

[Table 2]

In FIG. 2, among various functions of each CPU, the functions related to the process dispatch realized by executing the process dispatch control program 18 in each CPU are shown in blocks. Has been done.

The timer 26 holds the current time,
The CPU consumption time detection unit 28 monitors a change in the time held in the timer 26, and outputs a timer interrupt signal to the process dispatch control unit 30 each time a fixed time has elapsed. In the CPU consumption time measurement / holding unit 32 of the process dispatch control unit 30, when the timer interrupt signal is input, the CPU of the running process table 20D is processed as described above.
The usage time is updated, and the value is reset (returned to 0) each time a new process is started by the CPU.

The external event detection unit 34 detects an interrupt signal input in response to the occurrence of an external event such as data input, data output completion, resource release, etc., and the process of the process dispatch control unit 30 is detected. The selection control unit 36 is notified of the occurrence of the external event. Process selection control unit 3
In 6, the CPU that selects the process to be executed in response to the occurrence of the external event is selected.

In the selected CPU, if the process to be executed is a real-time process, the real-time process dispatch control unit 38 dispatches the real-time process to be executed, and the process to be executed is the time-division process. In this case, the time-sharing process dispatch control unit 40 dispatches the time-sharing process to be executed. Further, when no event has occurred, the time-sharing process dispatch control unit 40 dispatches the time-sharing process.

Further, the time slice value changing unit 42 judges whether or not the carry-over boundary time T2 of the time-division process to be started should be changed at the time of starting the time-division process, based on the execution status of the process, and should be changed. If it is determined that the carry-over boundary time T2 is changed. Priority changing unit 44
Changes the priority of the time-sharing process to which the carry-forward process described later is applied.

Next, the operation of this embodiment will be described. First, FIG.
The event reception process executed when an event occurs will be described with reference to the flowchart of FIG. In addition,
The event reception process is a process corresponding to the process selection control unit 36 and the real-time process dispatch control unit 38 described above. Further, since the information processing apparatus 10 according to the present embodiment is a symmetric multiprocessor system, it is uncertain which of the CPU1 to CPUn will perform the event reception processing described below, and a CP that executes the event reception processing.
U is determined by the operating system (OS).

In step 100, it is determined whether the generated event is an event for activating a real-time process.
Even in the time-sharing process, the process may shift to a state of waiting for the occurrence of a predetermined event. If the occurring event is an event that should activate the time-sharing process, the above determination is denied, and step 102 After performing the time-sharing class event process that starts the time-sharing process corresponding to the event that occurred in step 1, the process ends. If the determination is affirmative, the process proceeds to step 104, and the CPU 1
~ Among the processes being executed by CPUn, the carry-over boundary time T2 or more has passed since the start of execution (that is, C
PU usage time ≧ carrying boundary time T2) It is determined whether or not there is a time division process.

If the determination is positive, step 106
Then, the CPU executing the time division process starts the switching process to the real-time process and ends the process. The details of this switching processing will be described later, but the processing of the time-divisional process being executed is interrupted, the real-time process corresponding to the event that has occurred is started, and the condition (CPU usage time ≧ carrying boundary time T2) is satisfied. Therefore, the carry forward process is applied to the time division process whose execution is interrupted.

On the other hand, when the determination at step 104 is negative, the routine proceeds to step 108, where CPU1 to CPUn
Of the processes being executed, it is determined whether or not there is a time division process in which the time slice time T1 expires (that is, CPU usage time = time slice time T1) within the allowable delay time. More specifically, this determination uses the priority of the real-time process to be activated as a key and the activation delay table 20C.
Value obtained by subtracting the CPU usage time from the time slice time T1 set in the executing process table 20D for each of the CPUs executing the time-divisional process, while taking in the allowable delay time in each CPU. (Time until expiration of time slice) is calculated, and the calculated value is compared with the accepted allowable delay time.

If the above determination is affirmative, after the time slice time has expired and the processing of the time-sharing process is stopped, the CPU that was executing this time-sharing process executes the real-time process. In step 112, the occurrence of the event is stored as information, and the process ends. Even if the above determination is denied, the process being executed spontaneously shifts to the waiting state within the allowable delay time, or the CPU use time by the time-sharing process being executed is carried forward to the boundary time T2 or more. Therefore, in step 110, a process switching timer that times out when a predetermined allowable delay time elapses is started, and in step 112, the occurrence of an event is stored as information and the processing ends.

