JP4905974B2 - Program profiling apparatus, program profiling method, and program - Google Patents

Description

  The present invention relates to a program profiling apparatus, a program profiling method, and a program for measuring a processing time of each method and analyzing a call relationship between methods when a program composed of a plurality of methods is executed.

For the purpose of improving the processing performance of a program composed of a plurality of methods, it is necessary to identify a method that greatly affects the performance degradation of the entire program (becomes a bottleneck). Therefore, processing such as measuring the processing time of the method is performed.
Actually, the processing time may vary depending on the route by which the method is called. Therefore, when measuring the processing time of the method in order to identify the bottleneck, it is necessary to analyze the calling relationship between the methods. Then, the processing time of the method is measured based on the analyzed call relationship. Such a series of processes is called profiling.

  For example, Patent Document 1 discloses a technique of processing a source program or an execution program, inserting probes before and after a method to be measured, and performing profiling by processing of the probes.

  However, in the technique disclosed in Patent Document 1, all the probes inserted in the method executed when the program is executed are executed, and as a result, even probes with a low need for measurement are executed. It will be. Therefore, the more detailed measurement is performed, the greater the load on the execution environment due to the execution of the probe, and the accuracy of the measurement result decreases.

Therefore, the following profiling methods have been proposed.
(I) First, a first measurement is performed to measure statistical data with almost all methods as objects of measurement. (Ii) Next, based on this measurement result, narrowing down methods for performing more detailed measurement is performed. (Iii) Then, only for this narrowed-down method, each time the method is executed, a more detailed measurement is performed such as measuring the processing time including the execution time of a method having a series of calling relationships. A second measurement process is performed.

In the profiling method described above, since methods for performing detailed measurement are narrowed down, profiling with a small load on the execution environment can be realized.
JP 2004-94374 A

  However, in the above-described profiling method, it is necessary to perform the first measurement and the second measurement separately. For each measurement, the measurement target system is set, the measurement scenario is executed, and the measurement result is analyzed. Procedures will occur and are inefficient.

  The present invention has been made in view of the above circumstances, and an object thereof is to provide a profiling apparatus and the like capable of efficient profiling.

In order to achieve the above object, a profiling apparatus according to a first aspect of the present invention includes:
Statistical data measuring means for measuring the processing time of the executed method when a program composed of a plurality of methods is executed;
Processing time determination means for determining whether the processing time of the method measured by the statistical data measurement means is equal to or greater than a predetermined threshold;
Main method setting means for setting a method determined by the processing time determination means to have a processing time equal to or greater than a predetermined threshold as a main method that is a main factor of execution time when the program is executed;
Detailed data measuring means for acquiring processing information of the main method and route information indicating what method the main method has been followed by when the main method is executed;
Method detailed information storage means for storing the processing time and route information of the main method acquired by the detailed data measuring means for each main method;
Main method setting release means for releasing the setting of the main method set by the main method setting means for the main methods stored in the method detailed information storage means exceeding a predetermined number;
With
The main method setting means sets a main method during execution of the program,
The main method setting release means cancels the setting of the main method during the execution of the program,
During one execution of the program, the measurement by the statistical data measurement means and the measurement by the detailed data measurement means are performed.
It is characterized by that.

The statistical data measuring means may accumulate the processing time of the executed method for each method, and the accumulated processing time may be the processing time of the method.

The threshold value may be a value based on the maximum value of the processing time measured by the statistical data measuring means within a predetermined time.

A waiting time obtaining means for obtaining a waiting time during which the main method is waiting for processing;
The detailed data measuring means may store the waiting time obtained by the waiting time obtaining means for each main method.

In order to achieve the above object, a profiling method according to a second aspect of the present invention includes:
Statistical data measurement step that measures the processing time of the executed method when a program composed of multiple methods is executed,
A processing time determination step of determining whether the processing time of the method measured by the statistical data measurement step is equal to or greater than a predetermined threshold;
A main method setting step for setting a method determined by the processing time determination step to have a processing time equal to or greater than a predetermined threshold as a main method that is a main factor of execution time when the program is executed;
A detailed data measurement step of obtaining a processing time of the main method and route information indicating what kind of method the main method has been called by when the main method is executed;
A method detailed information storage step for storing the processing time and route information of the main method acquired by the detailed data measuring step for each main method;
A main method setting release step for releasing the setting of the main method set in the main method setting step for the main method stored in the method detailed information storage step exceeding a predetermined number;
With
Wherein in the main method setting step sets the key methods during execution of the program,
In the main method setting release step, the setting of the main method is canceled during the execution of the program,
During one execution of the program, the measurement by the statistical data measurement step and the measurement by the detailed data measurement step are performed.
It is characterized by that.

