JP2527055B2 - Shared variable detection processing method in task - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial application] The present invention detects inter-task shared variables that can be regarded as in-task shared variables from a source program written in a high-level programming language capable of describing execution by a multiprocessor. The present invention relates to a processing method for replacing with a shared variable in a task.

[Conventional technology]

In a source program consisting of multiple program units, a common block method is used to reduce the storage area allocated for multiple (arrays) and to simplify the exchange of data between program units. Mostly used. For example, the FORTRAN language uses the COMMON statement.

By the way, in a high-level programming language capable of describing execution by a multiprocessor, for example, a source program written in FORTRAN language, a variable declared by COMMON statement is shared by a plurality of tasks to which the program unit including the declaration belongs. The variable becomes a shared variable between tasks, and exclusive control is performed at the time of execution of the program in order to maintain the validity of the contents for reference / writing from a plurality of tasks.

However, not all the elements of the inter-task shared variable declared in the source program are actually shared by multiple tasks, and probably contain unnecessary declarations, which is why unnecessary exclusive control is performed. Therefore, there is a problem that the execution efficiency of the program is reduced.

To solve the above problem, it is clear that the inter-task shared variable declared in the source program is a variable that is shared only among program units within a specific task (in-task owned variable). In some cases, TASKCOMMON
It is effective to replace it with the declaration of the shared variable in the task by the statement or to remove the unnecessary declaration.

Conventionally, the above processing is performed exclusively by humans,
While checking the task configuration of the source program, the reference information lists of variables for each program unit were visually compared, and if a substantial shared variable in the task was found, the source program was modified by an editor etc. .

[Problems to be Solved by the Invention]

As described above, conventionally, processing such as detecting inter-task shared variables that can be regarded as in-task shared variables from the source program and replacing them with in-task shared variables has been left to humans, which makes the work complicated and burdens the person in charge. Has a drawback that it is heavy and takes a long time to process. Further, since it is a human work, there are cases in which detection omissions, correction omissions, etc. occur, and there is a drawback that the effect of sufficiently shortening the execution time cannot be obtained.

The present invention has been proposed in view of the above points, and an object of the present invention is to automatically perform the process of detecting an inter-task shared variable that can be regarded as an in-task shared variable from a source program and replacing it with the in-task shared variable. It is to provide a shared variable detection processing method in a task that can be performed.

[Means for solving the problem]

In order to achieve the above object, the present invention detects an inter-task shared variable that can be regarded as an in-task shared variable from a source program written in a high-level programming language capable of describing execution by a multiprocessor and replaces it with an in-task shared variable. In the processing method, a source program is input, and for each program unit, a program unit shared variable that indicates whether or not the task-to-task shared variable declared in the program unit is referenced by the executable statement in that program unit. Program unit analysis processing means for generating information, and generating program configuration information about a program unit that is started as a task from the program unit and a program unit that is called as a procedure, and its startup method, and refer to the program configuration information. The program unit It is separated for each task unit to which it belongs, and the inter-task shared variable declared in the task unit is referred to in each task unit by referring to the program unit shared variable information. Task-unit shared variable information indicating whether or not the task-unit shared variable information is referred to, and which task-unit shared variable is referred to by each task-related shared variable information by referring to the task-unit shared variable information. Inter-task analysis processing means for generating inter-task shared variable information, and referring to the inter-task shared variable information, for each program unit in the source program, among the declarations of inter-task shared variables included in the program, What can be regarded as the in-task shared variable of the task to which the unit belongs is replaced with the declaration of the in-task shared variable or the instruction line indicating that it is the in-task shared variable is displayed. Inserted into start program, which can be regarded as another task in a task shared variable is deleted from in the source program, and so and a program conversion processing means for generating a new source program.

[Action]

In the in-task shared variable detection processing method of the present invention,
The program unit analysis processing unit inputs a source program to generate program unit shared variable information and program configuration information, and the task unit analysis processing unit refers to the generated program unit shared variable information and program configuration information to share task units. Variable information is generated, and the generated task unit shared variable information is referenced by the inter-task analysis processing means to generate inter-task shared variable information, and the generated inter-task shared variable information is referenced by the program conversion processing means, Of the declarations of shared variables between tasks in the source program, those that can be regarded as shared variables within the task of the invoking task should be replaced with the declaration of shared variables within the task, or a directive line indicating that they are shared variables within the task should be inserted into the source program. , What can be regarded as the shared variable in the task of another task is deleted from the source program and a new source is created. To generate a program.

