JP4857325B2 - Task switching device, method and program - Google Patents

Description

  The present invention relates to task switching in an operating system, and more particularly to a task switching device that switches tasks to be executed in a processor by switching time slots to which tasks are assigned.

  The main functions of the operating system are hardware management, task management, data management and input / output management. In particular, task management manages task execution order, and is an important function for efficiently operating a CPU, memory, input / output device, and the like.

  A task is a control unit that collectively manages the flow of program activation, execution, and termination. A program operating under the management of the operating system is treated as a task, and all operations of the operating system relating to program execution are performed in units of tasks.

  One of the algorithms for determining the task execution order is a time-sharing scheduling method or a switching method based on priority. The time-sharing scheduling method is a method in which an execution time in a task is assigned, the execution right of the processor is given to the task during the assigned time, and the execution right is transferred to another task after the assigned time has passed. As a result, execution rights are assigned to all tasks at an equal and determined time.

Regarding the switching method based on priority, both Patent Documents 1 and 2 disclose a scheduling device that switches tasks according to task priority.
JP 2000-20323 A Japanese Patent Laid-Open No. 4-101233

  However, according to the above prior art, in order to ensure the required performance of each task, it is difficult to design a program in that the programmer specifies individual task priorities, and once it is designed, the flexibility of change is poor. There is a problem.

  In order to solve the above problems, an object of the present invention is to provide a task switching device that facilitates program design that specifies priorities for satisfying the required performance of each task and has flexibility in program design. .

  In order to achieve the above object, a task switching device of the present invention is a task switching device that switches execution of a task assigned to a time slot by switching the time slot in a processor, and a pointer that associates a task with a time slot. And time slot information storage means for storing the allocation times of the plurality of first type tasks having designation of allocation time, and task storage for storing the priorities of the plurality of second type tasks having priority And for each of a plurality of time slots in one cycle, each of the plurality of tasks of the first type is assigned to one time slot in a one-task-to-one time slot based on the pointer, Multiple tasks ahead in multi-task vs. 1 time slot based on the pointer Based on an allocation unit allocated to one specific time slot to which a first type task is not allocated, and an allocation time specified for each of the plurality of first type tasks stored in the time slot information storage unit Time slot switching means for switching the time slot, and when the time slot is switched to a time slot other than the specific time slot, the task assigned to the time slot is selected and switched to the specific time slot. A task selecting unit that selects one task from a plurality of second type tasks based on the priority stored in the task storage unit.

  According to this configuration, as a result of the first type task being assigned a one-to-one time slot, the first type task is selectively executed at least once within one period including all time slots. Continuous processing performance is guaranteed. On the other hand, the second type task is assigned many time-to-one and one time slot is allocated, so that the processing performance is not continuously guaranteed. As a result, the programmer need only classify the task into tasks of the first type that need to satisfy the processing performance without considering the priority. In addition, the programmer may classify tasks that do not require processing performance into the second type. As a result, it is not necessary to assign priorities to all tasks individually, program design for ensuring processing performance can be facilitated, and program design flexibility can be ensured.

  Here, the second type task may have a priority, and the task selection unit may select one task according to the priority among the plurality of second type tasks.

  According to this configuration, since the second type task is executed from the highest priority, the programmer does not need to consider the priority for the first type task and does not require continuous processing performance. Since tasks may be classified into the second type and given priority, program design and modification can be facilitated.

  Further, the assigning means is configured so that a remaining time obtained by subtracting a total time of time slots to which the first type task is assigned from a predetermined period of time is used as the time of the specific time slot. Also good.

  According to this configuration, since the second type task is executed in the remaining time, it is possible to eliminate as much as possible that the execution of the second type task affects the processing performance of the individual tasks of the first type. Can do.

  Furthermore, the allocation means may be configured to recalculate the remaining time every time a new task of the first type is allocated to the time of the specific time slot.

  According to this configuration, it is possible to dynamically allocate the maximum remaining time to the second type task while guaranteeing the processing performance of the individual tasks of the first type.

  In addition, the first type task is a task having designation of an allocation time, and when the allocation means intends to add a new task of the first type, the total allocation time of the tasks already allocated and the new allocation time are newly added. If the total sum with the allocation time of a task exceeds one cycle time, the time slot allocation of the new task may be rejected.

  According to this configuration, it is possible to guarantee the processing performance of the existing first type task until the addition of a new first type task is eliminated.

  In addition, the task switching device includes storage means for storing lock information indicating whether or not a resource is locked by access from any task, and a task being executed is locked. Change means for changing the task from an executable state to a waiting state when attempting to access a resource in a state, and changing the task in a waiting state to an executable state when the locked state of the resource is released, The task selection means may be configured to exclude a task in a waiting state from a selection target.

  The task switching device may further include a transition unit that shifts the processor to a power saving state when there is no task that is executable among the first type task and the second type task. Good.

  Here, the processor includes at least two register sets for holding a context of a task, and the task switching device further provides one of the register sets for use of an executing task, and another one of the register sets. It is also possible to provide a switching means for returning the context of the task to be executed next to the register set by background processing and switching the register set when switching the time slot.

  According to this configuration, it is only necessary to switch register sets instead of programmatically saving and restoring contexts, so that task switching can be speeded up. Further, since the context of the task to be executed next is restored by background processing, it is possible to speed up task switching by efficiently using idle processing of processor time.

