JP2500515B2 - CASE structure processing method that allows range specification - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is a CA whose range can be specified.
CA in a computer system having a compiler of a source program written in a language that supports the specification of SE structure
This is related to the processing method of the SE structure, and especially when the CASE statement appears, the value specified in the CASE statement is changed to SELECT C
Each ASE block is stored and compared with a previously stored value to determine the optimal branch destination when that block is not executed for a range of values,
The present invention relates to a CASE structure processing method capable of specifying a range.

[0002]

2. Description of the Related Art In a computer system equipped with a compiler that compiles a source program written in a language that supports the specification of a CASE structure capable of specifying a range,
By arranging the values that determine the branch destination for the process specified by the conventional CASE statement in ascending or descending order, unnecessary judgment and branch instructions are reduced, and branching by the branch table is performed between the range specification of the CASE statement. There was a technique of generating (for example, Japanese Patent Laid-Open No. 3-4332 (A) FOR
CASE structure processing method in TRAN).

[0003]

In the above-mentioned prior art,
The processing specified in the CASE statement in order to reduce extra judgment and branch instructions when processing a range-capable CASE structure, and to generate a branch by a branch table between the range specification of the CASE statement. The values that determine the branch destinations for are arranged in ascending or descending order.

However, in this method, even if the user consciously describes the CASE block that is most likely to be executed in the first part, the intention of the user is not reflected and the execution speed is rather slowed down. There was a flaw.

[0005]

The processing method of the CASE structure capable of specifying a range according to the present invention is a CAS specifying a range.
C in a computer system having a compiler for a source program written in a language that supports the use of E structures
In the processing method of the ASE structure, each time a CASE statement appears, the value specified by the CASE statement is changed to SELECT C
An ASE block is stored for each ASE block and the magnitude relationship is compared with the already stored value to determine the optimum branch destination when the block is not executed. Is determined for the result of evaluation of the expression, if it is true, an intermediate code for branching to the corresponding CASE block is generated, and if false, an intermediate code is generated for branching to the optimum branch destination obtained by the next-stage determination branch processing determination unit.
The text that branches to the processing block corresponding to the E sentence is sequentially
And a CASE structure processing section having a decision branch processing generation section for outputting .

[0006]

According to the present invention, unnecessary judgment and branch instructions are reduced without changing the order of blocks described by the user, and branches are generated by a branch table between the range specification of the CASE statement. Will be able to.

[0007]

EXAMPLES [First Example] The following is the Fortr of the present invention.
An embodiment of an compiler will be described in detail with reference to the drawings.

FIG. 2 is a diagram showing the structure of the entire compiler to which the present invention is applied. The syntax analysis unit 22 of FIG.
The source program 21, which is an rtran program, is input, the syntax is analyzed, and the intermediate code 13 is converted. The semantic analysis unit 23 analyzes the meaning of each sentence using the CASE structure processing unit 1 or the like, and converts it into the intermediate code 14 in a format close to the object code. This is optimized by the optimization processing unit 24, and the code generation unit 25 executes the object code 2
7 is generated.

The present invention relates to the CASE structure processing unit 1,
This is to improve the execution efficiency of the object code 27. The operation of the portion related to the present invention in the CASE structure processing unit 1 will be described.

FIG. 1 is a block diagram of a CASE structure processing method capable of specifying a range, which is an embodiment of the present invention. FIG.
In the above, the CASE structure processing unit 1 is a means for processing a sentence related to the CASE structure, and the next decision branch processing decision unit 11 stores the CASE value specified in the CASE sentence for each SELECT CASE block every time the CASE sentence appears. Then, it is a means for determining the optimum branch destination when the block is not executed for the range of CASE value by comparing with the value already stored,
The decision branch process generation unit 12 is the next decision branch process determination unit 11
The branch table generation judging unit 121 is a means for judging the result of evaluating the expression related to the CASE structure on the basis of the information obtained in 1. and generating a branching intermediate code. Using the branch by the branch table based on the obtained information is a means for determining the group of determinations that the object program can execute earlier, and the table branch intermediate code generation unit 122 determines the group by the branch table generation determination unit 121. In response to the above determination, it is a means for outputting the intermediate code of the branch by the table branch, and the intermediate codes 13 and 14 correspond to the intermediate code which is being converted from the source program 21 to the object code 27.

