JPH06266562A - Object code size optimizing system in object-oriented language processing system - Google Patents

Abstract

PURPOSE:To reduce the size of an object code by generating only one real body of a method address storage table inside the object code. CONSTITUTION:In an object-oriented language processing system for generating output codes for the respective input files of an input file group 11 and realizing the execution time decision of the method of an object-oriented language by using the method address storage table statically generated inside the object code 17, a syntax semantic analysis part 13 compares an input file name with the name of a class to which the method belongs at the time of the semantic analysis processing of a class declaration, generates the real body and the declaration of the method address storage table inside the output code 15 in the case of coincidence and generates only the declaration of the method address storage table inside the output code 15 in the case of noncoincidence. A linker 16 solves reference relation among the plural output codes 15 and generates one object code 17.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention is an object-oriented language (eg, programming languages C ++, LOOPS, T).
AO, etc.), particularly, an output code corresponding to each of a plurality of input files is generated, and then a reference relationship between the plurality of output codes is resolved to generate one object code, and The present invention relates to a method of optimizing a target code size in an object-oriented language processing system, which realizes a runtime decision of a method of an object-oriented language by using a method address storage table statically generated in the target code.

[0002]

2. Description of the Related Art Object-oriented languages have a function called method runtime decision. This is a function for collectively handling objects belonging to different classes, and at the time of execution, determines a method activated for a message given to the object based on the class to which the object belongs.

In some object-oriented language processing systems, this function is realized by using a method address storage table statically generated in the object code. Since the method to be actually executed is determined based on the class to which the object belongs, the contents of the method address storage table are different for each class (more precisely, for each class and each class inheritance relationship).

Conventionally, if a declaration of a class to which a method determined at execution time belongs is included in an input source program file (hereinafter, simply referred to as an input file), unless a user of the object-oriented language processing system specifies otherwise. The substance of the method address storage table was always generated in the output code.

The prior art will be described below in the case of C ++, which is one of the object-oriented languages. C ++ has a mechanism for runtime decision of methods called virtual functions. To specify that it is a virtual function, vi
The keyword rtual is used.

In order to realize a virtual function, it is necessary to determine a method to be called at the time of execution (at the time of program operation) for a given message call. In order for the virtual function to be called correctly, the method to be executed must be determined at the time of program operation based on the information held by each object.

A naturally conceivable method for realizing this is a method in which the C ++ processing system embeds information representing class attributes in the output code and each object holds a pointer to that information. As a matter of course, the information indicating the attribute of the class includes the information of the method to be executed. During program operation, a pointer is traced from the object to access this information to determine the method to be executed.

The virtual function table is a method address storage table which the C ++ processing system embeds in the output code to realize the virtual function function. The virtual function table may be considered to be a simplification of “information indicating the class attribute”. That is, it is not a table that has all the information about the class, but a table that has the minimum information necessary to determine the method to be executed. Here, it may be considered that the virtual function table has the address (position) of the method to be executed as information.

The virtual function table is necessary when a virtual function (function with virtual) is used in a class declaration. The virtual function table must be generated each time the virtual function is rewritten, in other words, each class in which the virtual function is rewritten is declared.

In order to use an object in the input file, the declaration of the class to which the object belongs must be included in the input file. If you want to use an object of a class in multiple input files, the input files must contain declarations for the class. For this reason, the class declaration is usually put in a header file (the file name often ends in .h), and the header file is included in the input file that uses the object (#include is specified in the C ++ language specification). Often specified).
When the header file include is specified, the C ++ processing system replaces the part in which #include is specified with the contents of the included header file.

When the C ++ processing system embeds the virtual function table in the output code, the virtual function table is embedded in the same form as the external variable in the source program. Like variables, virtual function tables need to be given names (similar to variable names). Given the name this way, it is possible for an object to hold a pointer to a virtual function table (which can be referenced by name). Conversely, as long as the name of the virtual function table is known, the substance of the virtual function table (the area secured in the output code in the same form as the external variable described above) is not necessary.

However, what is important is that at least one virtual function table entity is required in the object code generated for the entire program composed of a single or a plurality of input files. Otherwise, during program execution, the object will hold the address of the location where the virtual function table was not actually created.

Therefore, it suffices for the C ++ processing system to perform a process of examining all programs at the same time and generating only one entity of the virtual function table in the target code, but to examine all the programs at the same time. It is not realistic because of problems such as processing time and memory consumption. For this reason, in many C ++ processing systems, an output code is generated for each input file, and names are combined later (in one output code, only the variable name is used for reference, and in another output code, the variable area is referred to). If there is, the process of rewriting the name of the referenced part to the address of the actual area) is performed.