  Further, the task switching method and the program thereof according to the present invention have the same configuration as described above and have the same operations and effects.

  As described above, according to the task switching device of the present invention, the programmer only needs to classify the task into tasks of the first type that need to satisfy the processing performance without considering the priority. Well, tasks that do not require processing performance may be classified into the second type. This eliminates the need to assign priorities to all tasks individually, facilitates program design to ensure processing performance, and ensures program design flexibility. There is.

  Further, since the second type task is executed from the task with the highest priority, the programmer does not need to consider the priority with respect to the first type task, and the task that does not require continuous processing performance is the second. What is necessary is just to classify | categorize into a type and to give a priority, and program design and a change can be made easy.

〔overall structure〕
FIG. 1 is a block diagram showing a configuration of a main part of a program execution device that performs task switching according to an embodiment of the present invention. This figure schematically shows functions realized by executing software for performing task switching as a part of functions of the operating system in the processor. The hardware configuration of the processor may be a general one.

  This program execution device handles two types of tasks, a task with an assigned time and a task with a priority, in one processor, ensuring the required performance of the former task, and selecting the latter task according to the priority Configured to run.

  As shown in the figure, the program execution device includes a task scheduling unit 10, a program storage unit 20, a timer control unit 30, and an execution control unit 40.

  The task scheduling unit 10 schedules two types of tasks. That is, for the former task, the task scheduling unit 10 assigns one time slot to each task having an assigned time specification, and selects each task at least once within a predetermined period. I do. For the latter task, the task scheduling unit 10 performs scheduling for selecting one task among a plurality of tasks having priority in the remaining time of the total time of the time slots in one cycle. Here, the time slot is obtained by dividing the time for executing the program in the processor for each allocated time in the cycle.

  The program storage unit 20 stores a program that is a main body of a task to be scheduled by the task scheduling unit 10 and information related to the program.

  Each time the allocation time is set by the task scheduling unit 10, the timer control unit 30 starts counting time and outputs a timeout when the allocation time is reached. This timeout is notified to the task scheduling unit 10 in order to notify the time slot switching timing.

  The execution control unit 40 executes the task selected by the task scheduling unit 10. The execution control unit 40 corresponds to hardware that executes a processor task.

[Configuration of Task Scheduling Unit 10]
The task scheduling unit 10 includes a task reception unit 11, a time slot storage unit 12, a task storage unit 13, a time slot switching unit 14, a task selection unit 15, a mutex storage unit 16, and a mutex control unit 17.

<1. Task reception unit 11>
The task reception unit 11 receives a task addition request in accordance with an instruction from a user operation or a user program, and reads “task information”, “priority”, and “allocation time” as information about the task from the program storage unit 20. Thus, time slot information and task management blocks are generated so that the task becomes a target of scheduling, and set in the time slot storage unit 12 and the task storage unit 13.

  Here, “task information” includes a program start address and a stack pointer. The program start address is the first address where the task is written. The stack pointer is position information indicating a storage location for temporarily saving the task state when the task is switched.

  The “priority” is a standard for distinguishing between a task having an assigned time designation and a task having a priority, and has two meanings indicating the priority in the latter case.

  In other words, a task with priority 0 is a task having an assigned time (hereinafter referred to as a type A task), and a task having a priority other than 0 is a task having a priority (hereinafter referred to as a type B task). have.

  The type A task is a task having designation of an allocation time, and is allocated in a one-to-one correspondence with the time slot in the task scheduling unit 10, and is reliably executed during the allocation time of the time slot. Since the processing performance is determined by the length of the execution time, a so-called time-driven task should be classified as type A. For example, it is desirable to set a type A task that requires constant processing performance such as decoding / encoding processing of moving image data and decoding / encoding processing of audio data.

  A type B task is assigned to one time slot together with other type B tasks, and a task with a higher priority is executed during the assigned time held by the time slot. Here, the priority is assumed to take values of 1, 2, 3,. A so-called event-driven type task should be type B. For example, it is desirable that a task that does not require constant processing performance, such as a process of displaying a menu consisting of characters and still images with a user operation as an event, but that occurs occasionally and occasionally occurs is type B. In this case, the priority may be set higher for a task that processes an event that requires a strict response speed, for example.

  “Allocation time” is valid only when the task is type A, and is a value that specifies the allocation time of the time slot corresponding to the task. It is assumed that the allocation time and the priority are specified by the program.

  In addition, the task reception unit 11 receives the task stored in the time slot storage unit 12 and the task storage unit 13 when there is an instruction from a user operation or a user program, or when the execution of the task is finished. Delete time slot information and task management blocks.

<2. Time slot storage unit 12>
The time slot storage unit 12 stores time slot information for generating a time slot serving as a reference for task switching.

  FIG. 2 is a diagram showing a specific example of time slot information in the time slot storage unit 12 and a task management block in the task storage unit 13. As shown in the figure, the time slot storage unit 12 stores a plurality of pieces of time slot information 100, 110, 120,. The time slot information 100 indicates a time slot to which a plurality of type B tasks are assigned. The time slot information 100 may be generated in advance even if there is no type B task. Other time slot information 110, 120,... Indicates time slots to which type A tasks are assigned.

  The time slot information 100 corresponds to one of the time slots and includes an allocation time 100a, an execution flag 100b, and a pointer 100c. The same applies to other time slot information.