In order to achieve the above object, a program according to the third aspect of the present invention provides:
Computer
When executing a program composed of a plurality of methods, statistical data measuring means to measure the processing time of the method was executed,
Processing time determination hand stage processing time method measured by the statistical data measuring means for determining whether more than a predetermined threshold value,
The processing time by the processing time determination means methods that is determined to be equal to or larger than a predetermined threshold value, the main method setting means to set the main method that is the main factor of the execution time for executing the program,
Wherein when the primary method is performed, and acquires the route information indicating whether the has been invoked following a what methods the processing time of said principal methods and said principal methods, detailed data measurement hand stage,
Method detailed information storage means for storing the processing time and route information of the main method acquired by the detailed data measuring means for each main method,
Main method setting release means for releasing the setting of the main method set by the main method setting means for the main method stored in the method detailed information storage means exceeding a predetermined number;
A program that functions as
The main method setting means sets a main method during execution of the program,
The main method setting release means cancels the setting of the main method during the execution of the program,
During one execution of the program, the measurement by the statistical data measurement means and the measurement by the detailed data measurement means are performed.
It is characterized by that.

  According to the present invention, statistical information and detailed information can be acquired simultaneously by executing a program once, so that efficient profiling is possible.

Hereinafter, a program profiling apparatus according to an embodiment of the present invention will be described with reference to the drawings.
As shown in FIG. 1, the program profiling apparatus 1 includes a communication unit 11, an input unit 12, an output unit 13, a storage unit 14, an input / output I / F unit 15, and a control unit 16. Each unit is connected to each other via a bus 17.

The communication unit 11 communicates with an information processing apparatus and the like on the network via the communication network 2 and includes a communication interface and the like. For example, when the communication unit 11 receives an execution program to be profiled, the program profiling apparatus 1 captures the profiled target.
It is assumed that the execution program to be profiled in this embodiment is composed of a plurality of classes, and each class has one or more methods.

  The input unit 12 is used to input various information to the program profiling apparatus 1 and includes input devices such as buttons and keys. For example, the input unit 12 receives analysis instruction start instruction information, which will be described later.

  The output unit 13 outputs various information and includes a display device such as a display. For example, the output unit 13 displays a list screen of processing times for each method measured in an analysis process described later.

The storage unit 14 stores various information, programs, and the like, and includes an auxiliary storage device such as a hard disk. For example, the storage unit 14 stores in advance an execution program or the like captured via the communication unit 11 or the like before performing the analysis process.
Further, as shown in FIG. 2, the storage unit includes a probe insertion target storage table 141, a method statistical information storage table 142, a call graph information storage table 143, a method information storage table 144, and a method detailed performance log table 145. And memorize.

  As shown in FIG. 4, the probe insertion target storage table 141 stores information indicating whether or not a probe is to be inserted into each class that constitutes an execution program to be profiled.

  As shown in FIG. 5, the method statistical information storage table 142 stores the total execution time, the total waiting time, and the number of executions for each method when the program is executed. The total waiting time is the accumulated waiting time (waiting time) from the start to the end of the program until the processing of the method called from the method is completed. It was time.

  As shown in FIG. 6 (A), the call graph information storage table 143 has a call relationship between processing (methods) (caller method name and callee method name) and the number of times the call relationship is called (call count). ) Is memorized. When information as shown in FIG. 6 (A) is stored in the call graph information storage table 143, a graph (call graph) representing an inter-call system as shown in FIG. 6 (B) from this stored information. Can be created.

As shown in FIG. 7, the method information storage table 144 includes, for each method included in the execution program, a method ID, a tracing ID, a method name, identification information (class name) of a class to which the method belongs, Measurement availability information is stored.
The method ID is method identification information issued in a probe insertion process to be described later.
The trace ID is identification information of a method used when acquiring a stack trace, which is acquired in probe insertion processing described later. For example, when the execution program is JAVA (registered trademark), the tracing ID corresponds to an ID issued by the JAVA virtual machine profiler interface.
The measurement availability information is information indicating whether or not the method is a target of a profile performed by the measurement unit.

