JPH03212742A - Parallel processing control system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed Description of the Invention] [Summary] Concerning a parallel processing control method in which devices are controlled by an operating system that performs parallel processing, parallel processing programs can be created easily and efficiently, maintainability is easy, and the operation of the device is optimized. For the purpose of providing a processing control method, each task, in which the execution conditions of multiple tasks that are processed in parallel are related to each other, is divided into processing units of a predetermined length and assigned a number to each, and the task management section , a task execution condition determination means is provided corresponding to each task to set mutual activation conditions for each task to be processed in parallel and determine whether or not the task can be executed, and the task execution condition determination means is activated by an instruction from the task execution condition determination means. When activated, the task execution unit executes the divided processing with the processing execution unit and increments the corresponding unit counter, and the task execution condition determination unit increments the unit counter of each task execution unit. It is configured to perform judgment based on the value.

[Industrial Application Field] The present invention relates to a parallel processing control method for controlling a device having a mechanism section by an operating system that performs parallel processing.

2. Description of the Related Art In recent years, technology for program-controlling mechanical units that execute multiple processes in parallel has been used in various fields. In such control, there are many cases in which there is a mutual relationship between the operating states of multiple mechanical parts1. Conventional programming for this purpose uses methods such as using flags to associate operations, but the program It has become complicated9 and improvements are desired.

[Prior Art] Conventionally, controlling mechanisms that can be performed in parallel has been performed in various fields. For example, a mechanism that takes out one part (or partial device or element) from a certain location and transports it to a processing position, a mechanism that performs the first processing by controlling the transported device such as positioning, and a mechanism that carries out the first processing by controlling the transported device such as positioning, and a mechanism for transporting the
There is control such as performing another process at another position.

Such control operations are mutually related, but each mechanism can operate in parallel, and parallel processing control is performed by a program.

An explanatory diagram of a conventional example that performs such parallel processing control is shown in FIG.

Task A9, task B, and task C in FIG. 6 are programs that process each mechanical section that performs parallel operations, respectively.
The main task is a control program provided above each task A to C, and each of these programs is controlled by an operating system capable of parallel processing (for example, iRMX for 32 bits).

Previously, it was a task A program! Up to al, it can operate independently of other tasks, but after a+, if it cannot operate until another task C finishes at position C2 in the program, set a flag Fal (flag name) at position a1 to the task. When the process reaches this position, it looks at this flag, and if it is off (initially off), it stops the operation under the control of the main task. When this state is reached, the main task stops its processing and controls other tasks.5 When the other task C reaches the program position ic2, the flag Fal of task A is turned on under the control of the main task. As a result, processing of task A is restarted. Similarly, task C has a flag Fcl set in the middle of its program, and when it reaches this position, flag Fcl is set.
When cl is off, the program of task B is stopped under the control of the main task, and when the program of task B reaches position ibl, the flag Fcl is turned on.

In addition to this method of using flags to establish mutual relationships with other tasks, when a process cannot be executed until another process completes, that process is put into a "sleep" state, and other related processes are There is also a method such as putting the program into an "act" state upon termination.

[Problems to be Solved by the Invention] In the above-mentioned conventional method using flags, the main task that controls the entire task 5 has more flags as the number of child tasks (task execution units, etc.) increases. Flag control (on/off control of each flag) becomes complicated. Furthermore, the management of flags (naming flags and defining their meanings) has become complicated, requiring a lot of time for programming and subsequent debugging, and problems have arisen in that the maintainability of the program is poor. A similar situation occurred when using a sleeve.

Conventionally, no method has been found for shortening the cycle time (the time required for all tasks to operate once) of mechanical units that are controlled in parallel through such processing.

SUMMARY OF THE INVENTION An object of the present invention is to provide a parallel processing control method that allows parallel processing programs to be easily and efficiently created, maintainability is easy, and optimizes device operation.

[Means to solve the problem] FIG. 1 is a diagram showing the principle configuration of the present invention.

In FIG. 1, 10 is a task management unit, 11 is provided corresponding to each task 1, task 2...task n, and is used to determine whether or not preset conditions are satisfied. 12 is a task execution unit that executes the divided unit corresponding to task I-n; 13 is a process execution unit for each task execution unit; 14 is a process execution unit for each process execution unit; A unit counter 15 counts the number of processing units, and a unit counter 15 counts the number of processing units 1. Processing 2 represents each task divided into tasks n.

In the present invention, each task is divided into a plurality of operation units (units) each having a predetermined length, a unit counter is provided for each task, and the corresponding unit counter is counted up each time a unit of operation is completed. Then, for each task, the execution conditions of its own task are determined. The determination is made by looking at the states of the unit counter of its own task and the unit counters of other tasks, identifying whether or not a preset condition is met, and issuing an instruction to execute the task if one condition is met.

