JPH0721032A - Program optimization processing system - Google Patents

Abstract

PURPOSE:To provide the program optimization processing system for automatically regenerating a target program while more accelerating the execution without necessity for a user to be conscious of a redundant program or a program with a lot of unnecessary parts in programs prepared by the user. CONSTITUTION:An instruction analysis part 8 retrieves a processing pattern equal to the processing pattern of an instruction in the target program from a data base part 9 and possesses a processing pattern corresponding to that retrieved processing pattern, these processing patterns are executed by an instruction execution part 7, the execution times are compared by an instruction execution time comparison part 6, and an optimizing part 3 replaces the processing pattern in the target program with the processing pattern more accelerating the execution.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a compiler and an optimizing method for automatically converting a processing pattern of a low-speed instruction into an optimum (high-speed) instruction processing pattern in each used instruction. .

[0002]

2. Description of the Related Art Generally, optimization means that the compiler does not need to be aware of redundant programs and wasteful programs among the original programs created by the user, and
The purpose is to create a target program with a faster execution time by considering the efficiency of U and the execution time of each instruction. Conventionally, the execution time of each instruction is compared to meet such a request. , There is a way to optimize the program. For example, the one disclosed in JP-A-60-221832 corresponds to this method.

FIG. 19 is a block diagram of a conventional program optimization method. When the target program is created, each instruction is fetched, and the equivalent processing pattern prepared in advance and the instruction in the table in which its execution time (estimated time) are stored are compared to determine the instruction in the target program. It is to convert to a faster execution time.

In FIG. 19, 19 is a compiler and 20 is a compiler.
Is a control unit, 21 is an instruction fetching unit that fetches instructions from the target program, 22 is an optimization unit that optimizes the instructions fetched by the instruction fetching unit 21, and 23 is a target that uses the instructions optimized by the optimizing unit 22. A target program generation unit for regenerating a program, 24 is an instruction execution time analysis unit for comparing the execution time of the instruction fetched by the instruction fetching unit 21 with the execution time of an equivalent processing pattern prepared in advance,
Reference numeral 25 is an instruction execution time storage table in which execution times of processing patterns are stored in advance. FIG. 20 is a flowchart showing the operation of the conventional program optimization method.

Next, the operation of the program optimization method will be described with reference to FIG. First, when the user's program is compiled, S26 starts and each instruction is traced by the instruction fetching section 21. Next, it is searched whether the traced instruction is stored in the instruction execution time storage table 25. If the instruction traced in S27 is not stored, the instruction is not changed (S29). If stored, the execution time of the instruction of the target program in S28 (however, the execution time is the estimated execution time)
And the instruction execution time (estimated execution time) stored in the instruction execution time storage table are compared by the instruction execution time analysis unit 24. If the execution time (estimated execution time) of the instruction stored in the instruction execution time storage table 25 in S30 is slower, the instruction is not changed (S32). When the execution time (estimated execution time) of the instruction of the object program is slower, the instruction stored in the table is converted by the object program generator 23 (S31). In this way, the process is repeated until all the instructions are completed, and the new target program with the minimum instruction execution time is generated (S33).

[0006]

In the conventional optimization method, even if the execution time of each instruction is reduced, if the program is not executed, the entire program is efficiently optimized. I didn't know if I was there or not. Since the instruction execution time (estimated execution time) differs depending on the model of the computer used, it is necessary to change the instruction execution time storage table depending on the model of the computer used, and the compiler designer must There was a problem in that the estimated execution time) had to be known.

The present invention has been made to solve such problems. The purpose is not to optimize individual instructions, but to optimize the entire program faster. It is also an object of the present invention to be able to perform optimization without depending on the type of computer used and without the compiler designer having to know the instruction execution time (estimated execution time).

