JPH0776926B2 - Loop control processing method - Google Patents

Description

DETAILED DESCRIPTION [Summary] When an original program of a computer system is translated to generate an object program that avoids loop execution when the initial control condition of each program loop is a predetermined condition Processing method. A loop having a condition that allows execution avoidance determination processing to be performed collectively for a plurality of loops is detected, and the individual determination processing is deleted. With this method, it is possible to reduce the instruction steps of the target program and improve the execution efficiency.

[Industrial application field]

The present invention translates an original program of a computer system,
The present invention relates to a loop control processing method when an object program is generated.

An original program written in the FORTRAN programming language or the like is translated by a known so-called compiler or the like to generate a target program to be executed by a required computer system.

Generally, various checks that are not explicitly specified in the original program are added to the target program, but it is desirable to minimize them and improve the execution efficiency of the program.

[Problems to be solved by conventional technology and invention]

As is well known, a programming language such as FORTRAN has a description function of a program loop (hereinafter referred to as a DO loop) that starts with a DO statement, and is often used in programs.

The DO loop uses the initial value, the closing value, and the increment value specified as its control parameters to change the value of the specified control variable from the initial value to the increment value at a value that does not exceed the closing value, 1 program in the DO loop for each value
It is to repeat execution once.

Here, as a language specification of a programming language such as FORTRAN, when the control parameter of the DO loop is a condition for ending the execution of the loop, that is, when the initial value of the control parameter exceeds the final value (increment value> 0 In the case of, the initial value> the final price, and in the case of the increment value <0, the initial value <the final value). Therefore, the DO loop is to be avoided without being executed even once.

The original program creator, by the control according to this specification,
In some cases, the program may be configured to control whether or not the DO loop needs to be executed.
It is necessary to configure the target program so as to avoid the DO loop and never execute it.

Therefore, in the translation of the original program, if a variable is specified in the control parameter of the DO loop and the value is not determined at the time of translation, the success or failure of the avoidance condition cannot be determined at the beginning of the DO loop at execution. A determination process is added so as to determine the avoidance condition.

Each determination process consists of two steps, for example, a determination instruction for determining the magnitude of the initial value and the final value, and a branch instruction according to the result. Such an instruction step is pre-positioned in each of the DO loops of the above condition, and is always As a result, the result is an increase in the number of execution steps of the target program.

[Means for solving problems]

FIG. 1A is a block diagram showing the configuration of the present invention,
FIG. 6B is a flow chart of the processing of the present invention.

In FIG. 1A, 20 is a processing device, 21 is a translation processing unit, and 22 is a loop avoidance determination optimization processing unit. FIG. 1B shows a flow of processing of the loop avoidance determination optimization processing unit 22, and
Is a processing step.

[Work]

The translation processing unit 21 executed in the processing device 20 of FIG. 1A translates the original program 23 held in the storage device to generate the target program 24. Next, the loop avoidance determination optimization processing unit 22 of the present invention executes the processing described below with respect to the object program 24, and the optimization object program 25
Is output.

Since the DO loop is generally a multiple loop having multiple levels of 1 or more, in the avoidance determination optimization processing,
In the processing step 1, the head DO loop is searched, and the head DO loop at the innermost level of the deepest multiple loop of the multiple levels is searched and retrieved.

In processing step 2, the same multi-level DO loop as the retrieved DO loop (hereinafter referred to as sibling loop)
If there is a sibling loop, process step 3
Then, with respect to the sibling loops that satisfy the predetermined condition, the processing is changed so that the avoidance determination processings are integrated.

In the next processing step 4, it is determined whether or not there is a multiple level outside by one level, and if there is, a determination is made in processing step 5 whether the DO loop of the level being processed satisfies a predetermined condition in relation to the outer DO loop. .

If the condition is met, in step 6 the inner DO
The loop avoidance determination process is deleted.

In processing step 7, a DO loop to be processed next is searched, and if it is a loop to be processed, the processing loops through processing step 8 and returns to processing step 2.

In processing step 7, it is assumed that the unprocessed first DO loop of the same level as the currently processed DO loop or the closest level to the current level is taken out, and if the outermost DO loop of the multiple loop is processed in this way, the processing ends. To do.

By the above processing, the number of places where the DO loop avoidance determination processing is inserted is reduced, and the program execution efficiency is improved.

〔Example〕

The loop avoidance determination optimization processing unit 22 of FIG. 1 (a) processes the DO loop of the target program 24 as follows, and outputs the optimized target program 25 with reduced avoidance determination processing.