The method detailed performance log table 145 is a table for storing the results of profiling. As shown in FIG. 8, the method detailed performance log table 145 includes, for each process of an executed method, identification information (name in the figure) of a thread to which the process to be executed belongs, a method ID, an execution time, and a wait Remember time and stack trace.
The stack trace is information indicating what method the process has been traced and called, and sequentially stores the trace ID of the traced method. For example, when information “11111, 22222, 33333” is recorded in the stack trace, the process having the stack trace traces each method having the trace IDs “11111”, “22222”, “33333” in order. Means that it has been called. Information such as the method name corresponding to the trace ID may be acquired with reference to the method information storage table 144 described above.

  Returning to FIG. 1, the input / output I / F unit 15 is an interface for connecting the input / output device 3. For example, an execution program or the like to be profiled can be fetched from the input / output device 3 such as a CD-ROM drive via the input / output I / F unit 15 instead of the communication unit 11.

  The control unit 16 performs calculation processing of data and controls the communication unit 11, the input unit 12, the output unit 13, the storage unit 14, and the input / output I / F unit 15 via the bus 17. A central processing unit (161), a read only memory (ROM) 162, a random access memory (RAM) 163, and the like. Specifically, the arithmetic processing and control processing in the control unit 16 are performed by the CPU 161 executing the control program stored in the ROM 162 while temporarily storing various data using the RAM 163 as a work area. Done.

  For example, the control unit 16 controls the above units according to the control program stored in the ROM 162 or the storage unit 14 so that the profiling process or the like is performed in the program profiling apparatus 1.

FIG. 3 is a block diagram illustrating a functional configuration example of the program profiling apparatus according to the present embodiment.
As shown in the figure, the program profiling apparatus 1 functionally includes a probe insertion unit 21, a program execution unit 22, a measurement unit 23, a detailed measurement target setting unit 24, a measurement method selection unit 25, and a measurement target. A setting unit 26. Note that the control unit 16 illustrated in FIG. 1 includes the communication unit 11, the input unit 12, the output unit 13, the storage unit 14, or the input / output I / F unit 15 illustrated in FIG. It can be realized by controlling.

  The probe insertion unit 21 inserts a probe for measurement into a method of a class that is set as an insertion target in the probe insertion target storage table 141 (set as acceptable), and uses the method as a method. Register in the information storage table 144.

  The program execution unit 22 executes the execution program in which the probe is inserted.

The measurement unit 23 measures information related to processing (method) of the execution program executed by the program execution unit 22. The measurement unit 23 includes a statistical data measurement unit 231 and a detailed data measurement unit 232.
The statistical data measurement unit 231 performs measurement using almost all methods included in the program as measurement targets.
The detailed data measurement unit 232 performs more detailed measurement than the measurement performed by the statistical data measurement unit 231 with a method that relatively affects the execution time when the program is executed as a measurement target.

The detailed measurement target setting unit 24 sets a method to be measured by the detailed data measurement unit 232.
Specifically, the detailed measurement target setting unit 24 activates the detailed monitoring flag embedded in the method when the processing time of the method stored in the method statistical information storage table 142 is equal to or greater than a predetermined threshold. The method is set as a measurement target by the detailed data measurement unit 232. Further, the statistical data measurement unit 231 refers to the method detailed performance log table 145 to determine a method that has been measured by the detailed data measurement unit 232 a predetermined number of times and is embedded in the determined method. The detailed monitoring flag is invalidated and the method is excluded from the measurement target by the detailed data measuring unit 232.

  The measurement method selection unit 25 performs a process of switching whether the above-described measurement process is performed by the statistical data measurement unit 231 or the detailed data measurement unit 232 of the measurement unit.

  The measurement target setting unit 26 updates the measurement availability flag in the method information storage table 144 and sets a method to be measured.

  Next, an outline of processing for profiling a program (execution program) in order to find a bottleneck when a program is executed by the program profiling apparatus 1 will be described.

  As a premise, an execution program to be profiled is stored in the storage unit 14 in advance, and information indicating which class is to be a probe insertion target among the classes constituting the execution program is also included. It is assumed that the probe insertion target storage table 141 is stored in advance by a user's preparation work or the like.

  When a user inputs an instruction for measuring an execution program from the input unit 12 and the input information is transmitted to the control unit 16, the profiling process shown in FIG. 9 is started.