  The allocated time indicates the time that the task corresponding to the time slot can be executed. When the time when the task is actually executed reaches the allocated time, it is switched to the next time slot. The allocation times 110a and 120a corresponding to the type A task are set simultaneously with the generation of the time slot information. The allocation time 100a corresponding to a plurality of type B tasks is the remaining time obtained by subtracting the total time of the allocation times 110a, 120a,... Of other time slots from the time of one cycle held in the cycle register 18. This remaining time changes each time a task is added or deleted.

  The execution flag indicates whether the time slot is valid or invalid. It is set to valid when generating time slot information, set to invalid when the resource being accessed is locked and waiting during task execution, and set to valid when returning from the wait state to the ready state. Is done. When the execution flag indicates invalidity, the time slot switching unit 14 regards that the time slot information does not exist.

The pointer indicates a task management block corresponding to the time slot.
These time slot information 100, 110, 120... Constitute an array, and the order represents the generation order of the time slots.

<3. Task storage unit 13>
The task storage unit 13 stores task management blocks 200, 201,... Corresponding to tasks assigned to time slots. The task management blocks 200, 201,... Correspond to one task and are information for managing the task.

  The task management block 200 includes task information 200a, a priority 200b, and a link address 200c.

  The task information 200a includes a program start address (or an address at which the program should be resumed) and a stack pointer. When the task management block is generated, the task information input by the task reception unit is held. When switching tasks, the execution address of the interrupted task and the value of the stack pointer at that time are shown.

The priority 200b indicates the priority of the task.
The link address 200c holds a pointer to a task management block which is the next element in the execution queue or wait queue when the task is connected as an element to the execution queue or wait queue.

  Specifically, in the task management block corresponding to the type A task, the link address (210c, etc.) is null (invalid) when the task is in an executable state, and the task is in a waiting state. Is a pointer to the next element in the wait queue. In the task management block corresponding to the type B task, the link address 200c is a pointer indicating the next element forming the execution queue when the task is in an executable state, and when the task is in a wait state. A pointer to the next element that forms the wait queue.

<4. Time slot switching unit 14>
The time slot switching unit 14 switches the time slot when the execution time of the task reaches the allocated time in the current time slot. Whether the allocation time has been reached is notified by a timeout by the timer control unit 30. Upon receiving the notification, the time slot switching unit 14 selects the next time slot information. In this embodiment, the time slot information to be selected next depends on the arrangement order. The time slot switching unit 14 acquires the allocated time from the selected time slot information and sets it in the timer control unit 30. As a result, counting of the allocation time of the next time slot is started.

<5. Task selection unit 15>
When the time slot switching unit 14 switches the time slot and there is an instruction from the program being executed, the task selection unit 15 displays the execution address, stack pointer, and the like of the task currently being executed. As task information, the task information is extracted from the task management block of the task to be executed next, and output to the execution control unit 40. At the same time, the context (register data, etc.) of the task being executed is saved, and the context of the task to be executed next is restored. As a result, the task to be executed next becomes an execution state.

  FIG. 4 is an explanatory diagram showing how tasks are switched by the time slot switching unit 14 and the task selection unit 15. In the figure, T is a period held in the period register 18, t0 is a time slot corresponding to the time slot information 100 corresponding to the type B task, and the length thereof indicates an allocation time. t1 to tn are time slots corresponding to n time slot information 110, 120,... corresponding to n tasks of type A, respectively, and the length thereof indicates an allocation time. S indicates a scheduling process in which task switching is performed by the time slot switching unit 14 and the task selection unit 15. As shown in the figure, in the scheduling process at the start of the time slot t0, the task B1 having the highest priority is selected and executed from the plurality of type B tasks B1, B2, and B3. In the scheduling process at the start of time slot t1, task A1 corresponding to time slot information 110 is executed. The same applies to time slots t2 to tn. In this way, the type A task is always executed once during one cycle T.

  FIG. 5 is an explanatory diagram showing task switching when a task being executed in time slot t0 ends. In the figure, when the task B1 completes its own processing, the task B1 notifies the task selection unit 15 accordingly. With this notification, task B1 deletion processing (E in the figure) and task switching processing (S1) are performed. In the deletion process E, since the task B1 is completed, the task management block of the task B1 is deleted. In the switching process S1, the task selection unit 15 selects and executes the task B2 having the highest priority after the task management block of the task B1 is deleted. In this way, in the time slot t0 to which a plurality of type B tasks are assigned, tasks are executed in order from the task with the highest priority.

<6. mutex storage unit 16>
Mutex (mutual exclusion) refers to a mechanism that performs arbitration when two or more tasks access one resource by competing, and is a function included in the mutex control unit 17 and includes a resource lock state and This is realized by an object (program) that manages the unlocked state. This mechanism is provided for each resource that may compete for access from two or more tasks. When a task succeeds in the mutex lock operation, the mutex (or resource) is locked and waits until it is unlocked even if another task performs the lock operation. As a result, the lock operation succeeds and the task can occupy the corresponding resource. When a task that has performed a lock operation performs a lock release operation, another task can perform the lock operation. When a task being executed accesses a resource, the task can access the resource only when the mutex corresponding to the resource is successfully locked. If it is already locked, it will wait until it is unlocked.

  FIG. 3 is a diagram showing mutex management blocks 300, 301... Stored in the mutex storage unit 16 and a waiting queue stored in the task storage unit 13. As shown in FIG.

  In the figure, mutex management blocks 300, 301,... Are provided corresponding to individual resources with which access from two or more tasks may compete.

