WO2012056569A1 - Performance measurement method, performance measurement device, and performance measurement program - Google Patents

Performance measurement method, performance measurement device, and performance measurement program Download PDF

Info

Publication number
WO2012056569A1
WO2012056569A1 PCT/JP2010/069335 JP2010069335W WO2012056569A1 WO 2012056569 A1 WO2012056569 A1 WO 2012056569A1 JP 2010069335 W JP2010069335 W JP 2010069335W WO 2012056569 A1 WO2012056569 A1 WO 2012056569A1
Authority
WO
WIPO (PCT)
Prior art keywords
program
execution time
performance
execution
extracted
Prior art date
Application number
PCT/JP2010/069335
Other languages
French (fr)
Japanese (ja)
Inventor
真哉 南谷
Original Assignee
株式会社日立製作所
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 株式会社日立製作所 filed Critical 株式会社日立製作所
Priority to PCT/JP2010/069335 priority Critical patent/WO2012056569A1/en
Publication of WO2012056569A1 publication Critical patent/WO2012056569A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3466—Performance evaluation by tracing or monitoring
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F11/00—Error detection; Error correction; Monitoring
    • G06F11/30—Monitoring
    • G06F11/34—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment
    • G06F11/3409—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment
    • G06F11/3419—Recording or statistical evaluation of computer activity, e.g. of down time, of input/output operation ; Recording or statistical evaluation of user activity, e.g. usability assessment for performance assessment by assessing time
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2201/00—Indexing scheme relating to error detection, to error correction, and to monitoring
    • G06F2201/865—Monitoring of software
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2201/00—Indexing scheme relating to error detection, to error correction, and to monitoring
    • G06F2201/88—Monitoring involving counting

Definitions

  • the present invention relates to a performance measurement method, and more particularly to a performance measurement method for measuring the performance of a program.
  • a program developer or the like inputs a specific input to the program, and the system corresponding thereto Measure the processing performance.
  • the developer or the like needs to specify a part of the program that has a problem with the processing performance.
  • a general method that has been used in the past is to use a function embedded in the program in advance to output a log to the program being executed, and the output time from the start to the end of the entire processing of the program based on the output log There is a way to get it.
  • such a method can acquire the execution time only with rough accuracy.
  • a profiler that measures an execution time of each method by using a profiling function provided in an OS or a virtual machine. According to this method, the execution performance of the program can be analyzed with more detailed accuracy by using the profiler. The problem can be solved when the program is optimized or there is a problem in the processing performance.
  • Patent Document 1 discloses a technique for measuring the execution time with detailed accuracy only for the main method in the second measurement.
  • Patent Document 1 measures all methods corresponding to a specific input and generates statistical information (average execution time, number of executions). For this reason, when measuring a program whose overall execution time is very short, the technique disclosed in Patent Document 1 cannot acquire an accurate execution time, and a method that is actually a bottleneck is not extracted as a main method. There is.
  • the predetermined threshold value used in the technique disclosed in Patent Document 1 is a constant value for any method. For this reason, for example, when comparing performance between two programs, a program that has been modified that affects performance and a program that has not been modified, the execution time is compared with the execution time of the entire program.
  • the technique disclosed in Patent Document 1 cannot be used for the purpose of identifying a method in which there is a significant difference in the performance of the method before and after the correction.
  • the present invention has been made in view of such problems, and can accurately specify a method that affects the processing performance, and can improve the processing performance by the measurement itself regardless of the scale of the program for measuring the processing performance.
  • the purpose is to provide a measurement method that reduces the influence.
  • an object of the present invention is to provide a performance measurement method for specifying a method in which a difference in performance occurs between a plurality of programs.
  • a typical example of the present invention is as follows. That is, a performance measurement method for measuring the performance of a program by a computer, wherein the computer includes a processor and a memory for storing a program executed by the processor, and the method includes: A procedure for first determining at least one first method among methods directly called by a program to be measured, and execution of each determined first method by the computer executing the program Using the time measurement procedure and the execution time of each measured first method, the computer can extract at least one second method that matches a predetermined condition from each first method.
