JPH01154237A - Executing device for time-division task - Google Patents

Abstract

PURPOSE:To complete the execution of a time-division task within a prescribed time by adding previously priority order to the task executing queues and heightening the order of a task of a low priority order that is not executed even though it scheduled executing time passed at the time-division parallel execution. CONSTITUTION:The register files are set in the same number as the task executing queues and at the same time each register file contains a data register for scheduled task executing time. Then the corresponding data register value is set to a counter of a system. The executing priority order of the relevant task is heightened in case the system counter exceeds the scheduled task executing time of the task executing queue when said task is carried out. Thus plural tasks are carried out in time division and in parallel with each other.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention is directed to a system controlled using a microprocessor, in which two or more tasks (jobs) are executed in a time-sharing manner by one microprocessor, and the priorities are determined. Time-sharing multitasking allows tasks to be executed in multiple queues, and each task waiting to be executed is queued to a designated queue, allowing time-sharing parallel processing of tasks up to the number of queues that exist. It concerns the execution device.

Conventional technology When a task is executed using a microprocessor, a group of registers,
Many systems prepare a set of register files consisting of stack pointers, status registers, etc., and execute tasks while transferring necessary data to the register file in response to instructions from a microprocessor. However, with this method, only one task can be executed at any time, resulting in poor execution efficiency.Therefore, conventionally, a set of register files provided for one microprocessor is used in a time-sharing manner to execute multiple tasks. Methods for executing tasks in time division have been considered.The most typical method is one in which tasks are switched for each microprocessor instruction.However, in this method, two or more tasks are executed in one Since each instruction is simply switched and executed sequentially, it is not possible to prioritize execution time between multiple tasks.Furthermore, each time a task is switched, the data previously stored in the register file is
It is necessary to first save the data to the memory (stack) area and then read the data needed for the next task from another memory space into the register file. During this data switching time, tasks cannot be executed, resulting in a large time loss.

On the other hand, instead of switching tasks for each instruction,
A method is also known in which a user program manages the time of multiple tasks and tasks are switched using interrupt processing or the like. According to this method, it is also possible to prioritize the execution times of a plurality of tasks in advance when creating a program. However, setting priorities using a program requires a sophisticated understanding of the program contents, and it is difficult for users using microprocessors to create programs that meet the user's specifications and set priorities. This is extremely difficult in practice. Moreover, even in this case, since there is only one set of register files for one microprocessor, it is necessary to repeatedly save and retrieve data each time a task is switched, and therefore time loss still remains unresolved.

In order to solve these problems, multiple register files are prepared for one microprocessor.
There is also a method in which multiplexers are switched according to instructions from a microprocessor (program), and multiple tasks are executed in a time-sharing manner while sequentially switching multiple register files. In this way, since one register file is prepared for one task, there is no need to save or read data at the time of switching, and therefore time loss is reduced. However, even in this case, the settings for switching between multiple tasks and the priority order of the execution time of each task must be set by a program. Therefore, creating a program according to the user's specifications imposes a considerable burden on the user.

Problems to be Solved by the Invention As described above, conventionally, when multiple tasks are executed by one microprocessor, even if there are multiple register files, data management, execution time management, and priority management for each task are difficult. There was a problem in that the ranking etc. had to be set programmatically, which placed a heavy burden on the user.

The present invention has a register file for each task execution queue, eliminates the time loss required for saving data when switching execution tasks, and also has data of scheduled task execution time for each task execution queue, so that multiple tasks can be executed. When a task is executed in parallel using time division, if the task does not finish within the scheduled execution time, the priority of the task is automatically increased, and even when the tasks are executed in parallel, the task is This solves the conventional problem by providing a function to automatically schedule tasks so that they end their execution.

Means for Solving the Problems In order to solve the above-mentioned problems, the present invention provides register files equal to the number of task execution queues, and also provides a data register for the scheduled task execution time in each register file. The data register value is set for the counter, and if the system counter exceeds the scheduled execution time of the task in the task execution queue while the task is being executed, the execution priority of the task is increased. This allows time-sharing parallel execution of multiple tasks.