The process switching timer started in step 110 is used when the real-time process corresponding to the generated event is activated within the allowable delay time due to the process being executed voluntarily shifting to the waiting state. Will be stopped, but otherwise timed out,
Along with this timeout, the process switching request timer timeout acceptance process (see FIG. 5) is executed. Less than,
The process switch request timer timeout acceptance process will be described with reference to FIG. Note that this processing is also processing corresponding to the process selection control unit 36 and the real-time process dispatch control unit 38 described above. Further, it is uncertain which of the CPU1 to CPUn will perform this processing, and the CPU that performs this processing is determined by the operating system (OS).

At step 120, the executing process table 20D is read, and at step 122, CPUs 1 to CP are read.
It is determined based on the read information whether or not there is a time-divisional process of which the carry-over boundary time T2 or more has passed from the start of execution among the processes being executed by Un. When the determination is affirmative, the process proceeds to step 128, and similarly to the above, the CPU executing the time division process activates the switching process to the real time process. In this case, the time division process whose processing is interrupted due to the activation of the real-time process is (CPU usage time ≧ carrying boundary time T
Since the condition of 2) is satisfied, the carry-forward process is applied to the time division process. And the next step 130
Then, the information indicating that an event has occurred, which is stored in the previous event reception process, is erased, and the process ends.

If the determination in step 122 is negative, the running process table 20 in step 124.
Based on the information read from D, it is determined whether the time division process is being executed by at least one of the CPU1 to CPUn. If the determination is affirmative, the processes of steps 128 and 130 are performed as described above. In this case, the time-sharing process whose processing is interrupted when the real-time process is started by switching to the real-time process is
Since the condition of (CPU use time ≧ carry-over boundary time T2) is not satisfied, the carry-over process is not applied to the time-division process. When a plurality of CPUs are executing the time-sharing process, in step 128, the CPU executing the time-sharing process having the lowest priority starts the switching process to the real-time process.

If the determination in step 124 is negative, all the processes executed by the CPUs 1 to CPUn are real-time processes. Therefore, in step 126, among the real-time processes being executed by the CPUs 1 to CPUn, It is determined whether or not there is a real-time process having a lower priority than the real-time process (the real-time process to be started) corresponding to the event in which the occurrence is stored. If the above determination is affirmed, the same as the above step 12
Any of the real-time processes that are being executed after processing 8 and 130 (the real-time process having the lowest priority if there are multiple real-time processes that have a low priority) Will be interrupted, and the real-time process with higher priority will be executed.

If the priority of each of the real-time processes executed by the CPU1 to CPUn is high, the determination at step 126 is denied and the process proceeds to step 132, the process switching timer is restarted and processed. To finish. This is an extremely rare case such as the occurrence of an event that should start a high-priority real-time process is congested, but whether any of the running real-time processes voluntarily shifts to the waiting state. , Or the activation of the real-time process in the activation waiting state is suspended until the process switching timer times out again.

Next, the switching processing to the real-time process executed by the corresponding CPU in accordance with the processing in step 106 of FIG. 4 or step 128 of FIG. 5 will be described with reference to FIG.
This will be described with reference to the flowchart in FIG. Note that this processing is processing corresponding to the real-time process dispatch control unit 38. In step 140, the processing of the process being executed is stopped, and in step 142, the information (such as the information set in the register) of the process that has stopped processing is set in the corresponding TCB. In the next step 144, the process execution interruption information indicating that the execution of the process is interrupted is set in the TCB, and in step 146, the corresponding information from the executing process table 20D (information of the process that has been performing the process up to now). ) Is erased.

In the next step 148, the previous step 1
At 40, it is determined whether the process that has stopped processing is a time-sharing process. If the determination is negative, the process proceeds to step 166, but if the determination is affirmative, step 150
The CPU usage time t by the time-sharing process whose processing has been stopped at is taken in. In step 152, it is determined whether the fetched CPU usage time is equal to or longer than the carry-over boundary time T2 set in the TCB of the time division process. If the determination is affirmative, the process proceeds to step 154.

In step 154, in order to apply the carry-forward process to the time-division process whose processing has been stopped, the carry-over time T3 when the time-division process is executed next time is executed.
(Time until expiration of time slice) is calculated according to the following equation (1).