  • a procedure for determining whether or not the at least one second method can be extracted A procedure for determining whether or not at least one third method called directly from each extracted second method can be extracted, and when the at least one third method can be extracted, A procedure for determining the extracted third method as a first method; a procedure for the computer to measure the execution time; and a procedure for determining whether or not the second method can be extracted; And a procedure for repeatedly executing a procedure for determining whether or not the third method can be extracted and a procedure for determining the third method as the first method.
  • FIG. 1 is a block diagram illustrating a configuration of a computer 100 that analyzes a performance difference between a plurality of programs according to the first embodiment of this invention.
  • the computer 100 includes an arithmetic device 101, a memory 102, an external storage device 103, an output device 104, and an input device 105.
  • the computer 100 is a device having a function of analyzing a difference in processing performance of a plurality of programs.
  • the arithmetic unit 101 is a processor such as a CPU.
  • the arithmetic device 101 executes a program in the memory 102.
  • the memory 102 is a storage device having an area for temporarily storing data.
  • the memory 102 holds programs and tables.
  • the external storage device 103 is a storage device such as a hard disk drive. An argument or the like input to a program executed by the arithmetic device 101 is held.
  • the output device 104 is a device such as a display or a printer. This is an apparatus for providing a result of processing by the computer 100 to a user.
  • the input device 105 is a device such as a keyboard or a mouse. A user inputs an argument or the like input to a program executed by the arithmetic device 101 to the computer 100 via the input device 105.
  • the memory 102 includes a control unit 106, a test execution unit 107, a method call verification unit 108, a performance difference factor method determination unit 109, a version switching unit 110, a first method call table 111, a second method call table 112, and a target method table 113.
  • the second program 121, the Java VM 122, and the JVMTI 123 are programs.
  • the first method call table 111, the second method call table 112, the target method table 113, the first execution information table 114, the second execution information table 115, and the measurement result table 116 are tables in which data is stored.
  • the control unit 106 causes the version switching unit 110 to switch to a state where the first program 120 or the second program 121 can be executed, and causes the test execution unit 107 to execute (test) the first program 120 or the second program 121.
  • the program executed in the memory 102 is controlled, for example, the measurement results obtained by executing the first program 120 or the second program 121 are synchronized.
  • the test execution unit 107 inputs a predetermined value designated in advance by the user to a program that is in an executable state of the first program 120 or the second program 121, and the first program 120 or the second program 121. Execute. Then, the first program 120 or the second program 121 is executed by applying a load necessary for measuring the performance.
  • the method call matching unit 108 sets information that indicates a method that matches a predetermined condition in order to set a method that matches a predetermined condition as a next measurement target among methods called directly by the initial method or the performance difference factor method. Are stored in the target method table 113.
  • the initial method is a method that is called first in each thread during the period when the first program 120 or the second program 121 is being tested.
  • the performance difference factor method is a method that causes the difference in performance most when there is a difference in performance between the first program 120 and the second program 121.
  • the performance difference factor method determination unit 109 calculates a value indicating performance based on the difference between the method start time and the method end time stored in the first execution information table 114 and the second execution information table 115. Then, the calculated value is stored in the measurement result table 116.
  • the performance difference factor method determination unit 109 calculates the average execution time and the number of executions of the measurement target method in the first program 120 and the second program 121 as the first execution information table 114 and the second execution information table 115. Extract from Then, based on the extracted value, an execution time difference and an execution time ratio between the first program 120 and the second program 121 are calculated. The calculated value is stored in the measurement result table 116. Further, a performance difference factor method is extracted based on the execution time difference and the execution time ratio output to the measurement result table 116.
  • the measurement target method is a method whose performance is measured in the present embodiment.