[Operation 1 In FIG. 1, each task 1-n is divided in advance by a predetermined processing amount. In that case, it is necessary to divide the program at a position where it cannot be executed until processing of other tasks is completed. In this way, each task 15 of task task n
As shown in FIG. 1, a predetermined number of processes 1. Processing 2... is divided into units.

In the task management unit 10, task execution condition determination means 11 is provided to determine the execution conditions of a plurality of tasks 1 to n that are processed in parallel, and each execution condition determination means determines whether each unit of each task is one If there is a relationship with other tasks, the execution condition of when the unit is completed is set in advance.

The task management unit IO is activated and drives each task execution condition determining means 11. Each task execution condition determining means 11 inputs the contents of the unit counter 14 of each task and determines whether or not a preset execution condition is satisfied, and when one condition is satisfied, outputs an execution instruction to the task execution unit 12. . When the processing execution means 13 starts each task execution unit,
The unit specified by the unit counter 14 at that time is processed. When processing is completed, unit counter 1
4 is counted up, and the count value is input to the task management section. Each execution condition determination means of the task management unit 10 can know not only the execution status of its own task but also the execution status of other tasks from each processing counter value. If the execution instruction for the primary unit is output and the 7 conditions are not satisfied (such as when the unit processing of a certain number in another task is not completed), the execution instruction is not generated.

In this way, each task execution unit manages the execution status of the divided tasks, and the task management unit identifies the execution status and determines the conditions for parallel processing programs. can be created efficiently and maintenance becomes easy.

[Example] Figure 2 (a) is a diagram showing the task configuration, Figure 2 (b) is the relationship between the processes of each divided task, Figure 2 (C) is a task transition diagram, and Figure 3 4 is a processing flow diagram of the main task of the embodiment, FIG. 4 is a processing flow diagram of task C, and FIG. 5 is a definition table of unit counter values.

The task configuration in FIG. 2(a) includes one main task M provided on a microprocessor and memory (not shown), which controls multiple tasks A to C, and tasks corresponding to multiple mechanical units constituting a control device (not shown). It consists of tasks A to C (referred to as child tasks to the main task). In addition,
The plurality of mechanical units of the control device perform input/output operations in response to the execution of each child task A to C.

Each task is divided into a plurality of units each having a predetermined length. In this example, task A is processed 1. Process 2 is divided into two, task B is divided into four processes 1 to 4, and task C is divided into three processes 1 to 3. The principle of dividing in this way will be explained using FIG. 2 (bl) and FIG. 2 (C1).

Analyzing the interrelationship of each task operation within one cycle of the control device, if the relationship is as shown in Figure 2 (b),
That is, first, task A, task B, and task C start processing in parallel, and once task A executes a unit of process 1, it becomes unable to proceed any further, and task C also executes a unit of process 1. Then, the process becomes unable to proceed. Considering the relationship shown in FIG. 2(b), each task ABC is divided into processes 1 at the times shown in the task transition diagram of FIG. 2(c).

Furthermore, according to FIG.
Since the condition is such that task B can be executed, task B is then divided at the position of process 2.

Tasks A and B have positions where they will not start until task C's process 1'I2 is completed, and task A's process 2. Task B is separated as process 4.

Note that Process 3 of Task B can be executed unconditionally after Process 2 of its own task is completed.

When each task is divided in this way, the time length of each divided process and the time from the start to the end of the process are shown in FIG. 2(C). In this case, the length of each divided processing unit in Figure 2 (C) can be adjusted as appropriate depending on the mutual relationship, but if the lengths are the same for each task, unnecessary waiting time will result. This allows for optimal division.

In this example, when tasks A, B, and C are divided as shown in FIG. 2(b) and FIG. 2(C), the division unit of each task (
As a unit counter that counts U
A, UB, and UC are provided. The meaning of the count value of each unit counter is defined as shown in the definition table shown in FIG. 5. The processing flow of the main task of the embodiment using the above unit counter is shown in FIG. The processing flow of task C is shown in FIG. 4 as a representative processing flow.

The processing flow of the main task shown in FIG. 3 will be explained.

First, reset the counters UA, UB, and UC (hereinafter simply referred to as UA, UB, and UC) to 0 (
Step 30) Wait. When there is a command to start processing (step 31), 1 is placed in each unit counter (
Step 32). In this state, each task enters the activated state, and in step 3, the conditions of each task A to C are determined.
3,3947 are processed in parallel. This processing is performed under the control of an operating system capable of parallel processing.

