JP2001092671A - Program execution optimizing method and program managing device and program executing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To attain the optimization of a program corresponding to the execution environment of the program by considering balance among the transferring time, executing time, and memory usage of the program. SOLUTION: A program managing device 1 dynamically optimizes and distributes a program according to the execution environment by referring to the available memory usage of a program executing device 2, program characteristics 141 storing the characteristics of a program to be executed by the program executing device 2, and client characteristics 142 storing the execution environment of the program executing device 2. The program executing device 2 measures the execution frequency data of a program module constituting the program, and communicates it to the program managing device 1. The program managing device 1 updates the program characteristics 141 and the client characteristics 142 according to the execution frequency data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for optimizing program execution, a program management device and a program execution device, and more particularly, to a program taking into account the balance among program transfer time, program execution time and memory usage. The present invention relates to a program execution optimization method, a program management device, and a program execution device for performing optimization according to the execution environment of a program.

[0002]

2. Description of the Related Art With the development of networks and the spread of terminals, a client-server system in which a program is stored in a server and the client downloads and executes the program as necessary has become widespread. In this client-server system, in order to reduce the download time, a method of compressing and transferring a program, and expanding and executing the program on a received client is widely used. Examples of such a method of compressing and transferring a program include a technique disclosed in Japanese Patent Application Laid-Open No. Hei 5-313904 and a Java JAR developed by Sun Microsystem Co., USA.
There is a format.

On the other hand, a virtual machine system is known in which a program is described in an environment-independent format called bytecode or intermediate code, and is sequentially converted into a format that can be directly executed in a loaded environment. In addition, in the virtual machine system, in order to avoid a decrease in execution speed due to execution while sequentially converting the format,
There is known a translator system in which a program described in a bytecode or an intermediate code is previously converted into a format that can be directly executed in an environment in which the program is loaded, and then transferred.

[0004]

The method of compressing and transferring a program can shorten the transfer time. Further, the memory for storing the downloaded program can be saved. However, since extra time is required for the development, there is a problem that the execution time becomes longer in terms of the time required from the download request to the execution of the program.

In the virtual machine system, extra time is required to convert a program described in byte code or intermediate code into a format that can be directly executed in an execution environment. Therefore, from the viewpoint of the time required from program loading to execution. There is a problem that the execution time becomes long.
Although this problem can be solved by using a translator, there is a problem that the transfer time is long and the memory usage is greatly increased.

In other words, the prior art does not take into account the balance between the transfer time of the program, the execution time, the amount of memory used, and the like, and there has been a problem that optimization according to the execution environment of the program has not been performed. . Therefore, an object of the present invention is to provide a program execution optimizing method, a program management device, and a program for performing optimization according to a program execution environment in consideration of a balance between a program transfer time, an execution time, and a memory usage. An execution device is provided.

[0007]

According to a first aspect, the present invention includes an execution frequency of each program module constituting a program executed on an electronic computer and designation of whether or not the program module should be compressed. A program module to be compressed is selected based on at least one of the optimization information, only the selected program module is compressed, and a program in which a compressed program module and a non-compressed program module are mixed is distributed. There is provided a program execution optimization method characterized by being developed and executed at a distribution destination. In the program execution optimization method according to the first aspect,
A program module that is executed less frequently or a program module that is specified to be compressed is compressed, but a program module that is frequently executed or a program module that is specified not to be compressed is not compressed. As described above, since the compressed program module and the non-compressed program module are mixed, the advantage of reducing the transfer time of the program, the advantage of saving the memory for storing the program, and the disadvantage of requiring extra time for decompression are provided. The balance can be achieved, and optimization according to the execution environment of the program can be performed.

According to a second aspect, the present invention provides at least one of optimization information including an execution frequency of each program module constituting a program to be executed on a virtual machine and designation of whether or not to convert the program module. Selects the program module to be converted to the base, converts only the selected program module into a format that can be directly executed in the execution environment, distributes a program in which the converted program module and the non-converted program module are mixed, and does not convert A module provides a method for optimizing program execution, wherein the module is converted and executed at a distribution destination. In the program execution optimizing method according to the second aspect, a program module having a high execution frequency or a program module designated to be converted is converted into a format that can be directly executed in the execution environment. Program modules specified not to be converted are not converted. As described above, since the converted program module and the non-converted program module are mixed, the advantage that the execution time of the program can be shortened and the disadvantage that the transfer time becomes longer and the memory usage increases can be balanced. Can be optimized according to the execution environment.