  First, the probe insertion unit 21 inserts a method measurement probe into the execution program (step S101).

  Next, the measurement target setting unit 26 performs settings for setting all methods included in the program to be measured by the statistical data measurement unit 231 (step S102).

Next, processing for measuring the processing time of a method processed at the time of program execution is performed (step S103).
This completes the profiling process.

  Next, the profiling process outlined above will be described in more detail with reference to FIGS.

(1) Probe insertion processing in step S101 in FIG. 9:
First, the probe insertion unit 21 reads one class constituting an execution program to be measured into the RAM 163 (FIG. 10, step S201).

  Subsequently, the probe insertion unit 21 determines whether the class read in step S201 is a probe insertion target based on information stored in the probe insertion target storage table 141 (step S202).

When it is determined that the class is not the probe insertion target class (step S202; No), the probe insertion unit 21 moves the process to step S207.
When determining that the class is a probe insertion target class (step S202; Yes), the probe insertion unit 21 issues identification information (method ID) to a method included in the class (step S203).

Then, the probe insertion unit 21 inserts the measurement probe before and after the method that issued the method ID (the portion before and after the execution of the method in the execution stage) (step S204).
In the following description, it is assumed that the probe inserted before the processing of the method is an execution start (for detection) probe, and the probe inserted after the method is an execution end (for detection) probe.

FIG. 11 is a diagram for explaining a method and a process of inserting a probe inserted into the method.
Note that the program described in FIG. 11 represents an execution program that is binary data in the JAVA (registered trademark) language format for easy understanding.
The execution program shown in FIG. 11A includes a class “A”, and the class “A” includes a method “sayHello” having no argument.
An example in which a measurement probe is inserted into this objective program is shown in FIG. In the class “A”, a monitoring flag “flag_1” indicating whether or not the probe processing is executed at the beginning of the processing, and a detailed monitoring flag indicating whether or not the detailed data measurement unit 232 executes detailed measurement “Dflag_1” is inserted. Further, the execution start probe “enterMethod” is inserted at the start time of the process of the method “sayHello”, and the execution end probe “exitMethod” is inserted at the end time of the process.
The execution start probe “enterMethod” and the execution end probe “exitMethod” have a detailed monitoring flag “dflag — 1” as an argument. When the detailed supervision flag is invalid, the probe performs only measurement by the statistical data measurement unit 231. When the detailed monitoring flag is valid, the probe performs measurement by the detailed data measurement unit 232 in addition to measurement by the statistical data measurement unit 231.

Subsequently, the probe insertion unit 21 acquires the tracing ID of the method included in the class (step S205).
Specifically, class information is acquired by an API (Application Program Interface) function provided by a virtual machine or the like for executing an execution program, and a trace ID is acquired from the class information.

  Subsequently, the probe insertion unit 21 associates, in the method information storage table 144, the method ID, the class name, the method name of the method in which the probe is inserted in step S204, and the trace ID acquired in step S205 with one method. It registers as an entry (FIG. 10, step S206).

Subsequently, in step S207, the probe insertion unit 21 determines whether there is another class that has not been read.
When it is determined that there is another class that has not been read (step S207; Yes), the probe insertion unit 21 repeats the processes of steps S201 to S206.
When it is determined that there is no other class that has not been read (step S207; No), the probe insertion process ends.

(2) Measurement object setting process in step S102 in FIG. 9:
The measurement target setting unit 26 sets the measurement enable / disable flag of all the entries in the method information storage table 144 registered by the probe insertion process to “Yes” indicating that the measurement unit 23 is a measurement target (see FIG. 12, Step S301). This completes the measurement target setting process.

(3) Measurement process in step S103 in FIG. 9:
When the program execution unit 22 executes the execution program with the probe inserted, the method in the statistical mode is measured by the processing of the probe embedded in the method of the execution program.

Here, the operation of the method in which the probe is inserted, which is performed by executing this execution program, will be described with reference to FIGS.
First, when the execution of a method is started, the program execution unit 22 determines whether or not the method is a measurement target (FIG. 13, step S401). Specifically, the program execution unit 22 makes this determination based on the measurement availability information in the method information storage table 144.

  When it is determined that the method is not a measurement target (step S401; No), the program execution unit 22 moves the process to step S408.

