JPH0795275B2 - Compile processor - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] A compile processing device for compiling a source program including a mathematical function such as a square root function, and whether or not the mathematical function can be converted into another mathematical function capable of higher speed processing. In a compile processing device for compiling a source program (9) including a mathematical function for the purpose of optimizing the processing of the mathematical function by recognizing and performing the conversion, the source program (9) is analyzed and intermediate A syntax analysis unit (3) for outputting a language, a condition storage unit (12) for setting conditions for converting a mathematical function into another mathematical function in a usage pattern, contents of the condition storage unit (12) and the syntax analysis A conversion unit (6) for converting the mathematical function into the different mathematical function based on the intermediate language output by the unit (3), and a conversion result of the conversion unit (6). Constructed from object code generator for generating an object code and (7).

[Industrial application field]

The present invention relates to a compilation processing device, and more particularly, to a compilation processing device for compiling a source program including a mathematical function such as a square root function.

In many programs used in the field of scientific and technological calculation, many basic mathematical functions (square root function, exponential function, logarithmic function, trigonometric function, etc.) occupy a large cost (occupy a large proportion of execution time).

In such a program, the speedup of the processing speed by optimizing the processing of the mathematical function has a great effect on the improvement of the execution performance, that is, the substantial reduction of the execution time of the entire program.

[Conventional technology]

FIG. 4 is a diagram showing an example of conventional compilation processing of a mathematical function.

When a source program is input (), the compiler parses it () and generates an intermediate language (code) equivalent to the source program according to the analysis result ().

When the source program is described by FORTRAN as F = Q / SQRT (X + Y)-(1), the intermediate language is, for example, It is said that Where SQRT is the square root function and t1 and t2 are the work variables generated by the compiler.

Next, the compiler generates a target (object) code according to the generated intermediate language (). Therefore,
The program of the square root function SQRT is called from the function library that stores the program for processing mathematical functions ().

In the processing of the square root function SQRT, t2 = SQRT in equation (2)
When processing (t1), in reality, due to the processing convenience, The above processing is performed.

Therefore, equation (1) is actually Is processed as follows.

[Problems to be solved by the invention]

In the above-mentioned conventional technique, only the performance improvement of the function itself (function program stored in the function library) is mainly considered as means for optimizing the processing of the mathematical function, and how the mathematical function is used in the source program. The usage pattern of whether or not it was done was not considered.

That is, when the mathematical function SQRT that is processed as shown in equation (3) is used in the source program as shown in equation (1), By performing the conversion as described above, it is possible to omit the division twice and process at high speed as compared with the expression (4), but the processing according to the expression (5) is not performed. .

In other words, the usage form of the mathematical function is recognized, and the mathematical function is converted into another mathematical function, that is, the calling form from the function library is converted ((3) (Or only the corresponding functional program is called). That is, since the conventional compiler has no means for optimization by recognizing the usage form of the mathematical function, it only generates the target code corresponding to the mathematical function as it is in the description of the source program.

The present invention provides a compile processing device capable of optimizing the processing of a mathematical function by recognizing whether or not the mathematical function can be converted into another mathematical function capable of higher speed processing and performing the conversion. The purpose is to do.

[Means for solving problems]

FIG. 1 is a block diagram of the principle of the present invention, showing a processing device equipped with a compiler according to the present invention.

In FIG. 1, 1 is a processing unit including a central processing unit (CPU) and memory, 2 is a compiler, 3 is a syntactic analysis unit, 4 is an optimization processing unit, 5 is a recognition unit, 6 is a conversion unit, and 7 is Target code generator, 8 is a combined edit processor, 9 is a source program, 10 is an object module, 11 is a function library, 12 is a conversion table, 1
3 is a load module.

The source program 9 written in a high-level language such as FORTRAN is translated into machine language code by the compiler 2,
The set of object codes that are the translation results are stored as the object module 10. The linkage edit processing unit 8 such as the linkage editor links the function program called from the function library 11 and the target code (object module 10) according to the target code so that the function program can be immediately referred to (executable format Edit the load module 13 and store it.