  • the version switching unit 110 activates the program designated by the control unit 106 out of the first program 120 and the second program 121 to make the first program 120 or the second program 121 executable.
  • the first program 120 and the second program 121 are programs whose performance is measured by the computer 100 according to the first embodiment.
  • the first program 120 is a program before correction by the user
  • the second program 121 is a program after correction by the user.
  • the first program 120 and the second program 121 of this embodiment are programs created by Java or the like, and are executed in the Java VM (on the virtual machine) 122.
  • the Java VM 122 has a profile function for issuing a callback when a method called by the first program 120 and the second program 121 is started and ended.
  • the JVMTI 123 is a profile function for issuing a callback provided in the Java VM 122 of the present embodiment.
  • first program 120 and the second program 121 of the present invention are not limited to programs created in the Java language, but methods (or subroutines corresponding to methods) called by the first program 120 and the second program 121. ) May be created in any language as long as it includes a means for issuing a callback (or a signal corresponding to the callback) at the time of starting and ending.
  • the Java VM 122 and the JVMTI 123 may be any environment as long as it is an environment for executing the first program 120 and the second program 121 corresponding to the languages of the first program 120 and the second program 121.
  • the callback receiving unit 119 receives a callback issued by the JVMTI 123, and generates a time when the callback is received using a high-precision timer provided in the computer 100. Then, the generated time, information indicating the method included in the callback (class name, method name, method identifier, information indicating the calling method, frame number of the call stack, etc.) and the method are started. Information indicating whether or not the process has been completed is transmitted to the initial method recording unit 117 or the target method recording unit 118.
  • the initial method recording unit 117 is in a state where values are input to the first program 120 and the second program 121 by the test execution unit 107 (that is, while the first program 120 or the second program 121 is being tested). Extracts the highest method stored in the call stack. Then, information indicating the extracted method is stored in the first method call table 111 and the second method call table 112, respectively.
  • the method called by the first program 120 or the second program 121 is stored in the call stack for each thread and started. Then, the latest method stored in the call stack (PUSH) is acquired from the call stack earliest (POP). That is, the latest started method finishes earliest.
  • PUSH latest method stored in the call stack
  • POP call stack earliest
  • the method stored earliest in the call stack is described as the highest method, and the method stored later than the highest method in the call stack is described as a lower method.
  • the target method recording unit 118 is during the time when a value is input to the first program 120 or the second program 121 by the test execution unit 107 (that is, while the first program 120 or the second program 121 is being tested).
  • the start time and end time of the measurement target method stored in the target method table 113 and information indicating the method are received from the callback receiving unit 119.
  • Information such as the time received from the callback receiver 119 is stored in the first execution information table 114 or the second execution information table 115.
  • the target method recording unit 118 extracts information indicating a method directly called by the measurement target method while measuring the first program 120 or the second program 121. Then, information indicating the extracted method is stored in the first method call table 111 or the second method call table 112.
  • the first method call table 111 is a table in which information indicating a method called by the first program 120 is stored.
  • the second method call table 112 is a table in which information indicating a method called by the second program 121 is stored.
  • the target method table 113 is a table in which information indicating methods common to the first program 120 and the second program 121 among methods called by the first program 120 and the second program 121 is stored.
  • the first execution information table 114 is a table in which the start time and end time of the method called by the first program 120 are stored.
  • the second execution information table 115 is a table in which the start time and end time of the method called by the second program 121 are stored.
  • the measurement result table 116 includes the first program 120 and the second program 121 calculated based on the start time and end time of each method stored in the first execution information table 114 and the second execution information table 115. It is a table in which the value for comparing performance is stored.
  • FIG. 2 is an explanatory diagram illustrating the target method table 113 according to the first embodiment of this invention.
  • the target method table 113 stores information indicating a method whose start time and end time are measured.
  • the value stored in the target method table 113 is updated when a method called by the first program 120 or the second program 121 calls another method. That is, the target method table 113 is updated in order to set the method called by the measurement target method as the next measurement target method.