The DO loop is generally considered to be a multiple loop of one or more multiple levels, and the avoidance determination optimization process is performed sequentially from the innermost level loop of the multiple loop having the highest multiplicity to the outside.

In the processing step 1 of FIG. 1B, the first DO loop is searched, and the first loop at the innermost level corresponding to the above is searched and taken out.

In processing step 2, it is searched whether there is a sibling loop of the extracted DO loop, and if there is a sibling loop, in processing step 3, the processing is changed so that the avoidance judgment processing of those sibling loops satisfying a predetermined condition is integrated. To do.

If there is no sibling loop, the process immediately proceeds to processing step 4, and optimization processing with an outer loop described later is started.

FIG. 2 is a diagram showing a detailed flow of processing of the sibling loop of processing step 3.

When a sibling loop is detected in processing step 2 of FIG. 1 (b), the current processing loop will be referred to as a leading loop below,
Processing step 10, the next sibling loop is taken out and it is determined whether or not it is continuous. If it is continuous, processing step 11 determines whether or not a predetermined condition is satisfied in relation to the head loop.

For this determination, for example, the control parameters of the first loop (the above-mentioned initial value, final value, and increment value) and the control parameter of the DO loop being processed are the same, and in both loops, the variable of the control parameter is set. The condition is that it has not been changed.

If this condition is met, then in step 12 this DO
After deleting the loop avoidance determination process, the process returns to the process step 10.

In this way, the loop of processing steps 10 to 12 continues, and when it is determined in processing step 10 that there is no continuous sibling loop, the processing proceeds to processing step 13 and the branch destination when avoiding the avoidance determination processing in the first loop. Changes to branch immediately after the last sibling loop.

In processing step 14, it is judged whether or not all the sibling loops have been processed. If not processed, the next sibling loop is set as a new head loop and the process returns to processing step 10. If there is no unprocessed processing, this processing is terminated and the process shown in FIG. Go to process step 4) b).

If the above conditions are not satisfied in the determination in processing step 11, the process branches to processing step 15 and processing steps 10 to 12 are executed.
Loop is broken. In processing step 15, the branch destination of the avoidance determination processing of the top loop is changed so as to branch immediately before the last fetched sibling loop, the sibling loop becomes a new top loop, and the process returns to processing step 10.

With the above processing, the avoidance determination processing is reduced for the sibling loop so that the avoidance determination processing is performed only once for each continuous loop that satisfies the above condition.

In processing step 4 of FIG. 1 (b), it is judged whether there is a multiple level outside the processed loop of the sibling loop by one level, and if there is, a DO loop of the level being processed in relation to the outside DO loop. In step 5, it is determined whether or not satisfies the predetermined condition.

This determination is, for example, on the condition that the control parameters of the outer loop and the inner loop are the same, and that there is no process for changing the variables that are the control parameters between the start of the outer loop and the start of the inner loop. .

If there are loops that satisfy the above conditions in the inner loops, the avoidance determination processing for those loops is deleted in processing step 6.

In processing step 7, a DO loop to be processed next is searched, and if it is a loop to be processed, the processing loops through processing step 8 and returns to processing step 2.

In processing step 7, the unprocessed first DO loop, which is at the same level as or outside the current processing DO loop and is closest to the current level, is taken out, and thus the outermost DO loop of the multiple loop is extracted. Processing ends.

〔The invention's effect〕

As is clear from the above description, according to the present invention, in the process of translating the original program of the computer system to generate the target program, the number of places where the DO loop avoidance determination process is inserted is reduced, so that it is generated. There is a remarkable industrial effect that the execution efficiency of the target program is improved.

[Brief description of drawings]

 FIG. 1 (a) is a block diagram showing the configuration of an embodiment of the present invention, FIG. 1 (b) is a flowchart of the processing of the present invention, and FIG. 2 is a flowchart of sibling loop processing. In the figure, 1 to 8 and 10 to 15 are processing steps, 20 is a processing device, 21 is a translation processing unit, 22 is a loop avoidance determination optimization processing unit, 23 is an original program, 24 is a target program, and 25 is an optimization target program. Indicates.

Claims (1)

[Claims] 1. When translating an original program of a computer system, whether to execute the loop at all or at least once based on control information specified by an initial value, a final value and an increment of a variable for controlling the loop. Is a loop control processing method in a compiler that generates an object program of a loop to which the avoidance processing for avoiding the execution of the loop is added, and for a plurality of continuous loops at the same level, When the control information is the same, execution of all loops is avoided by one avoidance process, and when the control information of the outer loop and the inner loop is the same for the multiplexed loop, the inner loop The loop control processing method is characterized by performing only the avoidance processing of the outer loop without performing the avoidance processing of.