[0009] In a third aspect, the present invention is a program management device for holding a program and distributing the program to the program execution device in response to a program request from a program execution device. Optimizing means for selecting a program module to be optimized based on at least one of the execution frequency of each program module and information on optimization of the program module, and optimizing only the selected program module; And a program distribution means for distributing a program in which program modules not to be converted are mixed to the program execution device. In the program management device according to the third aspect, the optimized program module and the non-optimized program module are mixed, so that the advantages and disadvantages of the optimization can be balanced, and the optimal program module can be optimized according to the program execution environment. Can be performed.

According to a fourth aspect, the present invention provides the program management device according to the third aspect, wherein the optimizing means comprises:
There is provided a program management device characterized by optimizing by at least one of compression and conversion from a format executed on a virtual machine to a format directly executable in an execution environment. In the program management device according to the fourth aspect, the program execution optimizing method according to the first or second aspect can be suitably implemented.

According to a fifth aspect, the present invention is a program execution device for sending a program request to the program management device according to the third or fourth aspect and executing a distributed program. A program execution device comprising: an execution frequency measurement unit for measuring an execution frequency of the program; and an execution frequency notification unit for notifying the program management device of the execution frequency measurement result. In the program management device according to the fifth aspect, the program execution optimizing method according to the first or second aspect can be suitably implemented.

[0012]

Embodiments of the present invention will be described below with reference to the drawings. Note that the present invention is not limited by this. FIG. 1 is an explanatory diagram showing a configuration and an operation procedure of a computer system that executes a program execution optimizing method of the present invention. This computer system (100)
Is a program management device for transferring a program (1)
And a program execution device (2) for executing the program
And a network (3) connecting them. The program management device (1) corresponds to a server,
The program execution device (2) corresponds to a client.

As an alternative to the network (3), a hardware that directly connects hardware such as a so-called bus may be used. That is, the program management device (1) and the program execution device (2) may operate on one computer, or may operate on a distributed processing computer having a plurality of CPUs.

[0014] The program management device (1) is a program storage means (11) for storing one or more programs.
And a profile storage means (14) for storing program characteristics (141) and client characteristics (142). The program characteristics (141) are profile data relating to programs stored in the program storage means (11), and exist in the same number as the types of programs stored in the program storage means (11). The client characteristics (142) are profile data relating to the connected program execution devices (2), and exist in the same number as the number of the program execution devices (2).

The computer system (100) operates as follows. First, the program execution device (2) sends a program request to the program management device (1) (procedure 1000). At this time, parameters including a client identifier, a program identifier, and an available memory amount are also transmitted.

The program management means (1) receives the program request and the parameters (step 1001).
Next, the program management means (1) acquires profile data on the program from the program characteristics (141) based on the program identifier, and obtains profile data on the program execution device (2) from the client characteristics (142) based on the client identifier. Is acquired (procedure 1002). Next, program management means (1)
Is a program storage unit (1) based on a program identifier.
Obtain the program from 1) and optimize the program based on the parameters and the profile data (procedure 1)
003). Next, the program management means (1) sends the optimized program to the program execution device (2) (procedure 1004).

The program execution device (2) receives the optimized program (procedure 1005). Next, the program execution device (2) executes the program (procedure 1
006). During this execution, the execution frequency data of each program module of the program is measured. Next, the program execution device (2) sends the execution frequency data together with the client identifier to the program management device (1) (procedure 1007).

The program management device (1) receives the execution frequency data (procedure 1008). Next, the program management device (1) updates profile data included in each of the program characteristics (141) and the client characteristics (142) based on the execution frequency data (procedure 10).
09).

FIG. 2 is a configuration diagram of the program management device (1). The program management device (1) includes the program storage unit (11), the profile storage unit (14), an optimization unit (10) that performs program optimization based on profile data, and a program execution device (2). Load request processing means (12) for receiving and processing a program request from a computer, and a program execution device (2)
Execution frequency counting means (13) for receiving the execution frequency data sent from the server and updating the profile data. The load request processing means (12) and the execution frequency totaling means (13) are connected to a network (3).

FIG. 3 is a block diagram of the program execution device (2). The program execution device (2) includes a program execution means (20) for executing a program, a program request means (21) for sending out a program request, and an execution frequency management means (22) for measuring a calling frequency of each program module of the program. ) And execution frequency transmitting means (23) for transmitting execution frequency data after the program is executed.
And execution frequency recording means (2) for recording execution frequency data.
4). The program request means (2
1) and the execution frequency transmission means (23) are connected to the network (3).

The execution frequency recording means (24) may not be necessarily provided, but is desirably provided to increase the degree of freedom in the timing of transmitting the execution frequency data to the program management device (1).