Carry-over time T3 = time slice value T1-CPU usage time t (1) In step 156, T of the time-division process whose processing has been stopped
The carry-over time T3 calculated above is set in CB, and the carry-over process flag is set. Then, in the next step 158, the TCB set with the above information is linked to the end of the time division class queue 20A, and step 166
Move to. On the other hand, when the determination in step 152 is negative, the time slice remaining time Ts is determined in step 160.
(Time until expiration of time slice) is calculated according to the following equation (2).

Remaining time Ts = time slice value T1-CPU usage time t (2) In step 162, the remaining time Ts calculated above is set in the TCB of the time-division process whose processing has stopped, and in the next step 164 The TCB in which the remaining time Ts is set is linked to the head of the time division class queue 20A, and step 166 is performed.
Move to.

At step 166, the information about the real-time process to be started is fetched from the corresponding TCB, and at the next step 168, the necessary information about the real-time process to be started is registered in the running process table.
Then, by starting the processing of the real-time process in step 170, the switching processing to the real-time process is completed.

Next, the real-time process ending process executed by the CPU that was executing the real-time process when the real-time process that was executing next voluntarily shifts to the waiting state is shown in the flowchart of FIG. It will be described with reference to FIG. Note that this process is performed by the process execution CPU selection control unit 3
6. This is processing corresponding to the real-time process dispatch control unit 38 and the time-sharing process dispatch control unit 40. In step 180, the information of the real-time process whose processing is completed by shifting to the waiting state is set to the corresponding T
CB is set, and in step 182, the corresponding information (information on the real-time process which has been processed up to now) is deleted from the executing process table 20D.

At step 184, it is determined whether or not there is a real-time process which is interrupted by another real-time process and whose execution is suspended. This determination is made by determining whether or not there is a TCB for which process execution interruption information is set among the plurality of TCBs linked to 20B in the real-time class queue.

If the determination is positive, step 18
At 6, it is determined whether or not information indicating that an event has occurred is stored. If the determination is negative, the process proceeds to step 190. However, if the determination is affirmative, in step 188, it is stored that the execution-time suspended real-time process found in the previous step 184 has occurred. It is determined whether the real-time process corresponding to the current event has a higher priority.

If the determination is affirmative, it is determined that the real-time process corresponding to the event should be started, and the process proceeds to step 198. However, if the determination is negative, the execution is suspended. It is determined that the time process should be started, and the process proceeds to step 190. Step 190
Then, the information about the real-time process whose execution is suspended is fetched from the corresponding TCB, and in the next step 192, the necessary information about the real-time process to be activated is registered in the running process table. Then, in step 194, the processing of the real-time process is started to end the real-time process end processing.

On the other hand, if there is no real-time process whose execution is suspended, the determination in step 184 is denied, and it is determined in step 196 whether the occurrence of the event is stored. If the determination is affirmative, it means that the event has occurred but the corresponding real-time process has not been started yet. In step 198 and thereafter, the real-time process corresponding to the stored event is started.
If multiple types of event occurrences are stored,
The following processing is performed for the real-time process having the highest priority among the real-time processes corresponding to each event, or the real-time process corresponding to the event that occurs fastest.

In step 198, if the process switching request timer corresponding to the real-time process to be started is operating, this timer is stopped. The process switch request timer is operated when the determination in step 108 (see FIG. 4) is denied in the event reception process performed when the event corresponding to the real-time process occurs. In the next step 200, information about the real-time process that is about to start is provided in the corresponding TCB.
In step 202, necessary information about a real-time process to be activated is registered in the executing process table. In the next step 204, the information indicating that the event corresponding to the real-time process to be activated has occurred is erased, and the process of the real-time process is started in step 206, thereby ending the real-time process end process.

If there is no real-time process whose execution is suspended and the occurrence of an event is not stored,
When the determination in step 196 is negative, the process proceeds to step 208, and after the time-sharing process starting process is performed, the real-time process ending process is ended. The time division process start processing will be described later.

Next, when the time slice time T1 of the time-division process being executed expires or the time-division process voluntarily shifts to the waiting state, the CPU executing the time-division process The time division process end processing executed will be described with reference to the flowchart in FIG. This process is also a process corresponding to the process execution CPU selection control unit 36, the real-time process dispatch control unit 38, and the time-division process dispatch control unit 40. In step 220, the information of the time-division process that has completed processing is set in the corresponding TCB, and step 222
Then, the TCB in which the above information is set is linked to the end of the time division class queue 20A. In step 224, the corresponding information (information on the time-division process which has been processed up to now) is deleted from the process table 20D being executed.

