JP3921722B2 - Compiler processing device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a technique for converting an initial value of a DATA statement (DATA declaration instruction) of a compiler such as FORTRAN into an object program.
[0002]
[Prior art]
Conventionally, there are the following two methods for assigning initial values of DATA statements of compilers such as FORTRAN.
[0003]
The first method (hereinafter referred to as a batch saving method) will be described based on a program example 1 as shown in FIG. FIG. 1A shows a source program of an initial value setting portion by a DATA statement that is a part of the FORTRAN program as an example, and is assumed to be program example 1, but for the sake of convenience of explanation, initial values in the program examples will be hereinafter described. Are all 4 bytes with a sign.
[0004]
“INTEGER” in (1) is a declaration command for securing an integer work area (area), and “DATA” in (2) to (4) is a declaration command for setting an initial value in the work area. , “END” in {circle around (5)} indicates an end instruction of the source program.
[0005]
FIG. 2B shows the initial value setting and output process of the first program example. First, the work area shown in (1) is secured by the INTEGER instruction (1). Subsequently, initial values are set in (4) through (2) and (3) by the DATA commands (2) to (4), and all initial values are set.
[0006]
However, in this method, in the case of the program example 2 of (1) shown in FIG. 12, that is, 10,000,000 work areas are secured by the INTEGER instruction (1), and (2) and (3) A large array (variable) that sets the initial value "1" in the first element B (1) work area and the initial value "2" in the last element B (10000000) work area by the DATA instruction In this case, as shown in the initial value output result of (2), there is a problem that the area between them is wasted and it takes time to acquire the work area. Furthermore, there may be a case where the compilation fails because the work area cannot be secured. Note that, in the initial value setting by the collective storage method, although not omitted or expressed in the figure, there is a management block (described later) for managing the initial value.
[0007]
Next, as a second method (hereinafter referred to as a divided storage method), when an initial value is first given, a management block (4 bytes × 5 items = size of 20 bytes) for setting and managing the initial value is set. create. FIG. 13A shows the relationship between the contents of the management block and the initial value setting work area. The contents of the management block include “pointer to next management block” and “pointer to initial value”. , "Initial value allocation position", "Work area size", and "Initial value repeat count" are stored and retained, and the "Initial value pointer" is used to set the initial value setting work area (variable length). ). Each time an initial value is given, that is, as many management blocks as the initial value are created, these are chained by the above-mentioned “pointer to the management block” in the order of allocation positions to manage the initial values. However, when the initial values of the same value continue, the initial value can be managed by one management block only by setting the required number of “number of repetitions of the initial value” of the management block. (Program example 4)
Then, while referring to these management blocks, the object program is output in the order of the allocation position and completed.
[0008]
FIG. 13 (2) is a diagram when initial values are output by the divided storage method for the previous program example 1, and three management blocks of the number of initial values are created. The order in which the sentences appear in the sentence is sorted in the order of position), and is chained by a “management block pointer”. The initial values “1”, “2”, and “3” are chained to the initial value setting work area by “pointer to initial value”, and are set and stored.
[0009]
However, in this method, as the number of initial values increases, the number of management blocks for storing and managing the initial values increases, the work area of the management block increases, and the initial values are not given in the order of allocation positions. In such a case, the time for sorting the management blocks in the order of the allocation position also increases, which causes a problem that the compilation time becomes longer.
[0010]
As described above, the initial value setting method of the DATA statement includes the batch saving method and the divided saving method. In the batch saving method, the initial value is managed and stored in a continuous work area, and the management block is one. However, in the split storage method, each time an initial value is given, the initial value setting work area and its management block are secured and set, so the initial value is not always set to a continuous work area. The number of setting work areas and management blocks is basically plural. However, as described above, when the same initial value continues, the initial value can be managed in one management block by setting the “number of initial value repetitions” in the management block.
[0011]
[Problems to be solved by the invention]
When setting an initial value by a DATA statement, conventionally, the system (compiler) fixes whether to use the batch saving method or the divided saving method. Therefore, depending on the initial value setting structure of the DATA statement, a wasteful work area occurs, it takes time to acquire the work area, and if the number of initial values increases, the work area for the management block becomes large. In addition, in the case where the initial values are not given in the order of the allocation positions, the time for sorting the management blocks in the order of the allocation positions also increases, resulting in a problem that the compilation time becomes longer.