In this embodiment, CASE capable of specifying a range
New Fort as a language that supports the specification of structure
The ran language is taken as an example.

In the new Fortran language, the specification of the CASE structure capable of specifying the range consisting of the following syntax rules is added. SELECT CASE structure select-case statement [case statement block] ... end-select statement select-case statement [case structure name:] SELECT CASE (case expression) case statement CASE case selector [case structure name] block Collection of execution statements end -select statement END SELECT [case structure name] case selector DEFAULT or (list of case value range) case value range case value or case value: or: case value or case value: case value The FORSE rules for CASE structure are , CAS
E value range can be specified, multiple CASE values and CASE value range can be specified for one block,
Even if it is not possible to specify a CASE value or a range of CASE values such that there are two or more blocks to be executed, or if you do not explicitly specify to exit from the block like the break statement in c language, The feature is that only blocks are executed.

The following program fragment is an example of the CASE structure.

SELECT CASE (I + J)! CASE (10:20)! PRINT *, '10-20'! CASE DEFAULT! PRINT *, 'DEFAULT'! CASE (30)! PRINT *, '30'! END SELECT! This program fragment prints "10-20" if the value of (I + J) is 10 or more and 20 or less, and END SELECT
If the value of (I + J) is equal to 30, control is transferred to the next execution statement of the statement, "30" is printed, and END SELECT
The control is transferred to the next execution statement of the statement, and if none of them, "DEFULT" is printed and ENDSELECT
Operates to transfer control to the next executable statement in the statement.

Next, of the embodiments shown in FIG. 1 of the present invention, CA
The operation of the SE structure processing unit 1 will be described with reference to FIGS. 3, 4, and 5.

FIG. 3 shows a Fortran original program for explaining an embodiment of the present invention.

FIG. 4 shows an example of intermediate code generated from the program of FIG. 3 according to the prior art.

A source program as shown in FIG. 3 is converted into an intermediate code as shown in FIG. 4 based on the standard syntax / semantic analysis theory. That is, the information of the CASE statement is recorded, and when the END SELECT statement appears, an intermediate code for performing a decision branch corresponding to each CASE value is output.

The decision branch according to the range of CASE value of the sentence 3-1 (hereinafter simply referred to as 3-1) in FIG. 3 is 4-1 in FIG.
And 4-2 (hereinafter referred to as 4-1 and 4-2) are decomposed into two intermediate codes. If 4-1 is smaller than the lower limit of the range of CASE value, the block is not executed, so that the process branches to the intermediate code of the next judgment. Hereinafter, such a determination branch will be referred to as a lower limit determination branch. If 4-2 is smaller than the upper limit of the range of CASE values, it means that the condition that control is passed to 4-2 is greater than or equal to the lower limit, so branch to the branch destination for executing this block. To do. Hereinafter, such a determination branch will be referred to as an upper limit determination branch.

When 4-1 is true, that is, when the CASE expression is smaller than 20, the branching destination 4-3 is never true because the judgment condition is equal to 60. The decision is useless. In this case, 4
The judgments of -4 and 4-5 are also useless, and there is a possibility of becoming true next, that is, 4-6.

Similarly, in the case of 4-5, when 4-5 is false, that is, when the CASE expression is larger than 50, the branch destination 4-6 is never true.
It is useless to make this determination.

If the judgment branches of the CASE value are sorted, this useless judgment can be reduced, but if the judgment is not performed in the order intended by the user, the judgment that is likely to be executed will be performed later. Not good to do in.

The next-judgment branching processing decision unit 11 skips the unnecessary branching in the information of the CASE statement recorded so as to avoid the unnecessary judgment without changing the order specified by the user, and the magnitude relationship of the CASE values. The process of adding the information of is performed.