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of the in-task shared variable detection processing method of the present invention. In FIG. 1, the in-task shared variable detection processing method of the present invention is embodied as a program analysis / generation mechanism 2, and this program analysis / generation mechanism 2 can describe execution by a multiprocessor, for example, a high-level programming language. A source program 1 written in FORTRAN language is input, and a new source program 3 and shared variable information list 4 in which inter-task shared variables that can be regarded as in-task shared variables are replaced with in-task shared variables are output. . The program analysis / generation mechanism 2 has a program analysis mechanism 21 and a program generation mechanism 25 as functional units, and a configuration information area 28 and a shared variable information area 29 as information storage areas. Also,
The program analysis mechanism 21 is a program unit analysis processing means 22.
And task unit analysis processing means 23 and inter-task analysis processing means 24
The program generation mechanism 25 is composed of shared variable information editing / output processing means 26 and program conversion processing means 27. The configuration of each unit will be described through the operation described below.

FIG. 2 shows an example of the source program 1 written in the FORTRAN language.
It shall consist of 0,300,400,500.

The program unit 100 declares inter-task shared variable declaration statements 110, 111, 112 that declare inter-task shared variables AA, CC, DD, which are named common blocks, and EQUIVALENCE that declares area coupling of element DM2 and array C of inter-task shared variable DD. A statement 113, a task activation statement 120 that activates the procedure subprogram A of the program unit 300 as a task, and a procedure call statement 121 that calls the procedure subprogram D of the program unit 200,
It includes shared variable reference statements 130, 131 and 132 including references to elements AM3, DM1 and C of shared variables between tasks.

The program unit 200 is an inter-task shared variable declaration statement that declares an inter-task shared variable DD that is a named common block.
210 and shared variable reference statements 230 and 231 including references to the elements DD1 and DD2 of the inter-task shared variable.

The program unit 300 includes inter-task shared variable declaration statements 310 and 311 that declare inter-task shared variables AA and CC that are named common blocks, a procedure call statement 320 that calls the procedure subprogram B of the program unit 400, and a procedure of the program unit 500. Procedure call statement 321 that calls subprogram C, and elements CA3, CA2, CA1, AA1 of inter-task shared variables
Shared variable reference sentences 330, 331, 332, 333 including the reference of.

The program unit 400 is an inter-task shared variable declaration statement that declares an inter-task shared variable AA that is a named common block.
410 and a shared variable reference statement 430 including a reference to the element AB2 of the inter-task shared variable.

The program unit 400 includes inter-task shared variable declaration statements 510 and 511 that declare inter-task shared variables AA and CC that are named common blocks, and shared variable reference statements 530, 531 and 532 that include references to elements AC1, AC2, and CC1 of inter-task shared variables. Includes and.

Taking the source program 1 shown in FIG. 2 as an example,
The operation of the embodiment shown in FIG. 1 will be described.

The program analysis / generation mechanism 2 is a program unit 100,
Input the source program 1 consisting of 200,300,400,500.

Program unit analysis processing means of the program analysis mechanism 21
22 performs the following processing according to the type of sentence appearing in each program unit. That is, when the inter-task shared variable declaration statement 110 of the program unit 100 is detected, the shared variable name AA and its element names (name) AM1, AM2, AM3 are included as part of the program unit shared variable information 101 shown in FIG. And information related to the shared variable AA including the element length (length) 4, 8, 8 is stored in the shared variable information area 29. The reference information is unreferenced as an initial value. Inter-task shared variable declaration statement 111,11
The same process is performed for 2, and the task-to-task shared variables CC and DD are
The information related to the above is stored in the shared variable information area 29.

Next, the program unit analysis processing means 22 connects the element DM2 of the inter-task shared variable DD and the area of parallel C to EQUIVALEN.
When the CE statement 113 is detected, it is checked whether or not other elements of the inter-task shared variable are area-joined by this area joining. In this case, since the array C is a complex type, the element length is 40, and since it corresponds to both areas of the element DM2 having the element length 20 and the element DM3 having the element length 20, the element DM3 is also area-combined with the array C, and the element DM2, DM3 is one element. Therefore, the shared variable DD of the program unit shared variable information 101 (see FIG. 3) in the shared variable information area 29.
Information of the elements of length 4 with element name DM1 and element names DM2, D
Change to consist of 40 elements with length M3, C.