The mutex management block 300 includes a lock variable 300a and a pointer 300b.
The lock variable 300a indicates one of two states (1: locked state, 0: unlocked state). The initial value is 0, and it is necessary for the task to perform a lock operation to change from the unlocked state to the locked state. To change from the locked state to the unlocked state, only the task that performed the locking operation is allowed to operate.

  The pointer 300b points to the task management block at the head of the wait queue indicating the task that performed the lock operation in the lock state (the task that failed the lock operation). The task management blocks 200 ′, 201 ′,... Connected to the wait queue correspond to the tasks for which the lock operation has failed, and wait queues connected in order by the link addresses 200′c, 201′c,. Form.

<7. mutex control unit 17>
The mutex control unit 17 performs a lock variable operation when a mutex locking operation and unlocking operation are performed, a task management block operation request to the task selection unit 15, a notification to the time slot switching unit 14, and Request execution flag operation.

  FIG. 6 is an explanatory diagram showing a state where the type B task has failed in the lock operation. This figure shows a case where a lock operation (L: lock process in the figure) is performed on the mutex corresponding to the resource to be accessed by the task B1 being executed in the time slot t0 and fails. If the mutex is already locked, the task B1 fails to perform the lock operation. Therefore, the task B1 is moved from the execution queue of the task storage unit 13 to the waiting queue by the lock process L as shown in FIG. The task is switched to task B2 by S1. Since the task B1 is excluded from the execution queue until the mutex changes from the locked state to the unlocked state, the task B1 is excluded from the scheduling target.

  In this way, in the time slot t0 corresponding to the type B task, when the task being executed is in a waiting state, the task is switched to the next task.

  When task B1 succeeds in the lock operation in the lock process, execution is continued (accessing resources) during time slot t0 without switching to task B2.

  FIG. 7 is an explanatory diagram showing a state when the type A task fails in the lock operation. In the figure, when the type A task A2 fails in the lock operation and is moved to the waiting queue, the time slot is switched by the scheduling process (S in the figure) even if the allocation time of the time slot t2 has not been reached. Since only one task is assigned to the time slot corresponding to the type A task, the time slot is switched even if there is a remaining time slot.

  Further, when the task A2 succeeds in the lock operation, the task A2 is switched to the next time slot when the allocation time of the time slot t2 is reached.

  FIG. 8 is an explanatory diagram showing a situation when a task unlocks. As shown in the figure, when the currently executing task A1 unlocks the locked mutex, the task B1 in the waiting queue R corresponding to the mutex is returned to the execution queue B by the unlocking process R. As a result, task B1 is executed when the corresponding time slot t0 is selected thereafter.

  FIG. 9 is another explanatory diagram showing a situation when the task unlocks. In the figure, when the task A2 being executed unlocks the mutex that is in the locked state, the task B2 in the waiting queue S corresponding to the mutex is returned to the execution queue B by the unlock process R. Unlike FIG. 8, task B2 is returned between tasks B1 and B3. This is because they are returned to the execution queue so that they are arranged in order of priority. This is the case, for example, when task B1 with higher priority is generated while task B2 is in the waiting queue. Thereafter, when the time slot t0 is selected, the task B1 having the highest priority is selected and executed.

[Details of processing]
Hereinafter, as details of processing in the program execution device of the present invention, scheduling processing (S in FIGS. 4 to 9), mutex lock processing (L in FIGS. 6 and 7), and mutex lock release processing (FIGS. 8 and 9). R), task switching processing (S1 in FIGS. 5 and 6), task addition processing, and task end processing (E in FIG. 5) will be described in detail.

<1. Scheduling process>
FIG. 10 is a flowchart showing details of the scheduling process. The scheduling process of FIG. 6 is called and started when the timer control unit 30 notifies the time slot switching unit 14 of a timeout and when the task being executed fails in the lock operation (see FIG. 7). In the figure, i represents a variable for counting time slot information in which the execution flag is invalid, and (n + 1) represents the number of existing time slot information.

  First, the time slot switching unit 14 initializes a variable i used as a counter to 0 (S401), and saves the state of the currently executed task in the task management block (S402).

  Next, the task selection unit 15 adds 1 to the variable i (S403), selects the next time slot information in the time slot information array, and if there is no next element, the task selection section 15 starts the time slot information array. Are selected, and the selected time slot information is acquired (S404). Further, the time slot switching unit 14 takes out the allocated time included in the selected time slot information and sets it in the timer control unit 30 (S405). As a result, the timer control unit 30 starts counting and counts until the allocated time is reached and time-out occurs.

  Further, the time slot switching unit 14 determines whether or not the acquired time slot information is the time slot information 100. That is, it is determined whether the task to be executed next is type B or type A (S406).

  If the determination result is type A, the task management block pointed to by the pointer in the time slot information is acquired (S407). If the execution flag in the time slot information is 1 (valid) (S408), According to the task information in the task management block, the state of the task is restored (S411). Thereby, the task is switched to the type A task.

  When the determination result is type B, the time slot switching unit 14 determines whether the execution queue of the task management block is connected to the pointer in the time slot information 100 (whether the pointer is valid or invalid). Is determined (S412). If it is determined that the execution queue is connected, the first task management block with the highest priority is acquired (S413), and the state of the task is restored according to the task information in the task management block ( S411). As a result, the type B task is switched to the highest priority task. If it is determined that the execution queue is not connected, there is no type B task that can be executed. In this case, the process returns to S403 to select the next time slot information.