[0008]

A program optimizing system according to a first aspect of the present invention has the following elements. (A) First to perform equivalent functions and obtain equivalent processing results
And a storage unit that stores the second processing pattern in association with each other,
(B) an instruction fetching means for inputting a target program and fetching a processing pattern of an instruction used in the target program; (c) a first processing pattern stored in the storage means, and a fetched by the instruction fetching means. An instruction analyzing means for detecting a coincidence of the processing pattern of the target program, (d) a second instruction corresponding to the first processing pattern while executing the instruction of the processing pattern of the target program detected by the instruction analyzing means. Instruction execution means for executing instructions of the processing pattern, (e) execution time comparison means for comparing instruction execution times of the first and second processing patterns by the instruction execution means, (f) comparison result by the execution time comparison means Optimizing means for optimizing the target program using the second processing pattern based on

A program optimizing method according to a second aspect of the present invention is the same as the program optimizing method according to the first aspect, in that the instruction executing means executes the entire target program. The execution time comparison means compares the execution times of the entire target programs.

According to a third aspect of the present invention, there is provided a program optimizing processing method, which is different from the program optimizing processing method according to the first aspect of the present invention in that the storage means has a plurality of instructions arranged in a predetermined order. The second processing pattern in which the instructions are rearranged is stored for the processing pattern of.

A program optimization processing method according to a fourth aspect of the present invention is the program optimization processing method according to the first aspect of the present invention, in which the storage means has a plurality of instructions arranged in a predetermined order. The second processing pattern in which the relative positions of these instructions are changed with respect to the above processing pattern is stored.

A program optimization processing method according to a fifth aspect of the present invention is the same as the program optimization processing method according to the first aspect, wherein the storage means has at least first and second processing. One of the patterns is characterized by being composed of one instruction.

A program optimization processing method according to a sixth aspect of the present invention has the following elements. (A) A storage unit that stores the execution times of the first and second processing patterns while storing the first and second processing patterns that have the same functions as other models and obtain the same processing results. (B) an instruction fetching means for inputting a target program and fetching a processing pattern of an instruction used in the target program; (c) a first processing pattern stored in the storage means, and a fetched by the instruction fetching means. Command analysis means for detecting a match in processing pattern of the target program, (d) first command detected by the command analysis means
Execution time comparing means for searching and comparing execution times of the processing pattern of No. 1 and the second processing pattern having the same function as the first processing pattern from the execution times stored in the storage means, and (f) the execution Optimization means for optimizing the target program using the second processing pattern based on the comparison result by the time comparison means.

A program optimization processing method according to a seventh aspect of the present invention is the same as the program optimization processing method according to the sixth aspect, wherein the storage means corresponds to a plurality of models. It is characterized in that the first and second processing patterns and the execution time are stored.

[0015]

In the program optimizing processing method according to the first aspect of the present invention, the instruction executing means executes the instruction of the processing pattern of the target program, and corresponds to the first processing pattern corresponding to the processing pattern of the target program. The instruction of the processing pattern 2 is also actually executed, the execution time comparison means compares the two execution times, and the optimization means optimizes the target program based on the comparison result. Therefore, compared with the conventional method that optimizes using the estimated execution time of each instruction used in the target program, the optimization is performed using the actual execution time, so a more accurate optimization is required. can do.

A program optimizing method according to a second aspect of the present invention is the program optimization method according to the first aspect, wherein the instruction executing means executes the entire target program, and the execution time comparing means causes the execution time of the entire target program. Therefore, the entire target program can be regenerated into an efficient target program as compared with the first invention.

A program optimizing method according to a third aspect of the present invention is the same as the first aspect, wherein the second processing pattern stored in the storage means rearranges the instructions of the first processing pattern. It is a thing. Therefore, when the optimizing means optimizes the target program based on the comparison result of the execution time comparing means, it is possible to change a certain processing pattern in the target program to the corresponding second processing pattern. , It is possible to automatically rearrange the order of efficient instructions.

A program optimizing method according to a fourth aspect of the present invention is the program optimization method according to the first aspect, wherein the second processing pattern stored in the storage means is a relative instruction of the first processing pattern. The position has been changed. Therefore, when the optimizing means optimizes the target program based on the comparison result of the execution time comparing means, a certain processing pattern in the target program is changed to the corresponding second processing pattern. Thus, the address used in the program can be changed to an efficient address.