  When it is determined that the method is a measurement target (step S401; Yes), the statistical data measurement unit 231 indicates that the processing of the method is started together with the time at that time in FIG. As indicated by L1, it is recorded in the RAM 163 (step S402).

  Subsequently, the measurement method selection unit 25 determines whether or not the method is a method called from a method to be subjected to detailed monitoring (step S403).

Note that the determination in step S403 may be performed as follows, for example.
First, a counter Cnt (initial value 0) is prepared for each thread. The counter Cnt is set to 1 when a method set as a measurement target by the detailed data measuring unit 232 is executed, and is set to be incremented by 1 every time the method is executed thereafter. To do. Then, when the execution of the method is started, the counter Cnt is checked, and if the value of the counter Cnt is 1, it may be determined that the method is called from the method to be monitored in detail.
For example, it is assumed that there are methods A to D having a calling relationship as shown in FIG. That is, method A calls method B, method B calls method C, and method C calls method D. Note that the method B is a method set as a measurement target by the detailed data measurement unit 232.
Consider a process in which methods B to D are executed in order according to the calling relationship when method A is executed in such a state. In this case, the counter Cnt is set to 1 at the time when the method B, which is the target method for detailed monitoring, is executed. Subsequently, when the method C is executed, the counter Cnt is set to 2, and when the method D is executed, the counter Cnt is set to 3. That is, in this example, the counter Cnt is checked at the start of execution of the method, and it is possible to determine that the method C whose value is set to 1 is a method called from the method to be monitored in detail. .

  Returning to FIG. 13, when it is determined that the method is a method called from the method targeted for detail monitoring (step S403; Yes), the detail data measuring unit 232 selects the method targeted for detail monitoring of the call source. Is recorded in the RAM 163 as indicated by L3 in FIG. 15B together with the identification information (thread name) of the thread to which the process of the method belongs and the time at that time (step S404), The process moves to step S408.

  If it is determined that the method is not a method called from the method to be subject to detailed monitoring (step S403; No), the measurement method selection unit 25 determines whether the method is a measurement target by the detailed data measurement unit 232 or not. Is determined (step S405). Specifically, this determination is made based on the detailed monitoring flag embedded in the method.

  When it is determined that the method is not a measurement target by the detailed data measurement unit 232 (step S405; No), the measurement method selection unit 25 moves the process to step S408.

  If it is determined that the method is an object of measurement by the detailed data measurement unit 232 (step S405; Yes), the detailed data measurement unit 232 indicates that the process of the method has started, and the process to which the process of the method belongs. The identification information (thread name) and the time at that time are recorded in the RAM 163 as indicated by L5 in FIG. 15C (step S406).

Subsequently, the detailed data measurement unit 232 performs stack trace, acquires the method followed by the process of the method in the list format of the trace ID, and stores it in the RAM 163 (step S407). Then, the process proceeds to step S408.
The stack trace may be performed by an API function provided by a virtual machine or the like for executing the execution program. For example, in the case of JAVA (registered trademark), this process can be performed by using GetCallTrace (), which is a method for performing a stack trace provided by the JAVA virtual machine profiler interface (JVMPI). .

  In step S408, the program execution unit 22 executes processing of a method other than the probe inserted in the method (that is, processing of the method before the probe is inserted).

  Subsequently, when the process of step S408 ends, the program execution unit 22 determines whether or not the method is a measurement target (FIG. 14, step S409). This process is substantially equivalent to step S401.

  When it is determined that the method is not a measurement target (step S409; No), the program execution unit 22 ends the execution of the method.

  If it is determined that the method is a measurement target (step S409; Yes), the statistical data measurement unit 231 indicates that the processing of the method has been completed in L2 of FIG. As shown, it is recorded in the RAM 163 (step S410).

Then, the statistical mode measurement unit 231 calculates the processing time of the method from the information recorded in the RAM 163 by the processing of Step S402 and Step S410. Further, the statistical mode measurement unit 231 accumulates the calculated processing time in the total execution time of the method in the method statistical information storage table 142 and adds 1 to the number of executions.
For example, when information as shown in FIG. 15A is stored in the RAM 163, the processing time of the method stored in the entry having the method ID “01” in the method statistical information storage table 142 is 4 seconds. And the number of executions is incremented by one.