  If the execution flag is invalid in S408 (S408: no), the time slot switching unit 14 returns to S403 and S404 again (S409: no) and selects the next time slot. Further, when n + 1 task management information is invalid due to this repetition (S409: yes), that is, when there is no task that can be executed, the program execution device is shifted to the power saving state (S410). ).

  As described above, in the scheduling process by the task scheduling unit 10, the type A task and the type B task are differently scheduled. That is, when a type A time slot (a time slot corresponding to time slot information 1, 2,...) Is selected, the task scheduling unit 10 assigns only one type A task to the time slot. Switch. Further, when a time slot to which a type B task is allocated (a time slot corresponding to the time slot information 100) is selected, the task scheduling unit 10 has the highest priority among the plurality of type B tasks. Switch to.

  Further, the switching process S1 shown in FIGS. 5 and 6 may be substantially the same as the scheduling process of FIG. For example, the switching process S1 may start from S412 in the scheduling process of FIG.

<2. Task addition processing>
FIG. 11 is a flowchart showing details of the task addition processing in the task scheduling unit 10. This task addition process is started when a task addition request from a user operation or a user program occurs.

  In response to the task addition request, the task reception unit 11 acquires task information, priority, and allocation time of the task to be added (S501), and the type of the task to be added is type A or type B. (Whether the priority is 0 or not 0) (S502).

  When it is determined that the task to be added is type A, the task reception unit 11 checks whether a time slot can be added. That is, the task reception unit 11 reads the allocation time in the time slot information 110, 120,... Corresponding to the type A task from the time slot storage unit 12, calculates the total, and adds it to the total If the sum total of the allocation time of the task to be added does not exceed the period value held in the period register 18, it is determined that a time slot corresponding to the task can be added (S503). If the period value is exceeded, an error for the addition request is returned (S504).

  If it is determined that the task can be added, the task reception unit 11 generates time slot information in which an allocation time and an execution flag (0 at this time) are set (S505), and includes a task including task information and priority. A management block is generated (S506).

  After the time slot information and the task management block are generated, the task receiving unit 11 sets the storage position of the task management block to the pointer of the time slot information in order to indicate the correspondence relationship (S507). Finally, the execution flag of the time slot is validated (S508).

  If it is determined in S502 that the task to be added is type B, the task reception unit 11 generates a task management block for the task to be added (S509) and adds it to the execution queue corresponding to the time slot information 100. To do. In addition to the execution queue, the task management block is added by searching for a position to be added so that the execution queues are arranged in descending order of priority (S510). Further, 1 is set to the execution flag in the time slot information 100 (S508).

  As described above, in the addition process, when the task to be added is type A with respect to the existing time slot information and task management block (see FIG. 2), new time slot information and a new task management block are added. If the task to be added is type B, a new task management block is added at a position corresponding to the priority of the execution queue.

  In the task deletion process E shown in FIG. 5, when the task to be deleted is type A, the time slot information and the task management block corresponding to the task are deleted, and the task to be deleted is type B. In this case, the task is deleted from the task management block execution queue corresponding to the task.

<3. mutex lock processing>
FIG. 12 is a flowchart showing details of the lock processing by the mutex control unit 17. This lock process is the lock process L shown in FIGS. 6 and 7, and is called by the executing task to lock the mutex corresponding to the resource immediately before the executing task accesses one of the resources. It is processing.

  When called by the task being executed, the mutex control unit 17 reads the lock variable and writes the valid value “1” to the mutex management block corresponding to the resource (S601). In order to prevent a contradiction in value caused by other tasks performing the same processing between the reading and writing, reading and writing are performed with one instruction. This can be achieved by a read-modify-write instruction implemented in the processor.

  If the read lock variable is 0 (unlocked state) (S602: yes), the lock operation is successful, and “1” has already been set by the above writing, and the lock process is terminated. As a result, execution control returns to the task that called the lock process, and the task can access the resource corresponding to the mutex exclusively.

  If the read lock variable is 1 (locked state) (S602: no), the resource is already occupied by another task, so the calling task accesses the resource until it is unlocked. I can't. In this case, the mutex control unit 17 determines whether the calling task is type A or type B.

  If the determination result is type A (priority is 0) (S603: yes), the execution flag of the corresponding time slot information is invalidated (S604). As a result, the time slot information is excluded from the scheduling target. In addition, the mutex control unit 17 removes the task management block of the calling task from the execution queue and connects it to the wait queue of the mutex management block (S605). The reason for connecting to the waiting queue is to indicate which mutex is waiting for unlocking (which resource is waiting for releasing the exclusive state). Further, the mutex control unit 17 calls the scheduling process S (S606). As a result, the time slot is forcibly switched to the next time slot even during the allocated time (see FIG. 7).

  Instead of invalidating the execution flag in S604, the time slot information pointer may be changed to point to the head of the type B task execution queue. As a result, the type B task is also executed in the time slot.

  If the determination result is type B (priority is not 0) (S603: no), the mutex control unit 17 removes the task management block of the calling task from the execution queue (S607). As a result, the task is excluded from the scheduling target. Further, the task management block is added to the wait queue of the mutex management block to indicate which resource is waiting for the exclusive state to be released (S608). Further, the task switching process S1 is called to switch to the next rank task (S609). As a result, the task of the next order is executed in the remaining time of the time slot allocation time (see FIG. 6).