A program optimizing method according to a fifth aspect of the present invention is the program optimizing method according to the first aspect, wherein at least one of the first processing pattern and the second processing pattern is constituted by one instruction in the storage means. Therefore, the instructions in the target program can be optimized for each individual instruction.

In the program optimizing method according to the sixth aspect of the present invention, the storage means stores the first and second processing patterns of other models and the execution times of these processing patterns at the same time. Therefore, the execution time comparison means searches the two execution times stored in the storage means without executing the first and second processing patterns that match the processing pattern of the target program. Two execution times can be compared, and programs of other models can be optimized.

A program optimization method according to a seventh aspect of the present invention is the program optimization method according to the sixth aspect, wherein the storage means has the first and second processing patterns for a plurality of models, and these processing patterns. Since the execution time of the program is stored, if the model for executing the compiled program is designated in advance and then the compilation is performed, the program can be optimized even for the model different from the model for executing the program.

[0022]

【Example】

Example 1. FIG. 1 is an overall configuration diagram of a program optimization processing method in this embodiment. In FIG. 1, 10 is a general compiler flow. Reference numeral 1 is a compiler, 2 is a control unit, and 3 is an instruction execution time comparison unit, which is an optimization unit that selects an instruction having a fast instruction execution time (actual measurement value). Reference numeral 4 is a target program generation unit for converting into an instruction selected by the optimization unit, 5
Is an instruction fetch unit for fetching the instruction of the target program, 6
Is an instruction execution time comparison unit that compares execution times (actually measured values) of programs executed by the instruction execution unit. Reference numeral 7 denotes an instruction execution unit that executes the target program and a program modified into instructions equivalent to the instructions used in the target program. Reference numeral 8 is an instruction analysis unit that searches for equivalent processing patterns, and 9 is a database unit in which equivalent processing patterns are stored. Reference numeral 11 is a new-purpose program that has been optimized. Further, the processing can be executed even if the processing patterns of a plurality of instructions are stored in the database unit.

FIG. 2 is a flow chart showing the execution procedure of the program optimization processing method. FIG. 3 is a diagram showing a user program used in this embodiment, and FIG.
It is a figure which shows the processing pattern stored in the database part. Here, the processing pattern is a set of one or more instructions. FIG. 5 is a flowchart showing the procedure for measuring the instruction execution time.

In this embodiment, an equivalent processing pattern is retrieved from a database section in which processing patterns of instructions equivalent to the instructions used in the target program are stored. The instruction pattern used in the target program and the equivalent instruction stored in the database unit are executed and compared to select the optimum instruction processing pattern. A detailed description will be given below.

The procedure of the program optimization processing method will be described with reference to the flow charts of FIGS. 2 and 5 and using the program example of FIG. 3 and the storage example of the database section of FIG. First, compiling the user's program as shown in Fig. 3 generates the target program,
S11 starts and each instruction used in the target program is traced by the instruction fetching section 5. Next, the instruction analysis unit 8 searches whether the traced instruction is stored in the database unit 9. If the instruction traced in S12 is not stored, the instruction is not changed (S14).
If it is stored, in S13, the instruction execution unit 7 executes the processing pattern of the target program and the processing pattern of the instruction equivalent to the processing pattern of the used instruction. S1
The instruction execution time comparison unit 6 compares the execution time (actual measurement value) of the processing pattern of the target program executed in 5 with the processing pattern of the equivalent instruction.

The comparison of the execution times is performed by the procedure shown in FIG.
First, an instruction is set in the instruction execution time comparison unit 6 (S1
9), the instruction execution time comparison unit 6 checks the instruction execution start time with a timer (S20). Next, when the execution of the instruction is completed, the time at the end is checked by the timer (S2
1) The execution time is calculated from the start time and the end time (S22). The processing from S19 to S22 is performed for the processing pattern of the target program and the processing pattern of the equivalent instruction, the execution time is calculated, and these two calculated execution times are compared. If the execution time (measured value) of the processing pattern of the equivalent instruction is slower, the instruction is not changed (S17). If the execution time (measured value) of the processing pattern of the target program is slower, the processing pattern of the instruction stored in the database unit 9 is selected by the optimization unit 3 and selected by the target program generation unit 4. It is converted into an instruction (S16).