Subsequently, the statistical mode measurement unit 231 calculates the waiting time of the method from the information recorded in the RAM 163 by the processing of Steps S402 and S410, and the calculated waiting time of the method in the method statistical information storage table 142 is calculated. The total waiting time is accumulated (step S412).
For example, when information as shown in FIG. 15D is stored in the RAM 163 by the processing of step S402 and step S410, the waiting time of the method 01 is the processing time of the method 02 called by the method 01. It can be calculated as 3 seconds, and the total waiting time of the method 01 is increased by 3 seconds.
Further, the processing for obtaining the total waiting time may be obtained by using an API provided by a virtual machine or the like for executing the execution program.

Subsequently, when the method is called from another method and executed, the statistical mode measurement unit 231 stores the correspondence relationship between the call source and the call destination in the call graph information storage table 143 (Step S1). S413).
For example, when information as shown in FIG. 15D is stored in the RAM 163 by the processing of step S402 and step S410, the call source is the method “01”, the call destination is the method “02”, and the call Entry information with the number of times “1” is registered in the call graph information storage table 143. If there is a correspondence between the call destination and the call source that is already stored in the call graph information storage table 143, a process of adding 1 to the number of calls of the correspondence is performed.
Further, the correspondence relationship of the call destinations may be obtained by using an API provided by a virtual machine or the like for executing the execution program.

  Subsequently, the measurement method selection unit 25 determines whether or not the method is a method called from a method to be measured by the detailed data measurement unit 232 (step S414). This determination process is practically equivalent to step S403.

  If it is determined that the method is a method called from the method targeted for detail monitoring (step S414; Yes), the detail data measuring unit 232 cancels the waiting state of the method targeted for detail monitoring of the call source. This is recorded in the RAM 163 as indicated by L4 in FIG. 15B, together with the identification information (thread name) of the thread to which the process of the method belongs and the time at that time (step S415).

Then, from the information stored in the RAM 163 in step S404 and step S415, the waiting time of the method to be monitored in detail, which is the method call source, is obtained and stored in the detailed performance log table (step S416). Then, the process proceeds to step S421.
For example, when information such as that shown in FIG. 15B is stored in the RAM 163 by the processing of step S404 and step S415, the detailed data measurement unit 232 stores the thread “# 1” in the method detailed performance log table 145, The waiting time of the entry having the method 01 is stored as “10 seconds”.

  If it is determined that the method is not a method called from the method to be monitored in detail (Step S414; No), the measurement method selection unit 25 determines whether the method is a measurement target by the detailed data measurement unit 232 or not. Is determined (step S417). This determination process is practically equivalent to step S405.

  If it is determined that the method is not a measurement target by the detailed data measurement unit 232 (step S417; No), the measurement method selection unit 25 moves the process to step S421.

  When it is determined that the method is an object to be measured by the detailed data measurement unit 232 (step S417; Yes), the detailed data measurement unit 232 indicates that the process of the method has been completed and the process to which the process of the method belongs. The identification information (thread name) and the time at that time are recorded in the RAM 163 as indicated by L6 in FIG. 15C (step S418).

Then, the detailed data measuring unit 232 creates and registers entry information for registration in the method detailed performance log table 145 from the information recorded in the RAM 163 by the processing in steps S406 and S418 (step S419).
For example, when the information as shown in FIG. 15C is stored in the RAM 163 by the processing of step S406 and step S418, the detailed data measurement unit 232 includes the thread “# 1”, the method “01”, the execution time The entry of “4 seconds” is stored in the method detailed performance log table 145.

  Next, the detailed data measurement unit 232 registers the list of trace IDs stored in the RAM 163 in step S407 in the stack trace of the entry registered in step S419 (step S420), and moves the process to step S422.

In step S421, based on the information stored in the method statistical information storage table 142, the detailed measurement target setting unit 24 determines whether the total processing time of the method is a predetermined threshold (for example, 60 seconds) or more. Is determined.
This threshold value is a value that changes dynamically, such as a value based on the maximum value of the total processing time stored in the method statistical information storage table 142 within a predetermined time (for example, 80% of the maximum value). There may be.

  If it is determined that the total processing time of the method is not equal to or greater than the predetermined threshold (step S421; No), the detailed measurement target setting unit 24 ends the processing of the method.

  If it is determined that the total processing time of the method is equal to or greater than the predetermined threshold (step S421; Yes), the detailed measurement target setting unit 24 activates the detailed monitoring flag embedded in the method and sets the method as detailed data. It sets so that it may become a measuring object by the measurement part 232 (step S423), and the process of the said method is complete | finished.