The process of the next decision branch process decision unit 11 will be described below with reference to FIGS. 5, 6 and 7.

FIG. 5 shows the case of the standard parsing theory, the conventional technique and the CASE according to the present invention, using the source program of FIG. 3 as an example.
It is a figure explaining the difference of the information of a sentence.

In FIG. 5, a label is a label created for this determination. The content is a record of what kind of determination is made. The branch destination means which block is executed when this judgment is true. The output order is to record the order when the judgment is made in the order different from the order described in the source program. In this example,
They are sorted in ascending order.

According to the recording of the CASE information according to the present invention, the left branch and the right branch are added to the range designation of the recorded CASE sentence information. The left branch means a judgment that appears after this judgment and has a CASE value smaller than the lower limit of the range specification. That is, it is the determination to be performed next when the lower limit determination determines that the block corresponding to the determination is not executed. The right branch means a judgment that appears after this judgment and has a CASE value larger than the upper limit of the range specification. That is, it is the next determination made when the upper limit determination determines that the block corresponding to the determination is not executed.

Of the recorded CASE sentence information, only the right branch is added to the judgment other than the range designation. This is the next determination made when it is determined by this determination that the block corresponding to the determination is not executed.

The newly recorded CASE method is always a branch of CASE information somewhere. In this way, the order of judgments written in the source program is not changed.

Here, x means that there is no right or left branch, and-is not used here (not referenced).
Means that.

CASE based on standard parsing theory
According to the recorded information, for example, if a value of 15 (no block is executed) is determined, all determinations must be made.

Generally, the efficiency of execution is improved by bringing the judgment with the highest probability of execution first, but according to the conventional technique, for example, in the example of FIG.
2 most of the time when ASE blocks are executed
Even if the block of 0:30 is executed, the judgment of 10 is always made.

FIG. 6 shows a CASE according to the present invention as shown in FIG.
It is a block diagram explaining the process of the next determination branch process determination part 11 for recording information. CASE structural analysis unit 1
Every time the CASE statement appears, it calls the next decision branching process determining section 11 and records the information of the CASE statement.

Referring to FIG. 6, reference numeral 61 designates the information of the CASE sentence input as the initial setting as the current determination, and the SELE being processed.
The step of determining the information of the first CASE statement of the CT CASE information as a target, 62 is a step of checking whether the target determination is a range designation and moving to step 63 if the target determination is determined, and step 65 to step 65 if not determined. 63 is a step of checking whether or not the value of the current judgment is smaller than the value of the target judgment, and if it is smaller, the step is moved to step 64, and if it is larger, the step is moved to step 65. 64 is checked whether the left branch already exists in the target judgment, If it exists, the step moves to step 641; otherwise, the step moves to step 642; 641 sets the left branch of the target judgment as the target judgment, and moves the execution to step 62; 642 shows the current judgment the left branch of the target judgment. Then, the step 65 ends the processing, and the step 65 checks whether or not the right branch already exists in the target determination. If it exists, the step goes to the step 651, and if not, the step 652. Step 651 is the step of shifting the execution to step 651, the right branch of the target determination is the target determination, step 562 is the step of shifting the execution to the step 62, and the current determination is the right branch of the target determination to end the processing.

FIG. 7 shows the state of CASE information obtained by applying the present invention corresponding to the CASE information of FIG.
It is a figure when CASE information of the determination of ASE value 15 is added.

The process of the next decision branch process decision unit 11 will be concretely described with reference to FIGS. 6 and 7.

According to FIG. 6, the recording of the CASE information is shown in FIG.
When the CASE statement of CASE (15) is further read in the state of, the CASE information L15 is recorded by the following operation.

First, the current judgment L15 is the value 15, the target judgment L15.
20 becomes the value 20:30 (step 61). Since the target determination is the range designation (step 62), when the values of the current determination and the target determination are compared, the current determination is small, so the left branch of the target determination is checked (step 63). Since the left branch exists (step 64), the target determination is set to L10 of the left branch,
The process is repeated (step 641). Next, since the target determination is not the range designation (step 62), the right branch is examined. Since the right branch does not exist (step 65), the current judgment is set as the target judgment branch and the processing ends (step 65).
2).