In the compiler 2, the syntax analysis unit 3 analyzes the source program 9 and expands it into an intermediate language (before optimization).
The optimization processing unit 4 optimizes the intermediate language by considering the usage form of the mathematical function. With respect to this optimized intermediate language, the object code generator 7 generates an object code and expands it in the object module 10.

The optimization processing unit 4 performs the following processing.

The recognition unit 5, which receives the intermediate language from the parsing unit 3, as input,
Recognize whether or not a mathematical function included in the intermediate language can be converted into another mathematical function that can be processed at higher speed. This recognition is performed by checking whether the mathematical function is used in a manner satisfying a predetermined condition.

For this purpose, for example, the conversion table 12 is prepared. In the change table 12, a predetermined condition that should be satisfied for conversion, another mathematical function after conversion, and its use condition are registered. That is, the recognizing unit 5 first recognizes the usage form of the mathematical function, and then determines whether or not this matches a predetermined condition registered in the conversion table 12 by referring to the conversion table 12 (when the matching condition is satisfied, conversion is possible. And)). The conversion table 12 may be provided in the function library 11 or the optimization processing unit 4.

The conversion unit 6 converts the mathematical function convertible by the recognition unit 5 into another mathematical function by referring to the conversion table 12, and performs necessary processing associated with the conversion.

The function library 11 is configured to include a function program for a mathematical function before conversion and a function program for a mathematical function after conversion.

[Action]

The optimization processing unit 4 of the compiler 2 performs a compile process of automatically converting to another mathematical function when the usage form of the mathematical function satisfies a predetermined condition. Further, as a process associated with this conversion, the optimization processing unit 4 performs a compile process for converting a part of the intermediate language in order to use another mathematical function according to its use condition.

As a result, the load module 13 described by using another mathematical function capable of faster processing can be obtained regardless of the description in the source program 9 while maintaining the equivalence with the source program 9.

As described above, as a result of the optimization of the mathematical function for high-speed processing, when the load module 13 (source program 9) is executed in the processing device 1, the execution time can be shortened.

〔Example〕

FIG. 2 is an explanatory view of an embodiment of the present invention and shows a simple concrete example.

It is assumed that the source program 9 is written in FORTRAN, and there is a statement similar to the above-mentioned expression (1) in the figure, as shown in the figure.

When the source program 9 is input to the compiler 2, the syntax analysis unit 3 analyzes it and generates a first intermediate text 14 made of an intermediate language before optimization. At this time,
The equation (1) is developed in the same manner as the above equation (2) as shown in the figure.

The first intermediate text is input to the optimization processing unit 4, and the second intermediate text 15 composed of the optimized intermediate language is generated. At this time, equation (2) is It is being developed like. Here, XSQRT is a function for obtaining the reciprocal of the square root, and t3 is a work variable generated by the compiler 2. Also, regarding the relationship between the functions SQRT and XSQRT, if the variable is t4, It is represented as. Normal, To obtain, the Newton-Raphson method is used. Then, the process of taking the reciprocal is performed. That is, To obtain, the process of taking the final reciprocal is not necessary. Since the process of taking the reciprocal is realized by one division, Than the function SQRT (x) The function XSQRT (x) that obtains requires only one division process.

Comparing equations (2) and (6) is as follows.
The process of obtaining the variable t3 is
Since the number of divisions is one less than in the process of obtaining the variable t2, the process is performed at a higher speed. In addition, since the processing of Q * t3 can be processed faster than multiplication, it can be processed faster than that of Q / t2. Therefore, it can be said that the expression (6) is converted into a format suitable for processing the expression (2) at a higher speed.

In this way, the optimization processing unit 4 recognizes in the recognition unit 5 that the mathematical function SQRT can be converted into a mathematical function XSQRT that can be processed at higher speed, and the conversion unit 6 converts the conversion into the first intermediate text 14 (expression ( 2)) to obtain the second intermediate text 15 (equation (6)). This allows the mathematical function SQ
The RT is optimized in consideration of its usage pattern (used as a divisor of division) in the source program 9.

When the second intermediate text 15 is input, the object code generator 7 develops the object module 10 based on this.