Next, when the program unit analysis processing means 22 detects the task activation statement 120, the program name information, the argument information, and the program configuration information indicating that the task is activated are provided in the configuration information area 28.
To be stored. The procedure call statement 121 is similarly processed.

Next, when the program unit analysis processing unit 22 detects the shared variable reference statement 130 that refers to the element AM3 of the inter-task shared variable AA, the element AM3 of the inter-task shared variable AA of the program unit shared variable information 101 (see FIG. 3). Change the reference information of to refer to. The shared variable reference sentences 131 and 132 are similarly processed.

These processes are processed by other program units 200,300,400,500
To generate the program unit shared variable information 201, 301, 401, 501 as shown in FIG. 3 in the shared variable information area 29.

Next, the task unit analysis processing means 23 of the program analysis mechanism 21 refers to the program configuration information in the configuration information area 28 and classifies the program units for each task. That is,
The program unit 100 with the program name MAIN calls the program unit 200 with the procedure subprogram name D, and the program unit 200 does not call another procedure subprogram, so the program unit 100 and the program unit 200 have the same task. It is determined that it belongs to and is classified as such. Similarly, the program unit 300 of the procedure subprogram name A whose task is started from the program unit 100
Is calling program unit 400,500 with procedure subprogram name B, C, and program unit 400,500 is not calling another procedure, so program unit 300,400,500
Determine that they belong to the same task and classify them as such. That is, a task composed of program units 100 and 200 (hereinafter referred to as task 10) and a task composed of program units 300, 400 and 500 (hereinafter task 2).
Call it 0. ) And two tasks.

When the processing for classifying the program units for each of these tasks ends, the task unit analysis processing means 23 refers to the program unit shared variable information 101, 201 (see FIG. 3) of the program units 100, 200 constituting the task 10 and is declared. Reconfiguring the elements of the inter-task shared variables AA, CC, DD within the task, and generating the task-unit shared variable information 11 on the shared variable information area 29. In this case, the shared variable A between tasks
Since A and CC are declared only in a single program unit 100, the information in the referenced program unit shared variable information 101 is used as it is, and task unit shared variable information 11 is generated as shown in FIG. To do. On the other hand, the task shared variable DD
Is declared in multiple program units 100 and 200, so each element is associated. That is, as a result of the correspondence of the elements, the overlap of the second element does not match, so the information is changed so that the maximum value of the overlap of the areas is one element. As a result, the task-unit shared variable information 11 is generated as an inter-task shared variable DD with an element having a length of 4 and having a reference in the task and an element having a length of 40 and having a reference.

The same process is performed for the other tasks 20 to generate task unit shared variable information 21 as shown in FIG.

Next, the inter-task analysis processing means of the program analysis mechanism 21
24 is task-unit shared variable information 1 in the shared variable information area 29
1,21 (see Fig. 4), refer to the inter-task shared variables AA, CC,
The DDs are extracted one by one, and the inter-task shared variable information 31 with reference information attached to each of the inter-task shared variable parts is generated on the shared variable information area 29. That is, the shared variable AA between tasks
When extracting, the task that refers to this inter-task shared variable AA is searched for each part, and in this case, the inter-task shared variable
A unit of length 12 from the beginning of AA is a program unit 300,400,50
Referenced only in task 20 consisting of 0 and of length 8
Task 1 whose part consists of 100,200 program units
Task 20 consisting of 0 and program units 300,400,500
Since both are referred to, as shown in FIG.
The inter-task shared variable information 31 is generated by associating the information to that effect with the shared variable name AA.

The same processing is performed for other inter-task shared variables CC and DD.

Next, the shared variable information editing / output processing means 26 of the program generation mechanism 25 refers to the program unit shared variable information 101, 201, 301, 401, 501, the task unit shared variable information 11, 21, and the task shared variable information 31 in the shared variable information area 29. , And outputs it as a shared variable information list 4 as shown in FIG. This is used as a reference for program developers.

Finally, the program conversion processing means 27 of the program generation mechanism 25 inputs the source program 1 again and refers to the inter-task shared variable information 31 (see FIG. 5) in the shared variable information area 29 to refer to the source program 1 in the source program 1. Of the declarations of shared variables between tasks in each program unit, those that can be regarded as shared variables within the task of the invoking task should be replaced with the declaration of shared variables within the task, or a directive line indicating that they are shared variables within the task is specified in the source program 1. What is inserted and can be regarded as an in-task shared variable of another task is deleted from the source program 1 to generate a new source program 3.