As a result of this processing, CASE information as shown in FIG. 7 is obtained.

Next, the processing of the decision branch intermediate code generator 12 will be described with reference to FIGS. 7 and 8.

FIG. 8 is an intermediate code obtained when the present invention is applied to the source program of FIG.

The decision branching process generator 12 generates an intermediate code for making a decision in the order in which the CASE statement appears. At this time, the CAS recorded in the next determination branching process determination unit 11 is
A branch to the next determination is determined based on the magnitude relationship of the E value, and an intermediate code is generated.

That is, the CASE information is taken out in the recorded order, the intermediate code of the decision branch and the label for the decision of the branch of the CASE information are generated, and the intermediate code branching to the branch is generated when the block is not executed. To do. CA
If a branch is not recorded in the SE information, and if the block is not executed, DEFALUT is present if DEFULT exists.
Generates intermediate code for an unconditional branch that exits a block, or a SELECT CASE block if none.

In the example of FIG. 7, since the information of the label L20 is the range designation, when it is smaller than 20, the label L of the left branch is displayed.
Generates an intermediate code that branches to 10, then branches to BLK 20 when the number is 30 or less, and L in the right branch otherwise.
An intermediate code that branches to 60 is generated. Then label L6
Since the information of 0 is not the range designation, if it is equal to 60, BL
An intermediate code that branches to K60, and branches to L60 on the right branch otherwise is generated. By repeating the same process, the intermediate code of FIG. 8 is obtained. Comparing FIG. 4 and FIG.
The branch destinations of 4-1 and 4-4 are changed to optimal branch destinations that do not make a redundant determination.

If the user sorts and describes the order of this determination on the source program, the same effect as the conventional technique can be obtained. Therefore, when the user does not know which block has a high probability of being executed when programming, the CASE values are rearranged in the source program, and when the block with a high probability of being executed is known, the block with the highest probability is selected. It should be described in order.

[Second Embodiment] Finally, with respect to the second embodiment, the operations of the branch table generation determination unit 121 and the table branch intermediate code generation unit 122 will be described based on the examples of FIGS. 9 and 10.

FIG. 9 is a Fortr for explaining the second embodiment.
It is an original program.

FIG. 10 shows C obtained from the Fortran original program of FIG. 9 by the next decision branch process decision unit 11.
It is a figure of ASE information. However, the symbol “*” in FIG. 10 is a mark that indicates that the table branch is performed, which is recorded by the branch table generation determination unit 121, and the symbol “!”. Similarly, "is a mark indicating that this CASE information has been processed.

The branch table generation determining unit 121 determines, from the tree structure representing the magnitude relationship of the CASE values created by the next determination branching processing determining unit 11, the left branch of each branch, or the right branch or the right branch. Examine the branch that is the leftmost branch of. That is, a group of judgments having a value between a certain range specification and a certain range specification is detected, and it is judged whether or not the branch by the branch table is faster than the group of judgments under the branch. When it is determined that the branch by the branch table is fast, the mark that the table is branched is recorded in the branch.

The criteria for the judgment are the amount of memory required for the branch table and the amount of decrease of the judgment instruction by the branch table.

For example, referring to FIG. 10, it can be seen that the portion indicated by the numeral 101 is the portion surrounded by the range designation and the range designation. For this part, the branch by the branch table has a table size of 10 addresses, and two judgment instructions are required for the range specification, so there are five, which is a computer having a branch table branch instruction. In most cases it will be faster in the system. Then CAS
Record the table branch mark on the branch with E value 31.

The table branch intermediate code generator 122
When outputting the judgment intermediate code, if there is a mark for branching the table, the branch is traced from there to output the intermediate code of the branch by the table branch, and the processed mark is recorded for all of the target group of judgments. .