The binding edit processing unit 8 calls the mathematical function XSQRT from the function library 11 according to the input object module 10. Then, the object module 10 and the mathematical function are combined and edited to generate the load module 13. Therefore, the load module 13 is composed of a machine language code described by using the mathematical function XSQRT capable of higher speed processing, although the source program 9 is described by the mathematical function SQRT.

When the processing device 1 executes the load module 13, as compared with the case where the load module according to the compile process (process for obtaining the first intermediate text 14) directly corresponding to the description of the source program 9 is executed, The optimization can shorten the execution time. Since the processing for this purpose is automatically performed by the optimization processing unit 4, there is no burden on the programmer, and the portability is not impaired by introducing a special function into the source program.

FIG. 3 is a flow chart of optimization processing in the optimization processing unit 4.

When the first intermediate text 14 is input to the optimization processing unit 4, the recognition unit 5 recognizes that the mathematical function used is SQRT. In other words, it recognizes that the mathematical function called from the function library 11 is SQRT.

Further, the recognition unit 5 recognizes that the value (t2) of the result of the processing using the mathematical function SQRT is referred to as the divisor of division.

The recognizing unit 5 judges whether or not the use form of the mathematical function SQRT obtained by the processing conforms to the condition that the mathematical function SQRT can be converted into another mathematical function that can be processed at higher speed. For this purpose, the recognition unit 5 refers to the conversion table 12.

In the conversion table 12, when the value (t2) of the result of the processing using the mathematical function SQRT is referred to as the divisor of division, it is registered that it can be converted into another mathematical function XSQRT.

The recognition unit 5 determines whether or not the value (t2) obtained as a result of the process using the mathematical function SQRT is referenced elsewhere in the source program 9 (first intermediate text 14).

The following (a) mathematical function SQRT is used in the recognition unit 5.

(B) The value t2 of the processing result is used as a divisor for division. That is, this usage pattern is registered in the conversion table 12.

(C) The value t2 of the processing result is not referenced elsewhere.

When the instruction sequence is recognized, the conversion unit 6 performs the following processing.

The conversion unit 6 refers to the conversion table 12 and refers to the mathematical function SQ.
Convert RT to another mathematical function XSQRT. That is, t2 = SQRT (t1) in the equation (2) is converted to t3 = XSQRT (t1) in the equation (6).

Further, the conversion unit 6 refers to the conversion table 12 and
Learn the usage pattern of another mathematical function XSQRT when performing conversion such as processing, and perform the necessary processing. That is, the mathematical function SQ
The division (F = Q / t2 of the equation (2)) that refers to the value t2 of the processing result of RT as a divisor is multiplied by the value t3 of the processing result of another mathematical function XSQRT as a multiplier ((6 ) F =
Convert to Q * t3). As a result, the equation (6), which is the same as the equation (1) of the source program 9 as in the equation (2), is obtained.

〔The invention's effect〕

As described above, according to the present invention, in a compilation process of a source program including a mathematical function, the mathematical function can be converted into another mathematical function that can be processed at higher speed by considering the usage form of the mathematical function. Therefore, the execution time of the entire program can be shortened.

[Brief description of drawings]

FIG. 1 is a block diagram of the principle of the present invention, FIG. 2 is an explanatory diagram of an embodiment, FIG. 3 is a flow chart of optimization processing, and FIG. 4 is a diagram showing an example of conventional compilation processing. 1 is a processor, 2 is a compiler, 3 is a syntax analysis unit, 4 is an optimization processing unit, 5 is a recognition unit, 6 is a conversion unit, 7 is a target code generation unit, 8 is a combined edit processing unit, 9 is a source program , 10 is an object module, 11 is a function library, 12 is a conversion table, 1
3 is a load module, and 14 and 15 are intermediate texts.

Claims (1)

[Claims] 1. A compiling processor for compiling a source program (9) including a mathematical function, comprising: a parsing section (3) for analyzing the source program (9) and outputting an intermediate language; Based on the condition storage unit (12) for setting a condition for conversion into another mathematical function, the content of the condition storage unit (12) and the intermediate language output by the syntax analysis unit (3), A compilation processing device comprising: a conversion unit (6) for converting into another mathematical function, and a target code generation unit (7) for generating a target code for the conversion result of the conversion unit (6).