<4. Mutex unlock processing>
FIG. 13 is a flowchart showing details of the mutex lock release processing by the mutex control unit 17. This process is the lock release process R shown in FIGS. 8 and 9 and is called by the currently executing task when the currently executing task has finished accessing one resource.

  When called, the mutex control unit 17 takes out the first task management block from the wait queue of the mutex management block corresponding to the resource (S701), and determines the task in the wait state from the priority in the extracted task management block. The type is determined (S702). If the priority is 0 (type A), the mutex control unit 17 sets the execution flag in the time slot information corresponding to the extracted task management block to 1 (valid) (S703). The time slot information corresponding to the extracted task management block includes, for example, task management having the same task information as the extracted task management information among the task management blocks connected to the time slot information whose execution flag is 0. What is necessary is just to specify by searching a block. Further, the mutex control unit 17 writes 0 in the lock variable in the mutex management block (sets the lock release state) (S705).

  If the priority of the task management block taken out from the waiting queue is not 0 (type B), the mutex control unit 17 adds the task management block to the execution queue of the type B task. The addition to the execution queue is performed at the positions arranged in descending order of priority from the top (S704), and 0 is written to the lock variable in the mutex management block.

  In this way, the task that has been in the waiting state returns to the executable state, so that it becomes a target of scheduling, and the resource lock operation can be performed again when it enters the executing state next time.

  As described above, according to the program execution device of the present embodiment, the type A task is assigned one-to-one time slots, so that it is always selected and executed within one cycle, so that the processing performance is continuously improved. Can be guaranteed. On the other hand, a type B task is assigned one time slot in a many-to-one manner. As a result, the processing performance is not guaranteed continuously, but the task is executed from the one with the highest priority. As a result, the programmer does not need to consider the priority for tasks that need to continuously satisfy processing performance, and only needs to classify them as type A. For tasks that do not require processing performance, type B What is necessary is just to classify. As a result, it is not necessary to assign priorities to all tasks individually, program design for ensuring processing performance can be facilitated, and program design flexibility can be ensured.

  In the above embodiment, the example in which the designation of the allocation time is included in the task information has been described. However, the task reception unit 11 may be configured to receive the designation of the allocation time from the outside. Further, when the allocation time is not designated, the default allocation time may be used.

  In addition, although one time slot is assigned to one task for a type A task, two or more time slots may be assigned to one task.

  Furthermore, although one time slot is assigned to a plurality of tasks for the type B task, it may be assigned to any one of two or more fixed time slots. For example, a task with an odd priority may be assigned to time slot A and an even task may be assigned to time slot B.

  In addition, an owned task column that records the name of the locked task is provided in the mutex management block, and when the lock is released, the owned task column is rewritten with the task name taken from the wait queue, and the lock variable remains valid. Good. In this case, the task stored in the owned task column does not need to be locked, and the resources can be used reliably.

  Finally, a modification of the program execution device is shown in FIG. Compared with the configuration of the program execution apparatus shown in FIG. 1, the difference is that eight register sets RS0 to RS7 that can be switched simultaneously with time slot switching are added.

  One register set corresponds to one time slot and holds a task context.

  Since the plurality of type B tasks correspond to one register set, when switching to the time slot of the type A task, the task scheduling unit 10 saves and restores the register set RS0 as in the above embodiment. .

  Since the type A task corresponds to one register set, when switching to the time slot of the type A task, the task scheduling unit 10 switches the register set instead of saving and restoring the context. In this case, it is not necessary to save and restore the context, so that task switching can be speeded up. Thus, the period of the period register 18 and the time slot allocation time can be extremely shortened. For example, the time slot time of type A can be shortened to the time required for saving and returning, and one cycle can be shortened.

  Note that the number of register sets in FIG. 14 may be an arbitrary number. When the number of time slots is larger than the number of register sets, for example, one of the register sets is shared by a plurality of time slots, and the register set On the other hand, the context may be saved and restored when switching the time slot.

  In FIG. 14, the number of register sets may be two. In that case, one of the register sets may be a table register set used by the task being executed, and the other one may be a back register set for restoring the context of the task to be executed next by background processing. Good. According to this configuration, the context can be restored by background processing using the idle state, and the speed can be increased by efficiently switching between the two register sets of the back side and the front side.

  Furthermore, when the number of register sets is 2, if there is a task that has returned from the wait state to the executable state due to unlocking, and that task is the next task to be executed, the context that has returned to the back register set is It may conflict with the context of the next task to be executed. The following configuration may be used for this. That is, a step of setting a notification flag in the task management block of the task informing that there is a task that has returned from the waiting state to the executable state immediately before S705 in the unlock processing (see FIG. 13), In the scheduling process (see FIG. 10), if the notification flag indicates that the task has returned from the wait state immediately before S411, the context of the task is returned to the back register set and then the back of the register set. And a step for switching between the table and the table may be added.