The program execution device (2) may include a recording device for caching the downloaded program. When this is present, the number of times of reading the program via the network is reduced, so that apparently high-speed execution is possible.

FIG. 4 is a detailed block diagram of the optimizing means (10) in the program management device (1). The optimizing unit (10) has an optimizing unit (100), a translator (101), and a compressing unit (102). The optimization execution unit (100) performs processing 1003 in FIG.
The program is optimized using the compression means (102) and the translator (101). The translator (1
01), a plurality of different types may exist for each type of CPU of the program execution device (2). When a program to be executed by the interpreter is not handled, the translator (101) may not be present. A plurality of compression means (102) having different compression ratios and compression algorithms may exist.

FIG. 5 is a configuration diagram of the program characteristic (141). The program characteristics (141) include a program identifier (1410), a module identifier (1411) for each program module, and its optimization characteristics (141).
2) and an execution frequency characteristic (143P). The program identifier (1410) is information indicating a program corresponding to the program characteristic (141). For example, it is a program name or a uniquely defined program ID.

The module identifier (1411) is information indicating a program module corresponding to this set. For example, a program module name, a start address, or a module ID. The optimization characteristics (14
12) optimization to be applied to the program module, such as whether to optimize the execution speed or to optimize the memory utilization efficiency when optimizing the program module Are recorded.
For example, for a program module that is executed very frequently, optimizing to increase the execution speed rather than to increase the memory utilization efficiency is effective in increasing the execution speed of the entire program. The fact that optimization should be performed with emphasis on execution speed is recorded. Conversely, program modules that are executed only when certain conditions occur (ie, program modules that may not be executed)
Since the execution speed of the program is not affected even if the execution speed is increased, it is recorded that optimization should be performed so as to increase the use efficiency of the memory. Optimization characteristics (141
The recording of 2) is performed at the time of developing the program, and when the program is stored in the program storage means (11), the registration in the program characteristic (141) is performed. The change of the optimization characteristic (1412) is possible even after registering the program.

FIG. 6 is a configuration diagram of the execution frequency characteristic (143P). The execution frequency characteristic (143P) has a set of a module identifier (1432) for each program module and an execution frequency (1443P) of the program module. The execution frequency (1443P) is created by performing statistical processing (for example, replacing the execution frequency data with an average value) on the execution frequency data received from the program execution device (2) when the program was executed in the past.

FIG. 7 is a configuration diagram of the client characteristic (142). The client characteristics (142) include a client identifier (1420) and execution environment information (1421).
And the execution frequency characteristics of each program executed in the past (14
3C). The client identifier (1420)
Is information indicating the program execution device (2) corresponding to the client characteristic (142). For example, a client name or a uniquely defined client ID. The execution environment information (1421) includes, for example, C
Information related to the execution of the program, such as the PU type, the memory capacity, the operating frequency, and whether the compressed file can be expanded by hardware, is recorded. This execution environment information (1
421), one is selected from the plurality of translators (101).

FIG. 8 is a configuration diagram of the execution frequency characteristic (143C). The execution frequency characteristic (143C) has a set of a program identifier (1431) of a corresponding program, a module identifier (1432) of each program module of the program, and an execution frequency (1443C) of the program module. The execution frequency (1433)
C) is the execution frequency characteristic (143) in the program execution device (2) corresponding to the client characteristic (142).
The execution frequency data received from the program execution device (2) when the program corresponding to C) was executed in the past is created by statistical processing (for example, replacement with an average value or replacement with the latest value).

FIG. 9 shows a program (30) executed by the program execution device (2) (1006 in FIG. 1).
FIG. This program (300)
It is composed of the above program modules (3000). Each program module (3000) includes module compression information (3010) and nativeization information (302).
0) and an execution code (3030). The module compression information (3010) records whether the program module is compressed or, if compressed, the compression method and the module size before compression. The native information (3020) includes information indicating whether the program module is to be interpreted and executed.
The format in which the PU is directly executed is recorded. In the case of a program that is not executed by the virtual machine, the native information (3020) is unnecessary.

FIG. 10 shows a program execution (100 in FIG. 1).
It is a flowchart which shows the processing procedure when a module call arises in 6). In step 11000, the execution frequency management means (22) is notified and the execution frequency of the called program module is counted up. In step 11010, the process refers to the module compression information (3010) of the program module to be called, and if it is compressed, the process proceeds to step 11020, and if not compressed, the process proceeds to step 11030. Step 11020
Then, the module size before compression is obtained from the module compression information (3010), a memory area capable of storing the module size is secured, and the program module is expanded in the memory area. And step 110
Move to 30.