[0012]
In view of these points, the present invention aims to provide a means for holding a plurality of initial value setting means in the system and selecting the initial value setting means depending on the structure of the initial value setting by the DATA statement. To do.
[0013]
[Means for Solving the Problems]
The above problem is solved by a compiler processing device configured as follows.
[0014]
FIG. 1 is a block diagram of the present invention. That is,
In a computer program compiler processor,
Recognizing means for recognizing a sentence expressing initial value setting, analyzing means for analyzing the structure of the sentence, storing method selecting means for selecting a storing method of the initial value according to the result of analysis, and initial value by the storing method By providing an initial value setting means for executing and outputting settings, it is possible to select an initial value storage method according to the structure of the sentence, which leads to a reduction in work area and at the same time a faster initial value setting can be realized. .
[0015]
DETAILED DESCRIPTION OF THE INVENTION
In the following, the method for setting the initial value of the DATA statement will be described using some examples of the FORTRAN program. The entire flow of the initial value setting will be schematically described based on FIG. .
[0016]
First, a statement expressing the initial value setting in the (source) program by the recognition means 10 in the figure, that is, a declaration instruction such as “INTEGER” for securing an area in the main memory for storing the initial value, followed by It recognizes a “DATA” declaration command that sets an initial value given by a parameter in the reserved area.
[0017]
The parameter structure of the recognized declaration instruction (hereinafter simply referred to as an instruction) is analyzed by the analysis means 11, and as a result of the analysis, a plurality of initial value storage methods, that is, the batch storage method and the divided storage method are used. The initial value is set / stored or selected by the storage method selection means 12.
[0018]
Then, according to the selected storage method, the actual initial value setting is executed and output by the initial value setting means 13, and the setting of the target initial value is completed.
Subsequently, these processes will be described by giving specific program examples to explain which method is advantageous in which case and to show the structure of the present invention.
[0019]
FIG. 2 shows the general form of quoting a specific program example. (1) The name of the area to be secured and its size 20 are specified in the parameter immediately after the “INTEGER” instruction. . Here, for convenience, the “INTEGER” instruction is representatively described, but the same applies to other similar declaration instructions such as the “REAL” instruction.
[0020]
(2) The parameter of the “DATA” command consists of two items separated by “/”. The previous item expresses the position of the area secured by the “INTEGER” command. It is called a parameter 21 and the subsequent item expresses an initial value to be set and is called a second parameter 22.
[0021]
Program example 2 in Figure 3 (quoted in the prior art section) sets the initial value "1" to the B (1) position in the B area and sets the initial value "2" to the B (10000000) position. It is. If the initial value is set using the batch save method, there is one management block, but the B (1) and B (10000000) areas of the initial value setting work area are wasteful areas. In addition, when the initial value is set by the divided storage method, there are two management blocks, but the initial value setting work area only needs two 4-byte areas. Therefore, in the case of a huge variable such as program example 2, it can be said that it is a good idea to set an initial value by the divided storage method in terms of both area and processing speed.
[0022]
Program example 3 in Fig. 4 takes an element of `` 2000 x 2000 '' in matrix format, and increments `` I '' of C (I, I) from `` 1 '' to `` 2000 '' by the first parameter (1) However, the initial value “1” is set to “1000” elements C (I, I) by the second parameter. When this is set as the initial value using the batch save method, there is one control block, and an initial value of “1” is set every other initial value setting work area. To do. When the initial value is set by the divided storage method, an initial value setting work area of 4 bytes is required for each of the 1000 management blocks, but the required area is smaller than the former.
[0023]
In this way, it can be said that it is more advantageous to adopt the divided storage method for a program that requires a vast area compared to the experience value N, such as a huge variable or a matrix variable.