It is a block diagram which shows the structure of the principal part of the program execution apparatus which performs task switching in embodiment of this invention. It is a figure which shows the specific example of the time slot information in a time slot memory | storage part, and the task management block in a task memory | storage part. It is a figure which shows the mutex management block memorize | stored in a mutex memory | storage part, and the waiting queue memorize | stored in a task memory | storage part. It is explanatory drawing which shows the mode of the task switching by a time slot switching part and a task selection part. It is explanatory drawing which shows the mode of task switching when the task of type B is complete | finished. It is explanatory drawing which shows a mode when the type B task fails in locking operation. It is explanatory drawing which shows a mode when the type A task fails lock operation. It is explanatory drawing which shows a mode when a task cancels | releases a lock | rock. It is another explanatory view showing a situation when the task unlocks. It is a flowchart which shows the detail of a scheduling process. It is a flowchart which shows the detail of a task addition process. It is a flowchart which shows the detail of the lock process by a mutex control part. It is a flowchart which shows the detail of the mutex lock release process by a mutex control part. It is a figure which shows the structure of the modification of a program execution apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Task scheduling part 11 Task reception part 12 Time slot memory | storage part 13 Task memory | storage part 14 Time slot switching part 15 Task selection part 16 Mutex memory | storage part 17 Mutex control part 18 Periodic register 20 Program memory | storage part 30 Timer control part 40 Execution control part 100 , 110, 120 Time slot information 100a, 110a, 120a Allocated time 100b, 110b, 120b Execution flag 100c, 110c, 120c Pointer 200, 201, 202, 210, 220 Task management block 200a, 201a, 202a, 210a, 220a Task information 200b, 201b, 202b, 210b, 220b Priority 200c, 201c, 202c, 210c, 220c Link address 300, 301 Mutex management block 300a, 3 1a lock variable 300b, 301b pointer

Claims (19)