By doing this, you can prioritize the task execution queues in advance, and when multiple tasks are executed in parallel in a time-sharing manner, you can avoid situations where queues with lower priorities are hardly being executed. Even if the task is faced with a low queue and the execution is not completed even after the scheduled execution time, it is possible to automatically raise the priority and finish the task almost within the scheduled execution time.

Embodiment An embodiment of the present invention will be described in detail using the block diagram shown in the figure. This example describes a case where two task execution queues exist. In FIG. 1, register file 1 is register file A.

It consists of two parts B. Each register file includes a register group (consisting of general-purpose registers, data registers, data pointers, etc.), a stack pointer, a status register, and a program counter. One scheduled task execution time data register 2 exists for each register file, and is a data register that stores the scheduled execution end time of a task using that register file.

The system counter 3 is a counter that automatically counts based on a count clock (not shown) generated by the system. The execution queue designation circuit 4 is a circuit that determines which of register file A or register file B is used to execute a task execution queue. MPX (multiplexer) 5 is a circuit that selects which data from the task execution scheduled time data register A or B should be used. The comparator 6 compares the value of the system counter with the scheduled task execution time data register value, and the value of the system counter is
This circuit generates a signal to control the task execution queue priority determination circuit when it becomes larger than the value of the scheduled task execution time data register. The task execution queue priority determination circuit 7 is a circuit that receives the signal generated by the comparator 6 and generates a control signal to the execution queue designation circuit 4.

For example, a queue that uses register file A is queue A, and a queue that uses register file B is queue B.
shall be. Assume that the priority order is set for queue A to be executed continuously for the number of A count clocks, and for queue B1 to be executed continuously for the number of B count clocks, and that the conditions are (A count clock)>(B count clock). In this case, queue A continuously executes count clocks A times, then queue A stops, queue B continuously executes count clocks B times, and then queue A again continuously executes count clocks A times. In this way, queue A and queue B perform time-division parallel execution alternately. In this case, A and B have separate register files, so no overhead occurs when switching between queues. Here, according to the priority order shown in the above inequality, queue A has a higher priority order than queue B (it is set).

Here, if the execution of the task is completed under the conditions of (task execution time data register) < (system counter), the next task is queued with this priority setting, and the same process as the previous one is performed. Perform split parallel execution.

Next, during task execution, when the condition (task execution time data register)> (system counter) is satisfied (this condition is constantly checked by the comparator 6), the task execution queue priority determination circuit 7 The number of consecutive count clocks executed for queue B is automatically set to A times, the priorities of queue A and queue B are raised to the same level, and processing is executed. When this task is completed, the number of consecutive count clocks for queue B is automatically reset to the set number of B count clocks.

In this way, the user can automatically complete processing within the specified time by simply setting the task execution time data register 2 to indicate when the task to be executed should be completed in the system counter. . Therefore, there is no need for the user to manipulate priorities or perform special priority processing using an execution program, and the burden on the user program can be significantly reduced. Of course, in this case as well, since a register file is provided for each task, there is no need to save or retrieve data when switching tasks, and no loss occurs due to task switching (execution queue switching).

As described in detail, according to the present invention, a single microprocessor can time-share multiple tasks and set the priority of each task using a simple control circuit without placing any burden on the user program. It can be carried out.

[Brief explanation of the drawing]

The drawing is a block diagram of a time-sharing task execution device according to an embodiment of the present invention. 1... Register file, 2... Scheduled task execution time data register, 3... System counter, 4... Execution queue specification circuit, 5
...MPX (multiplexer), 6...
...Comparator, 7...Task execution queue priority determination circuit.

Claims (1)

[Claims] A device for switching tasks in a time-sharing manner using one microprocessor is provided with a plurality of task execution queues that can be prioritized, a plurality of register files corresponding to each of the queues, and each of the register files has a plurality of register files. A time-sharing task execution device comprising a data register for determining a scheduled task execution time and a counter for managing the execution time of the task execution queue.