For example, in S12, the instruction corresponding to the instruction of FIG. 4a is stored in the database unit 9 as shown in FIG. 4b, and the execution times are compared in S15. When the execution time of the instruction of FIG. 4b is faster, the instruction of FIG. 4a in the target program is converted into the instruction of FIG. 4b in S16. The above steps S11 to S17 are repeated until all the instructions are completed, and the new object program 11 with the minimum instruction execution time is generated. Further, it can be executed even if the processing patterns of a plurality of instructions are stored in the database unit.

As described above, in this embodiment, the database section in which the instruction processing type (pattern) equivalent to the instruction processing type (pattern) used in the target program is stored and the database section are searched. It has an instruction analysis unit, an instruction execution unit that executes instructions with a processing pattern equivalent to the processing pattern of the target program, and an instruction execution time comparison unit that compares instruction execution times of processing patterns equivalent to the processing pattern of the target program. However, based on the result of comparison of the instruction execution times of the processing pattern equivalent to the processing pattern in the target program, the instructions can be replaced so that the program can be optimized so as to be executed at a higher speed.

Example 2. Further, the first embodiment is a pattern optimizing means for performing optimization by comparing the execution time of the processing pattern of the instruction of the target program with the execution time of the equivalent processing pattern. However, instead of executing only the pattern,
A program optimizing means that performs optimization by executing the entire program after replacing it with an instruction processing pattern equivalent to the instruction processing pattern used in the program and comparing the execution times Is possible.

Example 3. Further, in the first embodiment, the program optimization is realized by replacing the instructions, but the program optimization can also be realized by rearranging the instructions. A detailed description will be given below.

FIG. 6 is a diagram showing a user program used in this embodiment, and FIG. 7 is a diagram showing processing patterns stored in the database unit. FIG. 8 is a flow chart showing the execution procedure of the program optimization processing method in this embodiment. The overall block diagram of the program optimization processing method in this embodiment is the same as that in FIG.

Next, referring to the flow chart of FIG.
The procedure of the program optimization processing method will be described with reference to the program example of FIG. 7 and the storage example of the database unit of FIG. First, when a user program as shown in FIG. 6 is compiled, a target program is generated, and the instruction fetching unit 5 traces each instruction of the target program (S1
1a), it is searched whether the traced instruction is stored in the database unit 9 (S12a). If it is stored, the instruction execution unit 7 executes the processing pattern of the target program and the corresponding processing pattern, and the instruction execution time comparison unit 8 compares these two execution times (S).
15a).

For example, if the database section stores the processing patterns as shown in FIG. 7, c and d in FIG. 7 are executed and the execution times are compared. As a result of the comparison, if c in FIG. 7 is faster, the instruction remains unchanged (S17a), and if c in FIG. 7 is slower, the processing pattern in c in FIG. , D in FIG.
Is converted into the processing pattern (S16a). Since the processing pattern of d of FIG. 7 is a rearrangement of the instructions in the processing pattern of c of FIG. 7, it means that a certain processing pattern in the target program is rearranged in the processing of S16a.

As described above, in this embodiment, the instructions are rearranged by using the processing patterns stored in the database. By comparing the instruction execution times and replacing the instructions with the rearranged processing pattern, the instructions appearing in the program are automatically rearranged in a more efficient order.

Example 4. In addition, the optimization method of the first embodiment is an optimization of the instruction processing pattern, and it is possible to change the address used in the program by using this method and arrange it at the optimum address. A detailed description will be given below.

FIG. 9 is a diagram showing a user program used in this embodiment, and FIG. 10 is a diagram showing processing patterns stored in the database section. FIG. 11 is a flowchart showing the execution procedure of the program optimization processing method in this embodiment. The overall block diagram of the program optimization processing method in this embodiment is the same as that in FIG.