The processing from the next step 226 is executed in steps 184 to 20 of the real-time process end processing described above.
8 (see FIG. 7). That is, step 22
At 6, it is determined whether or not there is a real-time process that is interrupted by another real-time process and whose execution is suspended. If the determination is affirmative, information indicating that an event has occurred at step 228 is displayed. It is determined whether or not it is stored.
If the determination is negative, the process proceeds to step 232, but if the determination is affirmative, in step 230, the detected real-time process in which the execution is suspended corresponds to the event stored as occurring. It is determined whether the real-time process to be executed has a higher priority.

If the determination is negative, step 232
In step 234, the information about the real-time process whose execution is suspended is fetched from the corresponding TCB, and in step 234, the necessary information about the real-time process to be activated is registered in the running process table. Restart and end the time-sharing process termination processing.

On the other hand, when the determination in step 226 is negative, it is determined in step 238 whether the occurrence of the event is stored. If the determination is affirmative, the routine proceeds to step 240, and if the process switching request timer corresponding to the real-time process to be started is running, this timer is stopped. In the next step 242, information about the real-time process to be started up is fetched from the corresponding TCB, and in step 244, necessary information about the real-time process to be started up is registered in the running process table. At the next step 246, the information indicating that the event corresponding to the real-time process to be activated has occurred is deleted, and at step 206, the processing of the real-time process is started.
The time sharing process termination processing is terminated. Also, step 2
If the determination in step 38 is negative, the time-sharing process start processing is performed in step 250, and then the time-sharing process end processing is ended.

Next, referring to the flowchart of FIG.
The time-division process activation processing performed in step 208 of FIG. 8 or step 250 of FIG. 8 will be described. It should be noted that this process is a process corresponding to the time-division process dispatch control unit 40. In step 260, the TCB linked to the head of the time division class queue 20A is taken out, and in the next step 262, the TCB extracted is taken out based on whether or not the process execution interruption information is set in the taken out TCB. The corresponding time-division process (time-division process to be activated) determines whether the previous execution was interrupted before the time slice time T1 expired. If the determination is negative, the time slice time T1 set in the TCB is set as the current time slice value (time slice value of the time division process to be activated) T in step 264, and the process proceeds to step 282. To do.

When the determination in step 262 is affirmative, the process proceeds to step 266, and the TC that has been taken out is taken.
Based on whether or not the carry-forward processing flag is set in B, it is determined whether or not the time-sharing process to be activated from now on is a time-sharing process to which the carry-forward processing is applied. If the determination is affirmative, the remaining time Ts set in the TCB is set as the current time slice value T in step 268. In step 270, the process execution interruption information is erased from the TCB, the remaining time Ts set in the TCB is reset in step 272, and then the process proceeds to step 282.

When the determination at step 266 is affirmative, the routine proceeds to step 274, where the sum of the time slice value T1 set in TCB and the carry-over time T3 is set as the current time slice value T. In the next step 276, a carry-over time elapsed timer is started which times out when the carry-over time T3 has elapsed. In step 278, for the time-sharing process to be started,
Set the priority for carry-forward processing (higher priority than normal). This step 278 corresponds to the priority changing unit 44 and is a process corresponding to claim 5.

As a result, even if an event for activating a real-time process occurs while a time-division process to be activated is being executed, if another CPU is executing another time-division process, The execution of the other time division process is likely to be interrupted. Therefore, it is less likely that the execution of the time-sharing process to which the carry-forward process is applied is interrupted due to the occurrence of the event. In the next step 280, the process execution interruption information is erased from the TCB, the carry-over time T3 set in the TCB is reset, and then the process proceeds to step 282.

At step 282, necessary information about the time-sharing process to be activated is registered in the executing process table, and at next step 284, the information registered in the executing process table is read. Then, in step 286, based on the read information, it is determined whether or not it is necessary to change the carry-over boundary time T2 of the time division process to be started.