Therefore, in a normal C ++ processing system, the virtual function table is inevitably generated in each output code. This is because, as long as the target is a single input file, it is not known whether the virtual function table entity is generated by the output code for another input file.

[0015]

In the above-mentioned conventional object-oriented language processing system, regarding the method address storage table, if the declaration of the class to which the method determined at execution time belongs is included in the input file,
Unless otherwise specified by the user of the object-oriented language processing system, the entity of the method address storage table was always generated in the output code, so it is essentially sufficient to generate one entity for each class in the target code. There is a problem that a certain method address storage table is generated by the number of input files in which the class declaration is included, and the size of the target code becomes unnecessarily large.

To solve this problem, the user of the object-oriented language processing system generates the substance and declaration of the method address storage table in the output code for each input file, or only the declaration of the method address storage table. There is also a method in which the number of entities of the method address storage table generated in the target code is reduced by specifying whether or not to generate and the reference relationship between a plurality of output codes is reduced (AT & T,
"AT & T C ++ LANGUAGE SYSTEM
RELEASE 2.1 Release Note
s ", pp. 4-16 to 4-18). However, there is a problem that it is difficult for the user of the object-oriented language processing system to judge whether the declaration of the class is included in the input file. It was

In order not to unnecessarily increase the size of the object code, the number of method address storage tables used in the method runtime determination should be reduced unless the meaning of the input code changes. Also, the process of reducing the number of method address storage tables should be automatically performed without the judgment of the user of the object-oriented language processing system.

In view of the above points, an object of the present invention is to store the method address in the object code by controlling the generation of the substance and / or declaration of the method address storage table in the output code by the object-oriented language processing system. An object of the present invention is to provide a target code size optimization method in an object-oriented language processing system in which only one table entity is generated and the size of the target code is reduced.

[0019]

The object code size optimizing method in the object-oriented language processing system of the present invention is such that the output code corresponding to each of a plurality of input files is generated and then a reference is made between the plurality of output codes. An object-oriented language processing system that resolves relationships to generate a single object code, and implements the runtime decision of a method in an object-oriented language by using a method address storage table that is statically created in the object code In, in the semantic analysis process of the class declaration, the class name to which the method belongs and the input file name are compared, and if they match, the method address storage table entity and declaration are generated in the output code, and if they do not match, the method address storage Set up a syntactic and semantic analysis part that generates only the table declaration in the output code. Characterized in that was.

[0020]

The present invention will be described in detail with reference to the drawings.

As an embodiment of the present invention, a source program described in an object-oriented language is input, the syntactic and semantic analysis section converts it into an intermediate language format, and the code generation section converts it into an output code. The following is an example of a compiled object-oriented language processing system that outputs a file to a file.

FIG. 1 is a block diagram showing the configuration of an object code size optimizing system in an object-oriented language processing system according to an embodiment of the present invention. The object code size optimizing method in the object-oriented language processing system of this embodiment is as follows.
, Output code 15, linker 16, purpose code 17
It consists of and.

Referring to FIG. 2, the object code size optimizing process of the syntactic and semantic analyzing unit 13 includes a class declaration and semantic analyzing process step S1, a method address storage table necessity judgment step S2, and an input file name / class name comparison. Step S3, method address storage table entity and declaration generation step S4, method address storage table declaration generation step S5, same name input file existence determination step S6, class declaration existence determination step S7
, Class declaration inclusion step S8, compiled judgment step S9, uncompiled setting step S10
, A same-name input file creation step S11, a class declaration inclusion step S12, and a compilation unprocessed setting step S13.

Next, the operation of the object code size optimizing method in the object-oriented language processing system of the present embodiment thus constructed will be described.

The compiler 12 receives a plurality of input files as the input file group 11. Compiler 12
Activates the syntactic and semantic analysis unit 13 and the code generation unit 14 in order to compile one by one for all the input files constituting the input file group 11.

In the compiler 12, the object code size optimizing process is performed by the syntactic and semantic analysis unit 13.

First, the syntactic and semantic analysis unit 13 performs a semantic analysis process for class declarations in the input file (step S).
1). For details, register the members of the class declaration in the symbol table.

Next, the syntactic and semantic analysis unit 13 checks whether or not the class declaration in the input file requires the method address storage table (step S2). this is,
Based on the specifications of the programming language described in the input file, it is determined by whether the class currently being processed requires the runtime decision of the method.
In the case of the C ++ processing system, it is determined based on whether or not there is a virtual declaration (including the case where #include is included). If no method address storage table is needed, no further processing is done for this class.