Next, the procedure of the program optimization processing method will be described with reference to the flow chart of FIG. 11 and using the program example of FIG. 9 and the storage example of the database section of FIG. First, when a user program as shown in FIG. 9 is compiled, a target program is generated, each instruction of the target program is traced by the instruction fetching section 5 (S11b), and the traced instruction is stored in the database section 9? If it is retrieved (S12b) and stored, the instruction execution unit 7 executes the processing pattern of the target program and the corresponding processing pattern, and the instruction execution time comparison unit 8 compares these two execution times ( S15b).

For example, if the database section stores the processing patterns as shown in FIG. 10, the steps e and f in FIG. 10 are executed and the execution times are compared. As a result of the comparison, when the speed of FIG. 10e is higher, the instruction remains unchanged (S17b), and when the speed of FIG. 10e is lower, the processing pattern of FIG. Converted to a processing pattern (S16)
b). In this case, this change causes the instruction address in the target program to shift by the amount of the "NOOP" instruction.
There is no delay in executing this instruction on a pipeline controlled computer.

As described above, this embodiment enables the optimization method in which the instruction address is changed at the same time as the instruction replacement is performed based on the result of execution of the object program.

Example 5. In the optimization method of the first embodiment, when the instruction processing pattern is optimized, the instruction replacement is performed. However, in the present embodiment, the same optimization means as in the first embodiment is used. The feature is that one instruction used in the program is changed. A detailed description will be given below.

FIG. 12 is a diagram showing a user program used in this embodiment, and FIG. 13 is a diagram showing instructions stored in the database unit. FIG. 14 is a flow chart showing the execution procedure of the program optimization processing method in this embodiment. The overall block diagram of the program optimization processing method in this embodiment is the same as that in FIG.

Next, the procedure of the program optimization processing method will be described with reference to the flow chart of FIG. 14 and using the program example of FIG. 12 and the storage example of the database section of FIG. First, a user program as shown in FIG. 12 is compiled to generate a target program, and the instruction fetching section 5 traces each command of the target program to read the DIVIDE command (S11c). The instruction analysis unit 8 searches for the DIVIDE instruction stored in the database unit 9 (S12c). If the DIVIDE instruction is not stored, the instruction is left as it is. If it is stored, the instruction execution unit 7 executes the processing pattern of the target program and the corresponding processing pattern. Instruction execution time comparison unit 8
And measure these two execution times in (S13c),
The measured execution times are compared (S15c).

For example, it is assumed that the database section stores the processing pattern as shown in FIG. First, Fig. 1
The DIVIDE command g of No. 3 is executed, and the corresponding SHIFT command h is executed to measure the execution time.
The two measured execution times are compared, and if g in FIG. 13 is faster, the instruction is left as it is (S17).
c). When the execution time of g in FIG. 13 is slower,
The instruction of g in FIG. 13 in the target program is converted into the instruction of h in FIG. 13 (S16c).

Assuming that the instruction fetched by the instruction fetching section 5 (S11c) is a division by a power of 2 (fixed point division) instruction, the instruction analyzing section 8 executes an equivalent process from the database section 9. Search for a shift instruction that is an instruction to execute. Next, the instruction execution unit 7 executes the retrieved shift instruction and division instruction. As a result of execution, the instruction execution time comparison unit 6 compares the instruction execution times. Since fixed-point division is generally low-speed, the instruction execution time comparison unit 6 selects a shift instruction having a short execution time. Based on the selected result, the optimization unit 3 converts the instruction (division instruction) of the target program, the target program is optimized, and a new target program is generated.

In the above-described first to fourth embodiments, a group of a plurality of instructions in the target program is treated as a processing pattern. However, in this embodiment, even if one instruction in the target program is processed, the optimum operation of the first embodiment is performed. It shows that the means of conversion can be used.