As an example, in all the time-division processes being executed by other CPUs, if the remaining time until the carry-over boundary time T2 elapses is relatively long, an event occurs within a relatively short period from the present time. Then, even if the execution of the time-sharing process is interrupted after the allowable delay time has elapsed and the real-time process is started, there is a high possibility that the carry-forward process is not applied to the time-sharing process whose execution has been interrupted, and the carry-forward process is applied. Compared with the case, the overhead is increased due to the increase in the number of process dispatches.

In the above case, the determination at step 286 is affirmative, and the carry-over boundary time T2 of the time-sharing process to be activated at step 288 is small, although it depends on the priority of the other time-sharing process being executed. After the carry-over boundary time T2 is changed and set to be a value, the process proceeds to step 290. This increases the possibility that the carry-forward process will be applied to the time-sharing process whose execution has been interrupted to start the real-time process, thus avoiding the increase in overhead due to the increase in the number of process dispatches. be able to.

In step 228, the carry-over boundary time T
The time-sharing process as a target for changing and setting 2 is not limited to the time-sharing process to be executed from now on, and the carry-over boundary time T2 of another time-sharing process being executed is changed and set. Good. For example, when a time-sharing process having a lower priority than the time-sharing process to be started from now on is being executed, the carry-over boundary time T2 of the time-sharing process having a lower priority may be changed and set. Good. The steps 286 and 288 correspond to the time slice value changing unit 42, and
Is a process corresponding to.

When the determination in step 286 is negative, the process proceeds to step 290. In step 290, the processing of the time division process is started, and the time division process activation processing is ended.

When the carry-over time elapse timer times out, the carry-over time elapse timer timeout acceptance processing shown in FIG. 10 is executed. That is, step 300
Then, it is determined whether or not the execution of the time division process corresponding to the time-out carry-over time elapsed timer is continued. If the determination is negative, it means that the time-sharing process started by applying the carry-over process is interrupted before the carry-over time T3 elapses due to the start of the real-time process accompanying the occurrence of the event. The process ends without performing any processing.

On the other hand, if the determination in step 300 is affirmative, the time-sharing process started by applying the carry-over process has been executed for the carry-over time T3 or longer, so that the priority of the time-sharing process is set in step 302. Original value (TC
The value set in B) is set. The step 302 corresponds to the priority changing unit 44. At the next step 304, the carry-over process flag of the TCB is reset and the process ends.

Next, with reference to FIG. 11, an example of the dispatch operation by the dispatch control processing described above will be described in comparison with the dispatch operation by the conventional method. Note that in FIG. 11, two CPUs (C
PU1 and CPU2), and a case where a total of four time-sharing processes of process A to process D are set as the time-sharing process is shown as an example, and the time slice time T1 of each process is the process A as an example. Is 30m
s, Process B is 20 ms, Process C is 10 ms, Process D
Is set to 5 ms, and the carry-over boundary time T2 of process B is
Is set to 10ms. Further, in FIG. 11, the numerical value in parentheses represents the execution time of each process.

When no event occurs as shown in FIG. 11A, each process has time slice time T1.
Is executed by CPU1 or CPU2 until the time expires, and when the time slice time expires, the time division class queue 20A
Another time-sharing process corresponding to the TCB linked to the head of the is dispatched. In FIG. 11 (A), the number of times the process is dispatched is eight while each of the processes A to D is executed twice (that is, per two cycles).

When an event for activating a real-time process occurs, execution of a predetermined time-division process is interrupted as shown in FIG. 11B (process B in FIG. 11) in the conventional dispatch method. ) When the real-time process voluntarily shifts to the waiting state after the real-time process is started, the execution of the time-division process that has suspended the execution is restarted. In this case, the number of process dispatches increases to 10 every two cycles.

On the other hand, in this embodiment, as shown in FIG. 11C, the process B at the time when the event occurs.
If the execution time of is greater than or equal to the carry-over boundary time T2 of the process B, the determination in step 104 of the event reception process (FIG. 4) is affirmed, and the CPU 2 that was executing the process B switches to the real-time process (FIG. 6) is activated. In the switching process to the real-time process, step 152
Is affirmed, the carry-over time T3 is calculated, the calculated carry-over time T3 is set in the TCB, and the carry-over process execution flag is set, and then the TCB is linked to the end of the time division class queue 20A. .