  In step S422, the detailed measurement target setting unit 24 determines whether or not a predetermined number (for example, 20) or more entries of the method detailed performance log table 145 by the process of the method are registered.

  If it is determined that the predetermined number or more are not registered (step S422; No), the detailed measurement target setting unit 24 ends the process of the method.

  If it is determined that a predetermined number or more are registered (step S422; Yes), the detailed measurement target setting unit 24 invalidates the detailed monitoring flag embedded in the method, and the detailed data measurement unit 232 measures the method. (Step S424), and the process of the method is terminated.

  Thus, the process when the method in which the probe is inserted is executed is completed. And the above-mentioned process is performed with respect to the method in which all the probes executed when the execution program is executed are inserted. Then, when the execution of the program is completed, the measurement process ends, and the profiling process shown in FIG. 9 ends.

As described above, according to this embodiment, measurement is performed by the statistical data measurement unit 231 with almost all methods executed when the program is executed as a measurement target, and the measurement is obtained by the measurement. Based on the total execution time of the method, a method to be monitored by the detailed data measuring unit 232 is set as needed. The detailed data measuring unit 232 performs measurement for the method for which this setting has been made.
Therefore, the number of program executions is one, statistical measurement with almost all methods as the measurement target, and the call relationship every time a method is executed with the method that may be a bottleneck as the measurement target. At the same time, it is possible to carry out detailed measurement for measuring the processing time, and it is possible to realize more efficient profiling with reduced load on the execution environment.

  In addition, this invention is not limited to the said embodiment, A various application is possible.

For example, although the detail monitoring flag is invalidated for a method in which a predetermined number or more entries are registered in the method detailed performance log table 145 (FIG. 14, steps S422 and S424), this process is not necessarily performed. .
In the above embodiment, whether or not to enable the detailed monitoring flag is determined based on whether or not the total execution time of the executed method is equal to or greater than a predetermined threshold. However, this determination may be made based on whether or not the execution time of the method (that is, the time not including the execution time when the method has been executed in the past) is equal to or greater than a predetermined threshold. In this case, since it is not necessary to perform processing for accumulating and storing the execution time for each method in order to obtain the total execution time, profiling can be performed at a higher speed.

Note that the program profiling apparatus of the present invention is not limited to dedicated hardware, and can also be realized by a normal computer system.
Specifically, in the above embodiment, the program profiling apparatus has been described as being prestored in a memory or the like. However, a program for executing the above-described processing operation is recorded on a computer-readable recording medium such as a flexible disk, a CD-ROM (Compact Disk Read-Only Memory), a DVD (Digital Versatile Disk), or an MO (Magneto-Optical disk). A program profiling apparatus that executes the above-described processing may be configured by storing and distributing the program in a medium and installing the program in a computer.

Further, the program may be stored in a disk device or the like included in a server device on a communication network such as the Internet, and may be downloaded to a computer by being superimposed on a carrier wave, for example. Furthermore, the above-described processing can also be achieved by starting and executing a program while transferring it via a communication network.
When the above functions are realized by an OS (Operating System) sharing or by cooperation between the OS and an application, only the part other than the OS may be stored in a medium and distributed. You may download it.

It is a block diagram which shows the structural example of the program profiling apparatus which concerns on embodiment of this invention. It is a figure which shows the structure of the memory | storage part of the program profiling apparatus shown in FIG. It is a block diagram which shows the function structural example of the program profiling apparatus which concerns on embodiment of this invention. It is a figure which shows the structure of a probe insertion object storage table. It is a figure which shows the structure of a method statistics information storage table. (A) is a figure which shows the structure of a call graph information storage table. (B) is a figure showing an example of a call graph. It is a figure which shows the structure of a method information storage table. It is a figure which shows the structure of a method detailed performance log table. It is a flowchart for demonstrating operation | movement of a profiling process. It is a flowchart for demonstrating operation | movement of a probe insertion process. It is a figure for demonstrating insertion of a probe. It is a flowchart for demonstrating operation | movement of a measurement object setting process. It is a flowchart for demonstrating operation | movement of the method in which the probe is inserted. It is a flowchart for demonstrating operation | movement of the method in which the probe is inserted. It is the figure which showed the example of the information memorize | stored in RAM by a measurement process. It is a figure for demonstrating the process which identifies the method called from the method made into the detail monitoring object.