For example, referring to FIG. 10, it can be seen that the branch with the CASE value 31 is first marked. Therefore, the intermediate code of the branch by the table for the branches below 31 is output. When searching, information of 20:30 and 40:50 is recorded, and a branch table from 31 to 39 is created. Also follow the branches 31, 36, 33: 3
While setting the branch destinations of 5 and 32 in the branch table, the processed mark is recorded in each branch. When outputting the intermediate code for judgment, it is necessary to confirm that this output mark is not recorded.

FIG. 11 is an example of intermediate code obtained by converting the source program of FIG. 9 according to the present invention.

Here, 11-1 is an intermediate code of the table branch which means to branch to the Rth label among the labels after (R).

Looking at FIG. 11, without changing the order of determination,
It becomes possible to output the intermediate code of the branch by the branch table, and it can be seen that there is one intermediate code of the branch by the branch table instead of the judgment of 31, 32, 33: 35, 36, and the speedup is achieved. .

[0057]

As described above, according to the present invention,
When the CASE syntax capable of specifying a range is realized, efficient instruction output can be performed without changing the output order of the instructions to be determined, and high-speed branching can increase the object execution speed.

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an embodiment of the present invention.

FIG. 2 is a diagram showing a structure of an entire compiler to which the present invention is applied.

FIG. 3 is a Fortr for explaining an embodiment of the present invention.
An primitive program.

4 is an example of intermediate code generated from the program of FIG. 3 according to the prior art.

FIG. 5 is a diagram for explaining the difference between the standard syntax analysis theory and the information of the CASE statement according to the present invention and the present invention, using the source program of FIG. 3 as an example.

FIG. 6 is a diagram illustrating a process of a next decision branch process decision unit 11;

FIG. 7 is a diagram of CASE information obtained when the present invention corresponding to the CASE information of FIG. 5 is applied.

8 is an intermediate code obtained when the present invention is applied to the source program of FIG.

FIG. 9 is a Fortran original program for explaining a second embodiment.

FIG. 10: Fo of FIG. 9 by the next determination process determination unit 11
Figure of CASE information obtained from the rtran source program.

11 is an intermediate code obtained by converting the source program of FIG. 9 according to the present invention.

[Explanation of symbols]

 In FIG. 1, 1 CASE information processing unit 11 next decision branch process determination unit 12 decision branch process generation unit 121 branch table generation determination unit 122 table branch intermediate code generation unit 13 intermediate code 14 intermediate code In FIG. 2, 1 CASE structure processing unit 13 intermediate Code 14 Intermediate code 21 Source program 22 Syntax analysis unit 23 Semantic analysis unit 24 Optimization processing unit 25 Code generation unit 26 Intermediate code 27 Object code

Claims (2)

(57) [Claims] 1. In a CASE structure processing method in a computer system having a compiler for a source program written in a language that supports the use of a range-specifiable CASE structure, a CASE statement is specified each time the CASE statement appears. The determined value for each SELECT CASE block, and compares the value stored with the stored value to determine the optimum branch destination when the block is not executed. Then, if the result of evaluating the expression related to the CASE structure is determined, if it is true, the intermediate code that branches to the corresponding CASE block, and if false, branches to the optimum branch destination obtained by the next decision branch processing decision unit is selected. Generate and process corresponding to each CASE sentence
A processing method of a CASE structure capable of specifying a range, which comprises a CASE structure processing unit having a judgment branching processing generation unit that sequentially outputs a text branched to a block . 2. The decision branch process generation unit is configured to execute the CAS.
Which of the processing using the branch table and the processing not using the branch table for all the judgments that occur after the range specification having a value between the range specification specified by the E statement and the range specification that appears later is performed. A branch table generation determination unit that determines whether high-speed processing is possible, and a range specification specified by a CASE statement that determines that the branch table generation determination unit can use the branch table for higher-speed processing. And a table branch intermediate code generation unit that generates an intermediate code for branching by a branch table for all judgments that occur after the range specification having a value between the range specification that appears next. The CASE structure processing method capable of specifying a range according to claim 1.