Example 6. In this embodiment, the execution times of the first processing pattern and the second processing pattern corresponding to the first processing pattern are stored in the database section for each machine type that executes the program, so that the execution time of FIG. It is characterized in that the instruction execution unit 7 is eliminated from the above. A detailed description will be given below.

FIG. 15 is an overall block diagram of the program optimization processing system in this embodiment. In FIG. 15, the constituent elements that make up the overall configuration diagram are the same as those in FIG. 1 of the first embodiment, but the instruction execution unit 7 does not exist. The database unit 9a also stores the execution times of the first processing pattern and the corresponding second processing pattern for each model that executes the program. FIG. 16 is a diagram showing the stored contents stored in the database unit.
FIG. 7 is a flow chart diagram showing an execution procedure of a program optimization processing method in the present embodiment.

Next, the procedure of the program optimization processing method will be described with reference to the flowchart of FIG. In this embodiment, the user program to be used uses the same program as that shown in FIG. 12 as in the fifth embodiment, and the execution model for compiling and executing this program is Assume model A. First, the model that executes the actually compiled program is specified at the time of compilation (S10). The instruction fetching section 5 fetches an instruction from the target program, and the instruction analyzing section 8 searches the database section 9a for an instruction corresponding to the instruction (S11).
d). As a result of the search, if the corresponding model and instruction are stored, the instruction execution time comparison unit 6 compares the execution times of the stored instructions (S13d). The instruction execution time comparison unit 6 selects the fastest instruction among the retrieved instructions (S15d) and changes it to that instruction (S16d).

For example, the instruction fetching section 5 fetches a DIVIDE instruction from the target program created by compiling the user program shown in FIG. 12, and the instruction analyzing section 8 fetches the "A" from the database section as shown in FIG. For the DIVIDE command. Since the DIVIDE command of the model A exists, the DIVIDE command is executed by the command execution time comparison unit 6.
Execution time for IDE command TimeA and SHIFT
The execution time TimeB corresponding to the instruction is compared. Ti
If meA> TimeB, the DIVIDE command is SH
Replace with IFT instruction. It is not changed when the instruction of the target program is the fastest (S17d). It also does not change the instruction when the corresponding instruction is not stored in the database. In this way, the processing from S10 to S17d is repeated until all the instructions are completed, and the new target program with the minimum instruction execution time for any model is generated.

As described above, this embodiment is provided with a database section having a table structure of machine types, instructions and instruction execution times, and stores the instructions and instruction execution times of a plurality of types of machines in this database section. There is. Then, the instruction fetch section and the instruction analysis section search for a processing pattern equivalent to the target program. If the data of the model you want to execute the compiled program is stored in the database part, do not actually execute the processing pattern and compare the execution time, but compare the execution time stored in advance in the database part. It is characterized by performing optimization by.

Example 7. Here, 10 in FIG. 1 is an example of a compiler, and the flow is not necessarily required. For example, when a target program is created from a source program created using a language in which syntax analysis-intermediate file creation processing is performed twice, 10 in FIG.
It becomes the flow like a. Even in such a flow, the method of optimizing the target program can be any method of the above-described first to sixth embodiments.

[0052]

As described above, according to the present invention, it is possible to optimize not the individual instructions but the entire program to be faster. Further, there is an effect that the compiler designer can perform the optimization without depending on the type of the computer to be used and without knowing the instruction execution time (estimated execution time).

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a program optimization processing system according to a first embodiment of the present invention.

FIG. 2 is a flowchart showing an execution procedure of a program optimization processing method according to the first embodiment of the present invention.

FIG. 3 is a diagram showing a user program according to the first embodiment of the present invention.

FIG. 4 is a diagram showing processing patterns stored in a database unit according to the first embodiment of the present invention.

FIG. 5 is a flowchart showing a procedure for measuring instruction execution time according to the first embodiment of the present invention.

FIG. 6 is a diagram showing a user program according to a third embodiment of the present invention.

FIG. 7 is a diagram showing processing patterns stored in a database unit according to the third embodiment of the present invention.

FIG. 8 is a flowchart showing an execution procedure of a program optimizing processing method according to the third embodiment of the present invention.