In step 11030, it is determined whether or not the program to be executed is in a format to be executed in the virtual machine. If the program is to be executed in the virtual machine, the process proceeds to step 11050. Step 1
Move to 1040. In step 11040, the program module to be called is directly called, and the process ends.

In step 1105, the program refers to the native information (3020) of the program module to be called.
To step 110 if the CPU is to execute directly
Move to 70.

In step 11060, the program module to be called is called via the interpreter module, and the process ends.

In step 11070, the program module to be called is called via a module calling process in which the CPU directly executes the program module, and the process ends.

If there is not enough memory, the expanded program module is discarded after execution. If there is enough memory, the expanded program module is retained and reused (in this case, the compressed program module may be discarded).

When the program execution (1006 in FIG. 1) is completed, the execution frequency of each program module measured by the execution frequency management means (22) is transmitted to the execution frequency sending means (2).
3) is sent to the program management device (1) as execution frequency data (1007 in FIG. 1).

Execution frequency data transmission (1007 in FIG. 1)
Is performed at one of the following timings. (A) Transmit execution frequency data immediately after program execution. (B) When the program execution device (2) re-executes a program that has already been executed, the execution frequency data measured in the previous execution stored in the execution frequency recording means (24) is used as another program request. Sent together with parameters. (C) When the program management device (1) has room or at regular intervals, the execution frequency totaling means (13) of the program management device (1) is sent to the execution frequency transmission means (23) of the program management device (1). Sends instructions and sends execution frequency data in response. (D) When the program execution device (2) has room or periodically, the execution frequency data is transmitted under the initiative of the program execution device (2).

According to the above (a), there is an advantage that the execution frequency recording means (24) can be omitted. According to the above (b), since the execution frequency data is transmitted only when necessary,
There is an advantage that communication can be minimized. (C) and (d) above
According to the method, since the execution frequency data is not updated after the program request comes, there is an advantage that the time required from the program request to the program execution can be shortened. Further, since the processing can be performed when the CPU is idle, there is an advantage that the CPU time can be used effectively.

FIG. 11 is a flowchart of the processing procedure of the optimizing means (10). In step 12000, if conversion is designated by the optimization characteristics (1412 in FIG. 5) of each program module, the program module is converted, and if compression is designated, the program module is compressed. For example, for a program module that can be dramatically accelerated by applying a translator, it is preferable that conversion is specified by its optimization characteristics. In addition, it is preferable that compression is specified by the optimization characteristic of a program module that hardly affects the execution efficiency even if it is compressed.

In step 12005, the amount of available memory of the program execution device (2) and the aforementioned
If the usable memory amount is larger than the converted or compressed program size at step 120, the process proceeds to step 120.
Then, the process proceeds to step 12050 if the size of the program to be executed is larger.

In step 12010, it is determined whether or not the program is of a format to be executed by the virtual machine. If the format is to be executed by the virtual machine, the process proceeds to step 12020. If the format is not to be executed by the virtual machine, the process is terminated. Step 1
In step 2020, the translator translates the most frequently executed program module among the program modules that have not been converted. However, if the conversion is prohibited by the optimization characteristic (1412 in FIG. 5) of the program module, the conversion is not performed. For example, a program module, for which almost no speedup can be expected even when a translator is applied, may be prohibited from conversion by its optimization characteristics. In step 12030, the available memory amount is compared with the converted program size. If the available memory amount is larger, the process returns to step 12020. If the converted program size is larger, the process proceeds to step 12040. . In step 120401,
Returns the previously converted program module to the format before conversion. Then, the process ends.

In step 12050, the least frequently executed program module among the uncompressed program modules is compressed. However, if compression is prohibited by the optimization characteristic (1412 in FIG. 5) of the program module, compression is not performed. For example, the main routine of a program is generally executed only once and executed less frequently, but since it is executed first, if this program module is compressed, it takes time to start execution of the program due to expansion. Therefore, such a program module may be prohibited from being compressed by its optimization characteristics. Note that the less frequently used ones may be compressed by a compression algorithm with a higher compression ratio (that is, an algorithm that takes longer to expand). In step 12060, the available memory amount is compared with the program size after compression. If the available memory amount is larger, the process ends. If the compressed program size is larger, the process returns to step 12050.

Note that the above steps 12020 and 1220
In 050, since the bias of the execution frequency for each program execution device (2) can be reflected and more efficient optimization can be performed, the execution is performed using the execution frequency characteristic (143C) in the client characteristic (142). Determine the order of frequency. If the program execution device (2) has not executed the program in the past and does not have the execution frequency characteristic (143C), the execution is performed using the execution frequency characteristic (143P) in the program characteristic (141). Determine the order of frequency.