Explanation of symbols

1 Program Profiling Device 141 Probe Insertion Target Storage Table 142 Method Statistical Information Storage Table 143 Call Graph Information Storage Table 144 Method Information Storage Table 145 Method Detailed Performance Log Table 21 Probe Insertion Unit 22 Program Execution Unit 23 Measurement Unit 231 Statistical Data Measurement Unit 232 Detailed data measurement unit 24 Detailed measurement target setting unit 25 Measurement method selection unit 26 Measurement target setting unit

Claims (6)

Statistical data measuring means for measuring the processing time of the executed method when a program composed of a plurality of methods is executed;
Processing time determination means for determining whether the processing time of the method measured by the statistical data measurement means is equal to or greater than a predetermined threshold;
Main method setting means for setting a method determined by the processing time determination means to have a processing time equal to or greater than a predetermined threshold as a main method that is a main factor of execution time when the program is executed;
Detailed data measuring means for acquiring processing information of the main method and route information indicating what method the main method has been followed by when the main method is executed;
Method detailed information storage means for storing the processing time and route information of the main method acquired by the detailed data measuring means for each main method;
Main method setting release means for releasing the setting of the main method set by the main method setting means for the main methods stored in the method detailed information storage means exceeding a predetermined number;
With
The main method setting means sets a main method during execution of the program,
The main method setting release means cancels the setting of the main method during the execution of the program,
During one execution of the program, the measurement by the statistical data measurement means and the measurement by the detailed data measurement means are performed.
A program profiling apparatus characterized by the above. The statistical data measuring means accumulates the processing time of the executed method for each method, and sets the accumulated processing time as the processing time of the method.
The program profiling apparatus according to claim 1, wherein: The threshold is a value based on the maximum value of the processing time measured by the statistical data measuring means within a predetermined time.
The program profiling apparatus according to claim 1, wherein the apparatus is a program profiling apparatus. A waiting time obtaining means for obtaining a waiting time during which the main method is waiting for processing;
The detailed data measuring means stores the waiting time obtained by the waiting time obtaining means for each main method.
The program profiling apparatus according to any one of claims 1 to 3 , wherein Statistical data measurement step that measures the processing time of the executed method when a program composed of multiple methods is executed,
A processing time determination step of determining whether the processing time of the method measured by the statistical data measurement step is equal to or greater than a predetermined threshold;
A main method setting step for setting a method determined by the processing time determination step to have a processing time equal to or greater than a predetermined threshold as a main method that is a main factor of execution time when the program is executed;
A detailed data measurement step of obtaining a processing time of the main method and route information indicating what kind of method the main method has been called by when the main method is executed;
A method detailed information storage step for storing the processing time and route information of the main method acquired by the detailed data measuring step for each main method;
A main method setting release step for releasing the setting of the main method set in the main method setting step for the main method stored in the method detailed information storage step exceeding a predetermined number;
With
Wherein in the main method setting step sets the key methods during execution of the program,
In the main method setting release step, the setting of the main method is canceled during the execution of the program,
During one execution of the program, the measurement by the statistical data measurement step and the measurement by the detailed data measurement step are performed.
A program profiling method characterized by the above. Computer
When executing a program composed of a plurality of methods, statistical data measuring means to measure the processing time of the method was executed,
Processing time determination hand stage processing time method measured by the statistical data measuring means for determining whether more than a predetermined threshold value,
The processing time by the processing time determination means methods that is determined to be equal to or larger than a predetermined threshold value, the main method setting means to set the main method that is the main factor of the execution time for executing the program,
Wherein when the primary method is performed, and acquires the route information indicating whether the has been invoked following a what methods the processing time of said principal methods and said principal methods, detailed data measurement hand stage,
Method detailed information storage means for storing the processing time and route information of the main method acquired by the detailed data measuring means for each main method,
Main method setting release means for releasing the setting of the main method set by the main method setting means for the main method stored in the method detailed information storage means exceeding a predetermined number;
A program that functions as
The main method setting means sets a main method during execution of the program,
The main method setting release means cancels the setting of the main method during the execution of the program,
During one execution of the program, the measurement by the statistical data measurement means and the measurement by the detailed data measurement means are performed.
A program characterized by that.

Images