[0024]
Program example 4 in FIG. 5 is for continuously setting the same initial value “1” to each element consisting of D (20000). When this is set to the initial value using the batch save method, there is one control block, and the initial value setting work area is continuously set to the initial value “1”. When the value is set, it can be said that it is a special case of the above-mentioned divided storage method. When setting the initial value of the same value as the continuous allocation position, set the “Repetition number of initial value” field in the management block to “20000”. The management block and the initial value setting work area are not newly created, and the initial value can be set with one management block and the initial value setting work area. Therefore, the split storage method is advantageous for a program like this case.
[0025]
Program example 5 in FIG. 6 sets different initial values “1”, “2”,..., “19999”, “20000” in the order of each element of E (20000). If this is set to the initial value using the batch save method, there is one control block, and the initial value setting work area is assigned from E (1) to E (20000) in the order of the assigned position, and the initial value “1” to “20000”. Is set. When the initial value is set by the divided storage method, an enormous number of initial values, that is, “20000” management blocks are required. In addition, the batch storage method is more advantageous in terms of processing time such as chain processing of management blocks.
[0026]
In program example 6 in Fig. 7, "I" of F (I) is incremented from "1" to "20000" by "2" (every other) by the first parameter, and "10000" by the second parameter. An initial value “1” is set in the element F (I). If this is set to the initial value using the batch save method, there is one control block, and an initial value of “1” is set in the initial value setting work area. When the value is set, an initial value setting work area of 4 bytes is required for each 10000 management blocks, and the area becomes larger than the former. Therefore, when the initial value is set discontinuously by specifying the increment value of such a DO type array (instruction for repeated operation within the range by the first parameter) and the number of elements of the initial value setting is greater than the experience value P, The batch storage method is advantageous.
[0027]
In the above, some specific program examples have been given to explain the superiority or inferiority of the initial value setting by the batch storage method and the divided storage method, mainly from the aspect of the occupied work area. Figure 8 shows. The figure shows the “number of management blocks”, “size of management blocks”, “size of the initial value setting work area”, and “total work area size” in which both are added for each of the above program examples. Is a table comparing the batch storage method and the divided storage method. The unit of the size of the area in the table is the number of bytes.
[0028]
Program example 2 and program example 3 are examples in which the experience value N is a huge variable. In such a program, it is a good idea to perform processing by the divided storage method. Program example 4 sets the initial value of the same value as a continuous allocation position, and it is also a good idea to process it using the same division preservation method. Program example 5 is a program having a large number of different initial values with respect to the experience value M. In such a program, it is a good idea to process by the batch storage method. The following program example 6 is a program in which the same initial value is set at discontinuous element positions, and the number of elements of the initial value setting is large relative to the experience value P. Such a program is also a batch storage method. It is a good idea to handle it.
[0029]
As described above, the structure of the present invention holds a plurality of initial value setting means of the batch storage method and the divided storage method in the system, and analyzes the structure of initial value setting by the DATA statement by the experience values N, M, P, etc. In addition, an appropriate initial value setting means can be selected.
[0030]
Subsequently, these processing procedures will be sequentially described with reference to FIGS. FIG. 9 is a flowchart relating to the analysis of instructions in the source program and the selection of the storage method associated therewith. First, in step 50, the flag indicating the batch storage method selection is set to OFF as an initial setting, and the step is set. In 51, the size S of the work area used by the variable is calculated from the declaration instruction such as “INTEGER”.
[0031]
Next, in step 52, it is determined whether or not this calculated S is larger than the conventional experience value N stored in the memory. If it is larger, the processing by the divided storage method is performed, and if not, step 53 is performed. The contents of the second parameter in the declaration command such as “DATA” are decoded, and a different number X of initial values is obtained. Whether X is larger than the conventional experience value M stored in the memory is determined in step 54. If it is larger, the process proceeds to a preparatory process by the batch storage method. As a result of decoding the contents of the two parameters, it is determined whether the initial value assignment structure is discontinuous by specifying an incremental value, etc., and the number of elements at the initial value setting position is larger than the conventional experience value P stored in the memory. In some cases, the process proceeds to a preparatory process based on the batch storage method.
[0032]
If the preparatory processing by the batch storage method is entered in step 54 and step 55, the flag indicating batch storage method selection is set to ON in step 56, and the batch storage method management block in step 57 A work area is secured at the same time.