FIG. 9 is a diagram showing a user program in Embodiment 4 of the present invention.

FIG. 10 is a diagram showing processing patterns stored in a database unit according to the fourth embodiment of the present invention.

FIG. 11 is a flowchart showing an execution procedure of a program optimizing processing method according to the fourth embodiment of the present invention.

FIG. 12 is a diagram showing a user program in Embodiment 5 of the present invention.

FIG. 13 is a diagram showing instructions stored in a database unit according to the fifth embodiment of the present invention.

FIG. 14 is a flowchart showing an execution procedure of a program optimizing processing method according to the fifth embodiment of the present invention.

FIG. 15 is an overall configuration diagram of a program optimization processing method according to a sixth embodiment of the present invention.

FIG. 16 is a diagram showing execution time by processing pattern for each model stored in a database unit according to the sixth embodiment of the present invention.

FIG. 17 is a flowchart showing an execution procedure of a program optimizing processing method according to the sixth embodiment of the present invention.

FIG. 18 is a diagram showing a flow of a compiler in the seventh embodiment of the present invention.

FIG. 19 is an overall configuration diagram of a conventional program optimization processing method.

FIG. 20 is a flowchart showing a procedure of executing a program optimization processing method in a conventional example.

[Explanation of symbols]

 1, 19 compiler 2, 20 control unit 3, 22 optimization unit 4, 23 target program generation unit 5, 21 instruction fetch unit 6 instruction execution time comparison unit 7 instruction execution unit 8 instruction analysis unit 9 database unit 10 general compiler Flow 11 New purpose program 24 Instruction execution time analysis unit 25 Instruction execution time storage table

Claims (7)

[Claims] 1. A program optimization processing method having the following elements: (a) A first function which has equivalent functions and obtains equivalent processing results.
Storage means for storing the second processing pattern and the second processing pattern in association with each other, (b) an instruction fetching means for inputting a target program and taking out a processing pattern of a command used in the target program, (c) stored in the storage means A first processing pattern and an instruction analyzing means for detecting a match between the processing pattern of the target program extracted by the instruction extracting means, and (d) executing an instruction of the processing pattern of the target program detected by the instruction analyzing means. In addition to the above
Instruction executing means for executing an instruction of the second processing pattern corresponding to the processing pattern of (e) execution time comparing means for comparing the instruction execution times of the first and second processing patterns by the instruction executing means, ) Optimizing means for optimizing the target program using the second processing pattern based on the comparison result by the execution time comparing means. 2. The program optimizing process according to claim 1, wherein the instruction executing means executes the entire target program, and the execution time comparing means compares the execution times of the entire target program. method. 3. The storage means stores, for a first processing pattern in which a plurality of instructions are arranged in a predetermined order, a second processing pattern obtained by rearranging these instructions. 1. The program optimization processing method described in 1. 4. The storage means stores a second processing pattern in which the relative positions of these instructions are changed with respect to the first processing pattern in which a plurality of instructions are arranged in a predetermined order. The program optimization processing method according to item 1. 5. The program optimization processing method according to claim 1, wherein said storage means is configured by at least one of one of the first and second processing patterns. 6. A program optimization processing method having the following elements: (a) The first and second processing patterns that have the same functions as those of other models and obtain the same processing result are stored, and And a storage means for storing the execution time of the second processing pattern, (b) an instruction fetching means for inputting a target program and fetching a processing pattern of a command used in the target program, (c) stored in the storage means. A first processing pattern and a command analyzing means for detecting a match between the processing patterns of the target program fetched by the command fetching means, (d) the first processing pattern detected by the command analyzing means, and Execution time comparing means for searching and comparing the execution time of the second processing pattern having a function equivalent to that of the processing pattern from the execution time stored in the storage means, Based on the comparison result by the execution time comparing means, optimizing means for optimizing using the second processing pattern object program. 7. The program optimization processing method according to claim 6, wherein said storage means stores said first and second processing patterns and execution times corresponding to a plurality of models.