A task switching device for switching execution of a task assigned to a time slot by switching the time slot in a processor,
Time slot information storage means for storing the allocation time of each of a plurality of tasks of the first type having a designation for associating a task with a time slot and an allocation time;
Task storage means for storing priorities of a plurality of second type tasks having priorities;
For each of a plurality of time slots in one cycle, each of the plurality of tasks of the first type is assigned to one time slot by one task vs. one time slot based on the pointer, and the plurality of tasks of the second type Assigning means to one specific time slot in which the first type task is not assigned in a multi-task vs. one time slot based on the pointer;
Time slot switching means for switching time slots based on the assigned time specified for each of the plurality of tasks of the first type stored in the time slot information storage means;
When the time slot is switched to a time slot other than the specific time slot, the task assigned to the time slot is selected, and when the time slot is switched to the specific time slot, the task is selected from a plurality of second type tasks. A task switching device comprising: task selection means for selecting one task based on the priority stored in the storage means. The second type of task has a priority;
The task switching device according to claim 1, wherein the task selecting unit selects one task according to a priority among a plurality of tasks of the second type. The allocating means sets a remaining time obtained by subtracting a total time of time slots to which a first type task is allocated from a predetermined period of time as a time of the specific time slot. The task switching device according to 1 or 2. 4. The task switching device according to claim 3, wherein the assigning unit recalculates the remaining time each time a new task of the first type is assigned to a time slot , and sets the time as the specific time slot. . The first type of task is a task having an assigned time designation;
When the allocation means intends to add a new task of the first type, if the sum of the allocation time of the tasks already allocated and the allocation time of the new task exceeds the time of one cycle, The task switching device according to claim 1, wherein the time slot assignment of the new task is rejected. The task switching device further includes:
Storage means for storing lock information indicating whether or not the resource is locked by access from any task for the resource accessible by the task;
Running task when trying to access the resource in a locked state, to change the task from READY state to the waiting state, changes the task wait state when the lock state of the resource is released into the executable state Change means and
The task switching device according to claim 1, wherein the task selecting unit excludes a task in a waiting state from selection targets. The task switching device further includes:
The task switching device according to claim 6, further comprising: a transition unit that shifts the processor to a power saving state when no task in an executable state among the first type task and the second type task exists. . The processor includes at least two register sets for holding a task context;
The task switching device further includes:
Switching one register set to use a running task, returning the context of the next task to be executed to another register set by background processing, and switching the register set when switching time slots The task switching device according to claim 1, further comprising: means. A task switching device for switching a task to be executed in a processor by switching a time slot to which a task is assigned,
For a time slot within one period, one time slot is assigned to one task with one task vs. one time slot for a first type task having an assigned time designation, and a task corresponding to each time slot. First generation means for generating time slot information including the allocated time of:
A plurality of tasks for a second type task having priority are assigned to a specific one time slot in which the first type task in the cycle is not assigned in a multi-task vs. one time slot, and a specific time a second generating means for generating a single time slot information including between time assignment slots,
A third generating means for generating a task management information including the address of the task for each task assigned to the time slot,
Storage means for storing the generated time slot information and task management information in correspondence with each other by a pointer;
Selecting means for selecting at least one time slot within one period based on the allocated time of the time slot information stored in the storage means;
When the time slot information to which the first type task is assigned is selected, the task indicated by the task management information corresponding to the time slot information is executed by the pointer, and the second type task is assigned. Control means for selecting one of a plurality of task management information corresponding to the time slot information according to the priority by the pointer and executing a task indicated by the selected task management information when the time slot information is selected; A task switching device comprising: The storage means stores task management information of the second type task as a queue arranged in order of priority,
The task switching device according to claim 9, wherein the control means selects a task corresponding to the task management information at the head of the queue. The second generation means sets the difference between the period and the total of the allocation time of all tasks of the first type in the time slot information corresponding to the specific time slot as the allocation time of the specific time slot. The task switching device according to claim 10. The second generation means recalculates the remaining time as the difference every time a time slot is assigned to a new task of the first type by the first generation means to obtain the assigned time of the specific time slot. The task switching device according to claim 11. The storage means further stores lock information indicating whether or not a resource accessible by a task is locked by access from any task,
The task switching device further includes:
When an executing task tries to access a resource in a locked state, the correspondence between the task management information and the time slot information of the task stored in the storage unit is disconnected, and the task management information is stored as a wait queue. The task switching device according to claim 12, further comprising queue management means for storing task management information in the waiting queue in association with time slot information when the resource is released from the locked state. The processor includes at least two register sets for holding a task context;
The task switching device further includes:
A register that uses one of the register sets for the task being executed, restores the context of the next task to be executed to the other one register set by background processing, and switches the register set when the time slot is switched The task switching device according to claim 13, further comprising set switching means. A task switching method for switching execution of a task assigned to a time slot by switching the time slot in a processor,
An allocation time storage step of storing each allocation time of a plurality of tasks of the first type having a designation for associating a task with a time slot and an allocation time;
A priority storing step for storing the priority of each of the plurality of tasks of the second type having priority;
For each time slot within one period, each of the plurality of tasks of the first type is assigned to one time slot in one task vs. one time slot based on the pointer, and the second type is different from the first type. Assigning a plurality of tasks to one specific time slot in which the first type task is not assigned in a multi-task versus one time slot based on the pointer;
A time slot switching step for switching a time slot based on the assigned time designated for each of the plurality of tasks of the first type stored in the assigned time storage step;
When the time slot is switched to a time slot other than the specific time slot, the task assigned to the time slot is selected, and when the time slot is switched to the specific time slot, the priority is given to a plurality of second type tasks. A task selection step of selecting one task based on the priority stored in the degree storage step. A program for switching execution of a task assigned to a time slot by switching the time slot in a processor,
The program is
An allocation time storage step of storing each allocation time of a plurality of tasks of the first type having a designation for associating a task with a time slot and an allocation time;
A priority storing step for storing the priority of each of the plurality of tasks of the second type having priority;
For each time slot within one period, each of the plurality of tasks of the first type is assigned to one time slot in one task vs. one time slot based on the pointer, and the second type is different from the first type. Assigning a plurality of tasks to one specific time slot in which the first type task is not assigned in a multi-task versus one time slot based on the pointer;
A time slot switching step for switching a time slot based on the assigned time designated for each of the plurality of tasks of the first type stored in the assigned time storage step;
When the time slot after switching is not the specific time slot, the task assigned to the time slot is selected, and when the time slot after switching is the specific time slot, the plurality of tasks of the second type A program that causes the processor to execute a task selection step of selecting and executing one task based on the priority stored in the priority storage step. A task switching device for switching execution of a task assigned to a time slot by switching the time slot in a processor,
Storage means for storing the priority of each of the plurality of tasks of the second type having the allocation time and priority of each of the plurality of tasks of the first type having a pointer that associates the task with the time slot and designation of the allocation time;
For each of a plurality of time slots in one cycle, each of the plurality of tasks of the first type is assigned to one time slot by one task vs. one time slot based on the pointer, and the plurality of tasks of the second type Assigning means to one specific time slot in which the first type task is not assigned in a multi-task vs. one time slot based on the pointer;
Time slot switching means for switching time slots based on an assigned time specified for each of the plurality of tasks of the first type stored in the storage means;
When the time slot is switched to a time slot other than the specific time slot, the task assigned to the time slot is selected, and when the time slot is switched to the specific time slot, the plurality of tasks of the second type store the memory. A task switching device comprising: task selection means for selecting one task based on the priority stored in the means. A task switching device for switching execution of a task assigned to a time slot by switching the time slot in a processor,
Storage means for storing respective allocation times and priorities of a plurality of second type tasks having priorities of a plurality of first type tasks having an allocation time designation;
For each of a plurality of time slots in one cycle, each of the plurality of tasks of the first type is assigned to one time slot by one task to one time slot, and the plurality of tasks of the second type are assigned to a multitask pair. Allocating means for allocating to one specific time slot to which the first type task is not allocated in a time slot;
Time slot switching means for switching time slots based on an assigned time specified for each of the plurality of tasks of the first type stored in the storage means;
When the time slot is switched to a time slot other than the specific time slot, the task assigned to the time slot is selected, and when the time slot is switched to the specific time slot, the plurality of tasks of the second type store the memory. A task switching device comprising: task selection means for selecting one task based on the priority stored in the means. A task switching device for switching execution of a task assigned to a time slot by switching the time slot in a processor,
Receiving a plurality of first type tasks having designation of allocation time and a plurality of second type tasks having priority, and assigning each of the plurality of first type tasks and a plurality of second type tasks Task accepting means for obtaining the priority of each task;
For each of a plurality of time slots in one cycle, each of the plurality of tasks of the first type is assigned to one time slot by one task to one time slot, and the plurality of tasks of the second type are assigned to a multitask pair. Allocating means for allocating to one specific time slot to which the first type task is not allocated in a time slot;
Time slot switching means for switching time slots based on the assigned time specified for each of the plurality of tasks of the first type acquired by the task receiving means;
When the time slot is switched to a time slot other than the specific time slot, the task assigned to the time slot is selected, and when the time slot is switched to the specific time slot, the priority is given to a plurality of second type tasks. And a task selection means for selecting one task based on the degree.

Images