As a result, the next process C is executed after the real-time process voluntarily shifts to the waiting state. Further, in the time-division process activation processing (FIG. 9) executed by the CPU 1 when the time slice time of the process D in the CPU 1 expires (in FIG. 11C), the process moved to the head of the time-division class queue 20A. B's T
The CB is taken out, and since the process execution interruption information and the carry-over processing flag are set in this TCB, the determinations at steps 262 and 266 are affirmative. Therefore, the carry-forward process is applied to the process B, and the time corresponding to the sum of the time slice time T1 and the carry-over time T3 (see FIG. 11).
In (C), the process B is executed only for "20 + X". In this case, the number of times the process is dispatched is 9 per 2 cycles, and it can be understood that the increase in the overhead due to the occurrence of an event is suppressed as compared with the conventional method.

Next, with reference to FIG. 12, an example of the dispatch operation by the aforementioned dispatch control processing will be further described in comparison with the dispatch operation by the conventional method. In the conventional method, as soon as an event occurs, the corresponding real-time process is started. Therefore, as an example, as shown in FIG. 12A, the processing of the process being executed is terminated after a very short time has elapsed (in FIG.
The execution of the process is interrupted even when the time slice expires), and after the processing of the real-time process is completed, the interrupted time division process is restarted to execute for a very short time slice remaining time. Inefficiency is high and dispatches are frequent.

On the other hand, in the present embodiment, as shown in FIG. 12B, the time-division process B being executed reaches the end of the time slice within the period from the occurrence of the event to the passage of the allowable delay time. In this case, the determination in step 108 of the event reception process (FIG. 4) is affirmed, and the activation of the real-time process C corresponding to the occurred event is suspended.
Then, when the CPU that was executing the time-division process B starts the time-division process termination processing (FIG. 8) when the time-slice process B has completed the time slice, if there is no real-time process whose execution is suspended, the step If the determination at 238 is affirmative, the real-time process C that has been suspended from being activated is activated. This prevents dispatching to execute the time-division process for a short time, which improves efficiency and reduces the number of dispatches.

Further, in this embodiment, as shown in Table 1,
Since the allowable delay time value is decreased as the priority of the real-time process increases, the real-time process with a high priority will start the real-time process in a very short time when an event occurs. , High responsiveness is secured.

Further, as shown in FIG. 13A, CP
U2 has already executed the time-sharing process B with a long time slice time, and there is no other CPU, or another C
In a situation where the PU is continuously executing a real-time process,
In the CPU 1, when the time slice process and the carry-over boundary time are relatively long (the time slice time is shown as T1 and the carry-over boundary time is shown as T2), when the time-sharing process A is started, the execution is interrupted due to the occurrence of an event. The period during which the carry-forward process is not applied to
The period L) shown in FIG. Therefore, even if the event occurs and the start of the real-time process is delayed by the allowable delay time, it is highly likely that the carry-forward process is not applied to the time-sharing process whose execution is interrupted.

On the other hand, in the above-described embodiment, in the time-sharing process starting process (FIG. 9) executed to start the time-sharing process A in the CPU 1, the affirmative judgment in step 286 results in step 288. The carry-over boundary time of the time division process A is short (T shown in FIG. 9B).
2 ') is performed, the period L during which the carry-over processing is not applied to the time-sharing process whose execution has been interrupted is also shortened, and the carry-over processing may be applied to the time-sharing process in which an event has occurred and whose execution has been interrupted. Becomes higher. When the carry forward process is applied, the number of dispatches can be smaller than when the carry forward process is not applied.
There is a high possibility that the increase in overhead due to the occurrence of an event can be further suppressed.

In the event reception process shown in FIG. 4, even if there is a time division process in which the time slice ends within the allowable delay time when the event occurs, the CPU usage time is the carry-over boundary time T2 or more. If there is a time-sharing process, the execution of the time-sharing process is interrupted and the real-time process is delayed, but the present invention is not limited to this. For example, when the determination in step 108 is performed before the determination in step 104, even if there is a time-sharing process in which the CPU usage time is carried over to the boundary time T2 or more, the allowable delay time when an event occurs If there is a time-sharing process whose time slice has expired, the start of the real-time process is delayed, so it is possible to reduce the temporary reduction in the execution rate of the time-sharing process due to the start of the real-time process. Further, the execution order of the above two determinations may be changed according to the priority of the real-time process to be activated.