[0033]
When either the batch storage method or the divided storage method is selected in this way, the initial value setting operation by each storage method is executed according to the procedure of the flowchart as shown in FIG.
[0034]
First, one initial value is acquired in step 60, and it is determined in step 61 whether the batch storage method flag is ON. In the case of the batch storage method (flag ON), after setting an initial value in the element allocation position, The process proceeds to step 66 which is the next initial value process. On the other hand, in the case of the division saving method (flag OFF), in step 62, it is determined whether the program is a program for setting the same initial value to the continuous allocation position (whether it is the same as the initial value of the existing management block). If YES, in step 64, the “initial value repeat count” of the management block is counted up and the process proceeds to step 66 which is the next initial value process. If NO in step 63, a new management block and work area are secured and created, and initial values are set. Then, the process proceeds to step 66, which is the next initial value process. If the initial value setting is not completed in step 66, the process returns to step 60 to acquire a new initial value, and the initial value is sequentially set by repeating the same processing.
[0035]
In the description related to the present invention, “FORTRAN” is described as an example of the program language, but it can be shared in other languages such as PL / 1.
[0036]
【The invention's effect】
As is clear from the above description, according to the present invention, a plurality of initial value setting means are held in the system, and the initial value setting means can be selected according to the structure of initial value setting by the DATA statement, so that the number is smaller. In the work area, the initial value of the DATA statement can be set at high speed, and there is a remarkable industrial effect that compiling time can be shortened.
[Brief description of the drawings]
[Fig. 1] Configuration diagram of the present invention [Fig. 2] Description of general form of instruction [Fig. 3] Comparison of saving methods of program example 2 [Fig. 4] Comparison of saving methods of program example 3 [Fig. 5] Saving of program example 4 Method comparison [Fig. 6] Preservation method comparison of program example 5 [Fig. 7] Preservation method comparison of program example 6 [Fig. 8] Preservation method comparison table of each program example [Fig. 9] Flow chart of instruction analysis and selection of preservation method [Fig. Fig. 10 Flowchart of execution of storage method Fig. 11 Conventional value output method of DATA statement (Part 1) Explanatory diagram Fig. 12 Problems of initial value output method (Part 1) Fig. 13 Conventional DATA statement Initial value output method (part 2)
10 Recognition means
11 Analytical means
12 Preservation method selection method
13 Initial value setting method
20 Area name and area size
21 Area position (first parameter)
22 Initial value (second parameter)

Claims (3)

In a compiler processing apparatus for generating an object program that is a sequence of instruction words that can be interpreted and executed by a computer from a sequence of command statements that define instructions to a computer described in a source program,
Declaration instruction for instructing the compiler apparatus to reserve a variable area that is a memory area used when the object program is executed by a computer, set an initial value in the reserved variable area, and output the object program Recognition means for detecting and recognizing a sentence from the source program;
Analyzing means for analyzing a setting structure of one or more initial values set in the variable area based on the definition of the declaration statement;
Corresponding to the definition of the declaration command statement, one or more initial values set in the variable area are collectively stored in one continuous work area and saved based on the analysis result by the analyzing means. For the storage method and one or more initial values set in the variable area, one work area managed by one management block is assigned to each initial value, and the one or more initial values are set. A storage method selecting means for selecting any one of the storage methods for storing the divided and stored in each of the work areas so as to reduce the total amount of the work areas;
Corresponding to the declaration statement, the one or more initial values are stored in the work area by the storage method selected by the storage method selection means, and the one or more initial values stored in the work area are stored in the work area. Initial value setting means for setting the variable area and outputting the object program;
A compiler processing apparatus characterized by comprising: The analysis of the setting structure of the one or more initial values by the analyzing means includes a size of a variable area defined by the declaration statement, a number of the one or more initial values, a continuity of the one or more initial values, The compiler processing apparatus according to claim 1, wherein the compiler processing apparatus performs the processing based on the number of set positions for the variable area. The source program is a FORTRAN program;
The compiler processing apparatus according to claim 1, wherein the declaration instruction statement is a type declaration statement with an initial value specification and a DATA statement.

Images