To explain the processing flow regarding UC (for task C) in steps 47 to 52, first uc=e o
o or o is determined. This is to determine whether or not the process has ended normally according to the definition table shown in FIG. At this time, since UC=1, the answer is no, so proceed to the next step 4.
Move on to 8. Here, it is determined whether UC=1, and since it is YES, the process moves to step 52. This step
c (UC);'' and is to execute task C.

As a result, the processing of task C shown in FIG. 4 is started.

In FIG. 4, first in step 60 it is determined whether uc=i. At this time, since UC=1, the process moves to step 61 and process 1 of task C is executed. When this is completed, UC+1 is executed in step 62 (UC=2), and the process returns (to the main task). During this time, step 33
-38 and 39-46, the processing regarding UA (task A) and UB (task B) is performed in the same way.

After the process 1 of task C is completed and UC=2, it is determined in step 54 whether UA, tJB, and UC are all O. In this case, the answer is NO, so next, step 47 is performed again according to the route (■). Judgment can be made. After this, in step 48, the determination of UC=1 becomes NO, so in the primary step 49, it is determined whether tJc-2 is more than 1. In this case, the answer is yes, so proceed to the next step 50, where UB=3
Determine whether This determination is to check whether Process 2 of another task B (see FIG. 2(b) and FIG. 2(C)) has been completed.

At this time, whether processing 2 of task B has been completed or not is determined by determining the UBO value. If it is UB-3, the process advances to step 52 and the processing of task C in FIG. 4 is executed. In this case, process 2 is executed via step 63 in FIG. 4 (step 64). Thereafter, UC is counted up and becomes UC=3 (step 65), and the process returns.

Then step 54. Step 51 after step 47
When this is reached, UC=3 is determined to be YES for 1 time, and task C is executed in step 52. In this case, in the process of FIG. 4, process 3 (step 67) of task C is executed via step 66, and UC=8
000 (hexadecimal display) (step 68
).

After this, return and in step 47 UC=8
If a yes result is obtained in the determination of 000, LIC=0 in step 53.

The above-mentioned processing is similarly performed for each task Δ and B, and as shown in the steps of the main task in Figure 2 (a
), the processing of each task is executed when the conditions shown in FIG. 2(b) are satisfied.

When all the tasks A to C are completed, the counters UA, UB, and UC become 0. In this case, the determination in step 54 is YES, and it is determined whether or not the next completion is to be performed. This determination is made by setting in advance how many times the control operation will be completed, with one cycle being the time from the start of processing to the completion of each of the above tasks A to C, and when the set number has been reached. This is to determine whether or not it is true. This set number is 1. For example, when controlling the machining of one part, the number of parts is set.

If the process is not completed, the process returns to step 3I via route (3), and the same operation is repeated again based on the instruction to start the process.

In the operations shown in FIGS. 3 and 4 above, if an abnormality occurs in the processing operation of a task, the value of the unit counter corresponding to the task is set to a value of 8001 or more. In this case 3, it is possible to later detect which task's processing ended abnormally.

In this way, a task can be activated only by controlling the unit counter, and that control is also controlled by the task l! It can be managed by two transfers (Fig. 2(C)).

[Advantageous Effects of the Invention 1] According to the present invention, the operation order and parallel operations of each task are centrally managed by a unit counter, and even if the number of tasks increases, it can be easily controlled, and the task transition diagram can be used to optimize the operation. It can also be easily converted into

As a result, programming and debugging periods can be significantly shortened, and program maintainability can be improved.

[Brief explanation of drawings]

Figure 1 is a diagram showing the principle configuration of the present invention, Figure 2 (a) is a diagram showing the task configuration, Figure 2 (b) is a diagram showing the relationship between the processes of each divided task, and Figure 2 (C) is a diagram showing the task configuration. Task transition diagram, Figure 3 is a processing flow diagram of the main task of the embodiment, Figure 4 is task C
5 is a definition table of unit counter values, and FIG. 6 is an explanatory diagram of a conventional example. 10 in FIG. 1: Task management section 11: Task execution condition determination means 12: Task execution section 13/processing execution means 14: Unit counter 15: Task

Claims (1)

[Claims] In a parallel processing control method for controlling a device by an operating system that performs parallel processing, each task is processed for a predetermined length, and the execution conditions of the tasks to be processed in parallel are mutually related. The task management unit (10) divides into units and assigns numbers to each unit, and the task management unit (10) sets mutual activation conditions for each task to be processed in parallel, and determines whether the task can be executed. (11) corresponding to the tasks, and each task execution unit (12) is activated in response to an instruction from the task execution condition determining means (11), and each task execution unit executes divided processing when activated. Means (13)
A parallel processing control method characterized in that the task execution condition determining means (11) performs the determination based on the value of the unit counter of each task execution section.