According to the optimization procedure shown in FIG. 11, since the program modules which are used less frequently are compressed in accordance with the available memory amount, the program can be loaded and executed with a small available memory amount. Frequently executed program modules are not compressed, so that the execution speed is not significantly reduced. Also, when a program is executed on a virtual machine, the translator converts the most frequently executed program modules into a form that can be immediately executed according to the amount of available memory. And the execution speed can be increased. Further, by utilizing the optimization characteristics, it is possible to perform a mixture of conversion and compression, such as converting a frequently executed program module by a translator and compressing a less frequently executed program module.

In another embodiment, the client characteristics (142) are managed for each user instead of for each program execution device (2). In this case, it is sufficient to have the client characteristics (142) by the number of users and set the user identifier as the client identifier (1420).

[0046]

According to the program execution optimizing method, program management device and program execution device of the present invention,
Since program modules optimized according to the execution environment of the program and program modules not optimized are mixed, the advantages and disadvantages of the optimization can be balanced. That is, the optimization according to the execution environment can be performed in consideration of the balance between the transfer time of the program, the execution time, the memory usage, and the like.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing a configuration and an operation procedure of a computer system that executes a program execution optimizing method of the present invention.

FIG. 2 is a configuration diagram of a program management device in FIG. 1;

FIG. 3 is a configuration diagram of a program execution device in FIG. 1;

FIG. 4 is a configuration diagram of an optimization unit in FIG. 2;

FIG. 5 is a configuration diagram of a program characteristic in FIG. 1;

FIG. 6 is a configuration diagram of an execution frequency characteristic in FIG. 5;

FIG. 7 is a configuration diagram of client characteristics in FIG. 1;

FIG. 8 is a configuration diagram of an execution frequency characteristic in FIG. 7;

FIG. 9 is a configuration diagram of a program module to be executed.

FIG. 10 is a flowchart of a module calling process.

FIG. 11 is a flowchart showing the operation of the optimization means of FIG. 4;

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 program pipeline device 2 program execution device 10 optimization means 22 execution frequency management means 101 translator 102 compression means 141 program characteristics 142 client characteristics 143C execution frequency characteristics 143P execution frequency characteristics

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Yasuko Yokoyama 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture Inside Hitachi, Ltd.System Development Laboratory (72) Inventor Kenji Kitagawa 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture Stock Company (72) Inventor Ritsuko Kawabata Ritsuko Kawabata 1099 Ozenji Temple, Aso-ku, Kawasaki City, Kanagawa Prefecture F-term (reference) 5B042 GA02 GA12 HH20 MA08 MC12 MC25 MC28 5B076 AB10 BA05 BB06 5B081 AA09 CC00 CC21 5B089 GA11 GA21 GB02 HA10 JA35 JB07 KA05 KB09 KB10 KC37 KC60 KH28 5J064 AA02 BA01 BB12 BD02 BD03

Claims (5)

[Claims] A program module for compressing a program module based on at least one of an execution frequency of each program module constituting a program executed on an electronic computer and optimization information including designation of whether or not the program module should be compressed. A program characterized in that only a selected program module is compressed, a program in which a compressed program module and a non-compressed program module are mixed is distributed, and the compressed program module is expanded and executed in a distribution destination. Execution optimization method. 2. A program module to be converted is selected based on at least one of an execution frequency of each program module constituting a program to be executed on the virtual machine and optimization information including designation of whether or not to convert the program module. Then, only the selected program module is converted into a format that can be directly executed in the execution environment, a program in which the converted program module and the non-converted program module are mixed is distributed, and the non-converted program module is converted at the distribution destination. A program execution optimizing method characterized by executing. 3. A program management device for holding a program and distributing the program to the program execution device in response to a program request from the program execution device, the program management device comprising: A program module to be optimized based on at least one of the information related to the optimization of the program module is selected, and an optimization means for optimizing only the selected program module is mixed with an optimized program module and a non-optimized program module. And a program distribution means for distributing the program to the program execution device. 4. The program management device according to claim 3, wherein said optimizing means optimizes by at least one of compression and conversion from a format executed on a virtual machine to a format directly executable in an execution environment. A program management device, characterized in that: 5. A program execution device that sends a program request to the program management device according to claim 3 or 4 and executes a distributed program.
A program execution device comprising: an execution frequency measurement unit that measures an execution frequency of each program module; and an execution frequency notification unit that notifies the program management device of a measurement result of the execution frequency.