That is, when the inter-task shared variable declaration statement 110 in the program unit 100 (see FIG. 2) is detected, the inter-task shared variable information 31 is referenced to determine whether the inter-task shared variable AA is an intra-task shared variable. To judge. In this case, the three-element AM of the inter-task shared variable AA in the inter-task shared variable declaration statement 110
Length 12 from the beginning corresponding to elements AM1 and AM2 among 1, AM2 and AM3
Since it is found that the part of is a substantial in-task shared variable of the task 20 other than the task 10 to which the program unit 100 belongs, the elements AM1 and AM2 are deleted from the inter-task shared variable AA, and FIG. As shown in, the inter-task shared variable declaration statement 110a of the inter-task shared variable AA composed of one element AM3
Is generated in program unit 100.

Next, when an inter-task shared variable declaration statement 111 is detected in the program unit 100 (see FIG. 2), all elements of the inter-task shared variable CC are assigned to the task 10 to which the program unit 100 belongs.
Since it is a shared variable in the actual task of task 20 which is different from
This declaration statement 111 is deleted.

Next, when an inter-task shared variable declaration statement 112 is detected in the program unit 100 (see FIG. 2), all elements of the inter-task shared variable DD are assigned to the task 10 to which the program unit 100 belongs.
Since it is an in-task shared variable, the declaration statement 112 is changed to an in-task shared variable declaration statement 112a as shown in FIG.

Perform the same process for other program units,
Finally, the source program 3 as shown in FIG. 7 is generated.

〔The invention's effect〕

As described above, in the in-task shared variable detection processing method of the present invention, processing such as detecting an inter-task shared variable that can be regarded as an in-task shared variable from the source program and replacing it with the in-task shared variable is automatically performed. Therefore, it is possible to relieve the human from complicated work and to significantly shorten the processing time. Moreover, since no omission of detection occurs, the efficiency at the time of executing the program can be sufficiently improved.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of the in-task shared variable detection processing method of the present invention, FIG. 2 is a diagram showing an example of a source program, FIG. 3 is a diagram showing an example of program unit shared variable information, FIG. 4 is a diagram showing an example of task-unit shared variable information, FIG. 5 is a diagram showing an example of inter-task shared variable information, FIG. 6 is a diagram showing an example of a shared variable information list, and FIG. 7 is a conversion. It is a figure which shows the example of a later source program. In the figure, 1 ... Primitive program 2 ... Program analysis / generation mechanism 21 ... Program analysis mechanism 22 ... Program unit analysis processing means 23 ... Task unit analysis processing means 24 ... Intertask analysis processing means 25 ... Program Generation configuration 26 …… Shared variable information editing / output processing means 27 …… Program conversion processing means 28 …… Configuration information area 29 …… Shared variable information area 3 …… Primitive program 4 …… Shared variable information list

Claims (1)

(57) [Claims] 1. A processing method for detecting an inter-task shared variable, which can be regarded as an intra-task shared variable, from a source program written in a high-level programming language capable of describing execution by a multiprocessor, and replacing the inter-task shared variable with the in-task shared variable. By inputting to generate program unit shared variable information indicating whether or not the inter-task shared variable declared in the program unit is referenced in the executable statement in the program unit for each program unit,
Program unit analysis processing means for generating the program unit name that is started as a task from the program unit and the program unit name that is called as a procedure and the program structure information regarding the starting method, and each program with reference to the program structure information. The unit is divided for each task unit to which it belongs, and with reference to the program unit shared variable information, the task-to-task shared variable declared in each task unit is shared in each task unit. A task unit analysis processing unit that generates task unit shared variable information indicating whether or not it is referenced, and in which task unit each intertask shared variable is referenced with reference to the task unit shared variable information. Between-task analysis processing means for generating inter-task shared variable information indicating With reference to the variable information, for each program unit in the source program, among the declarations of the inter-task shared variables included in it, those that can be regarded as the intra-task shared variable of the task to which the program unit belongs are Program conversion that creates a new source program by replacing it with a declaration or inserting a directive line indicating that it is a shared variable in a task in the source program and deleting what can be regarded as a shared variable in a task of another task from the source program. A shared variable detection processing method in a task, comprising: a processing means.