Further, in the above, when the carry-over process is applied to the predetermined time-division process whose execution has been interrupted, the carry-over time T3 is carried over when the predetermined time-division process is executed next time (this time slice time). Although the sum of the time slice time T1 and the carry-over time T3 is set as T), the invention is not limited to this. The carry-over time T3 is carried over in a plurality of times (for example, T3 is t ₃ 'and t _3). divided into the "(However T3 = t ₃ '+
t ₃ ″), the next time slice time T of a predetermined time division process is T1 + t ₃ ′, and the next time slice time T is T1 + t ₃ ″).

Although the information processing apparatus 10 which is a symmetric multiprocessor system has been described above as an example of the information processing apparatus to which the present invention can be applied, the present invention is not limited to this, and the asymmetric multiprocessor is not limited thereto. It goes without saying that the present invention is also applicable to a system or a system having only a single processor and performing processing by multitasking.

[0100]

As described above, according to the invention of claim 1, when the execution of a predetermined time-sharing process that is being executed to start a real-time process is interrupted when a predetermined event occurs, a predetermined time-sharing process is interrupted. If the elapsed time after the execution of the time-sharing process is started is equal to or longer than the carry-over boundary time of the predetermined time-sharing process, the execution of the predetermined time-sharing process is interrupted, the real-time process is started, and the time-sharing process is started. When it becomes a state where it can be executed, while executing another time-sharing process that should be executed following the predetermined time-sharing process, the shortage of the normal execution time of the predetermined time-sharing process caused by the interruption of execution is Since the predetermined time-sharing process is carried over when it is executed again, it is possible to prevent the overhead from increasing due to the occurrence of a predetermined event that should activate the real-time process. It was having an effect.

In a second aspect of the present invention, when a predetermined event for activating a real-time process occurs, it is determined whether or not a process currently being executed enters a standby state within a preset allowable delay time. However, if it is determined that the process that is currently running enters the standby state within the allowable delay time, the real-time process is started when the process enters the standby state. This has an excellent effect that it is possible to suppress an increase in overhead due to occurrence of a predetermined event to be performed.

According to the third aspect of the present invention, when a predetermined event for activating a real-time process occurs, the process currently being executed is in a standby state, or the process has elapsed since the predetermined event occurred. This has an excellent effect that it is possible to suppress an increase in overhead due to occurrence of a predetermined event for activating the real-time process when the time reaches a preset allowable delay time.

According to a fourth aspect of the present invention, in the second or third aspect of the invention, when a predetermined event for activating a real-time process occurs, the real-time process is being executed and carried over after the execution is started. If there is a predetermined time-sharing process that has passed the boundary time or more, the execution of the predetermined time-sharing process is interrupted, the real-time process is started, and when the time-sharing process becomes ready to execute, the predetermined time-sharing process In addition to executing another time-sharing process that should be executed, the shortage of the normal execution time of the predetermined time-sharing process caused by the interruption of execution is carried over to the execution of the predetermined time-sharing process again. Therefore, in addition to the above effects, there is an effect that the responsiveness of the real-time process activation to the occurrence of a predetermined event is improved.

According to a fifth aspect of the present invention, in the first or fourth aspect of the present invention, the shortage of the normal execution time caused by interrupting the execution of the predetermined time division process in the past is carried over to the predetermined time. When executing the time-sharing process of, the priority of the predetermined time-sharing process is set to be higher than that of the normal time, so in addition to the above effect, the execution rate of the specific time-sharing process is significantly reduced. It has the effect that it can be avoided.

According to a sixth aspect of the present invention, in the invention according to the first or fourth aspect, when a plurality of processing units execute a plurality of processes in parallel, the time division is performed by at least one processing unit. Since the process carry-over boundary time is changed according to the execution status of the time-sharing process in other processing units, in addition to the above effects, the overhead associated with the occurrence of a predetermined event is generated regardless of the execution status of the time-sharing process. This has the effect of avoiding an increase in

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an information processing apparatus according to an embodiment.

FIG. 2 is a schematic diagram showing an outline of the contents of a system database and showing the functions realized by each CPU as functional blocks.

FIG. 3 is a conceptual diagram showing the contents of TCB.

FIG. 4 is a flowchart illustrating the content of event reception processing.

FIG. 5 is a flowchart illustrating the contents of a process switching request timer timeout acceptance process.

FIG. 6 is a flowchart illustrating the contents of a process of switching to a real-time process.