When the class currently being processed requires the method address storage table, the syntactic and semantic analysis unit 13 checks whether or not the class name and the input file name are equal (step S3). However, if the input file name includes an additional part (extension) defined by the object-oriented language processing system itself or the information processing system on which the object-oriented language processing system operates, the additional part is excluded. Compare the part with the class name. After that, this process is performed when comparing the class name with the input file name.

If the class name and the input file name are the same, the syntactic and semantic analysis unit 13 generates the substance and declaration of the method address storage table in the output code (step S4), and no more for this class. Is not processed. As a result, the substance and declaration of the method address storage table will be generated in the output code only when the class name and the input file name match. C ++
In the case of the processing system, the virtual function table is generated in the data section in the output code 15, and the attribute of the virtual function table of the class is generated in the symbol table (see FIG. 3).
reference).

When the class name and the input file name are not equal, the syntactic and semantic analysis section 13 generates only the declaration of the method address storage table in the output code (step S5), and the input file having the same name as the class name is created. Compiler 1
It is checked whether the file exists in the input file group 11 given to No. 2 (step S6).

When an input file having the same name as the class name exists in the input file group 11, the syntactic and semantic analysis unit 13
Checks the contents of the input file having the same name as the class name to see if the declaration of the currently targeted class is included (step S7). If the class declaration is included then no further processing is done for this class.

If the class declaration is not included, the syntactic and semantic analysis unit 13 changes the contents of the input file having the same name as the class name to include the declaration of the currently targeted class (step S8). In the case of the C ++ processing system, the attributes of the virtual function table of the class are generated in the symbol table (see FIG. 3). Furthermore, if the compilation process has already been completed (YES in step S9), the compiler 12
Uncompiled input file group 11 recorded by
(Step S10).

In step S6, the input file group 1 in which the input file having the same name as the class name is given to the compiler 12
If it does not exist in 1, the syntactic and semantic analysis unit 13 newly creates an input file having the same name as the class name (step S11), and selects a suitable one of the information processing system operating in the object-oriented language processing system. Place it in the position and include the class declaration in this input file (step S1
2) Further, the input file is included in the uncompiled input file group 11 recorded by the compiler 12 (step S13). In this case, the substance and declaration of the method address storage table are generated in the output code 15 when the newly created input file is compiled.

Multiple output codes 1 by the compiler 12
5 is output, the linker 16 outputs a plurality of output codes 1
One reference code 17 is generated by solving the reference relationship between the five.

FIG. 3 shows the input file X. h, X. C and a. C and output code X. o and a. It is a figure which shows an example of o. Input file X. C and a. Header file X.C. h is included. Class name X and input file name X. The output code X. In o, the substance and declaration of the virtual function table of class X are generated. Also, the class name X and the input file name a. The output code a. Only the declaration of the virtual function table of class X is generated in o.

[0037]

As described above, according to the present invention, the substance of the method address storage table in the output code and //
Alternatively, the object-oriented language processing system has a function of allowing the object-oriented language processing system itself to select whether to generate a declaration and only one entity of the method address storage table is generated in the target code, which makes it difficult for the user of the object-oriented language processing system. The effect is that the size of the target code can be reduced without performing detailed analysis.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an object code size optimizing method in an object-oriented language processing system according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a target code size optimizing process in a syntactic and semantic analysis section in FIG.

FIG. 3 is a diagram showing an example of an input file and an output code in FIG.

[Explanation of symbols]

 11 Input File Group 12 Compiler 13 Syntax Semantic Analysis Section 14 Code Generation Section 15 Output Code 16 Linker 17 Purpose Code

Claims (3)

[Claims] 1. An output code corresponding to each of a plurality of input files is generated, and then a reference relation between the plurality of output codes is resolved to generate one object code, and an object-oriented language method is executed. In an object-oriented language processing system that realizes time determination by using a method address storage table that is statically generated in the target code, the class name to which the method belongs and the input file name are processed during the semantic analysis processing of the class declaration. A syntactic and semantic analysis unit is provided to compare and generate a method address storage table entity and declaration in the output code if they match, and generate only the method address storage table declaration in the output code if they do not match. Objective Code Size Optimization Method for Object-Oriented Language Processing Systems. 2. If the class name and the input file name do not match, all the input file names processed by the object-oriented language processing system are checked, and if there is an input file whose input file name matches the class name. For example, scan the contents of the input file and include the declaration of the class in the input file if it does not contain the declaration of the current target class. If there is no input file whose input file name matches the class name, class name 2. The object code size optimizing method according to claim 1, wherein an input file having the same name as is created, a class declaration is included in the input file, and the input file is included in the input file group. 3. The object code size optimizing method in an object oriented language processing system according to claim 1, wherein the object oriented language is C ++ and the method address storage table is a virtual function table.