FIG. 7 is a flowchart illustrating the contents of real-time process termination processing.

FIG. 8 is a flowchart illustrating the content of time division process end processing.

FIG. 9 is a flowchart illustrating the content of time-division process activation processing.

FIG. 10 is a flowchart illustrating the contents of a carry-over time elapsed timer timeout acceptance process.

11A is an example of a dispatch operation when an event does not occur, FIG. 11B is an example of a dispatch operation by a conventional method when an event occurs, and FIG. 11C is a method of the present invention. FIG. 7 is a conceptual diagram showing an example of a dispatch operation under the same conditions as in FIG.

FIGS. 12A and 12B are diagrams showing an example of a dispatch operation in a conventional method and an example of a dispatch operation in the method of the present invention, respectively, for explaining the effect of delaying the activation of a real-time process. It is a figure.

FIG. 13A is a diagram illustrating a problem when a time division process having a long time slice time is executed in parallel, and FIG. 13B is a diagram illustrating an operation when dynamically changing a carry-over boundary time. It is a conceptual diagram.

FIG. 14 shows a conventional process dispatch method.
(A) is only executing the time division process, (B)
Is a real-time process only, (C) is a time-sharing process and a real-time process,
It is a conceptual diagram which shows an example of operation | movement of dispatch.

[Explanation of symbols]

 10 information processing device 16 shared memory

Claims (6)

[Claims] 1. When executing a plurality of time-sharing processes to be executed regularly and a real-time process to be executed when a predetermined event occurs, normal execution is performed for each of the plurality of time-sharing processes. Time and a carry-over boundary time shorter than the normal execution time are set in advance, and when the predetermined event does not occur, each of the plurality of time division processes is sequentially processed by the normal execution time of each time division process. And when repeatedly executing and interrupting the execution of the predetermined time-sharing process that is being executed in order to start the real-time process when a predetermined event occurs, after executing the predetermined time-sharing process, If the elapsed time is equal to or longer than the carry-over boundary time of the predetermined time division process, the execution of the predetermined time division process is interrupted, the real time process is started, and the time division process is started. When the ready state is reached, another time-sharing process that should be executed after the predetermined time-sharing process is executed, and the shortage of the normal execution time of the predetermined time-sharing process caused by the interruption of the execution. Is carried over when the predetermined time division process is executed again. 2. When each of a number of processes including a real-time process to be executed when a predetermined event occurs, when the predetermined event occurs, the process is currently executed within a preset allowable delay time. If it is determined whether the current process is in the standby state and the currently executing process is in the standby state within the allowable delay time, the real-time process is executed when the process is in the standby state. A process dispatch method characterized by starting a process. 3. When executing a large number of processes including a real-time process that should be executed when a predetermined event occurs, when the predetermined event occurs, the currently executing process enters a standby state. Alternatively, the process dispatch method is characterized in that the real-time process is activated when an elapsed time from the occurrence of the predetermined event reaches a preset allowable delay time. 4. The plurality of processes include a plurality of time-sharing processes to be executed periodically, and a normal execution time and a normal execution time more than the normal execution time are provided for each of the plurality of time-sharing processes. If a short carry-over boundary time is set and the predetermined event has not occurred, each of the plurality of time division processes is sequentially and repeatedly executed for the normal execution time of each time division process, and the predetermined event When there is a predetermined time-division process that is being executed and has passed the carry-over boundary time or more since the start of execution, the execution of the predetermined time-division process is interrupted and the real-time process is executed. When the time-sharing process is started up and ready to be executed, another time-sharing process that should be executed is executed after the predetermined time-sharing process, and the execution is interrupted. 4. The process dispatch method according to claim 2, wherein the shortage of the normal execution time of the predetermined time division process is carried over to the time of executing the predetermined time division process again. 5. When the predetermined time division process is being carried out by carrying over the shortage of the normal execution time caused by interrupting the execution of the predetermined time division process in the past, 5. The process dispatch method according to claim 1 or 4, wherein the priority is set higher than that at the normal time. 6. When a plurality of processing units execute a plurality of processes in parallel, the carry-over boundary time of the time-sharing process executed by at least one processing unit of the plurality of processing units is set to another value. The process dispatch method according to claim 1 or 4, wherein the process dispatching method is changed according to the execution status of the time division process in the processing unit.