  • the target method table 113 includes a frame number 1131, method information 1132, and caller method information 1133.
  • the frame number 1131 is a number indicating the order of methods stored in the call stack. A smaller frame number 1131 indicates a higher method in the call stack.
  • the method information 1132 includes information indicating the measurement target method.
  • the value stored in the method information 1132 is an identifier that uniquely indicates the measurement target method, including the class name and method name of the measurement target method.
  • the caller method information 1133 includes information indicating a method that called the measurement target method indicated in the method information 1132.
  • the value stored in the caller method information 1133 has the same notation method as the value stored in the method information 1132. That is, the value stored in the caller method information 1133 is an identifier that uniquely indicates the caller method, including the class name and method name of the caller method.
  • FIG. 3A is an explanatory diagram illustrating the first method call table 111 according to the first embodiment of this invention.
  • the first method call table 111 stores information indicating a method called by the first program 120 when the first program 120 is measured.
  • the first method call table 111 includes a frame number 1111, method information 1112, and call source method information 1113.
  • the frame number 1111, the method information 1112, and the call source method information 1113 correspond to the frame number 1131, the method information 1132, and the call source method information 1133 in the target method table 113.
  • the method information 1112 of the first method call table 111 includes information indicating the method directly called by the initial method. Accumulated by the unit 118. Further, while the start time and end time of the measurement target method lower than the initial method are being measured, the method information 1112 of the first method call table 111 includes information indicating a method directly called by the measurement target method. Accumulated by the method recording unit 118.
  • FIG. 3B is an explanatory diagram illustrating the second method call table 112 according to the first embodiment of this invention.
  • the second method call table 112 stores information indicating a method called by the second program 121 when the second program 121 is measured.
  • the second method call table 112 includes a frame number 1121, method information 1122, and call source method information 1123.
  • the frame number 1121, method information 1122, and call source method information 1123 correspond to the frame number 1131, method information 1132, and call source method information 1133 in the target method table 113.
  • the information stored in the second method call table 112 is the same as the information stored in the first method call table 111, and a value extracted by executing the second program 121 is stored.
  • FIG. 4A is an explanatory diagram illustrating the first execution information table 114 according to the first embodiment of this invention.
  • the first execution information table 114 is a table that stores a start time or an end time of a method called by the first program 120.
  • the first execution information table 114 includes a frame number 1141, method information 1142, start / end information 1143, and time 1144.
  • the frame number 1141 and the method information 1142 correspond to the frame number 1131 and the method information 1132 of the target method table 113.
  • the value stored in the start / end information 1143 is “start” when the time 1144 included in each row indicates the start time of the method indicated by the method information 1142, and the time 1144 included in each row is indicated by the method information 1132. “End” indicates the end time of the method.
  • the time 1144 indicates the start time or end time of the method indicated by the method information 1142. Although the time 1144 shown in FIG. 4A is expressed by the system time used in the computer 100, the time may be indicated by any notation method.
  • FIG. 4B is an explanatory diagram illustrating the second execution information table 115 according to the first embodiment of this invention.
  • the second execution information table 115 is a table for storing the start time or end time of the method called by the second program 121.
  • the information stored in the second execution information table 115 is the same information as the first execution information table 114, and stores the value extracted by executing the second program 121.
  • the method to be measured by the computer 100 is the next measurement.
  • each table may be updated, and the frame number of each table may be updated. That is, the frame numbers included in the first method call table 111, the second method call table 112, the target method table 113, the first execution information table 114, and the second execution information table 115 all have the same value at the time of each measurement. May show.
  • each table may store the value by any method.
  • the frame number indicating the measurement target method is held by a variable
  • the measurement target method having the frame number indicated by the variable is A method of causing the computer 100 to measure may be used.
  • FIG. 5 is an explanatory diagram illustrating the measurement result table 116 according to the first embodiment of this invention.
  • the measurement result table 116 is a table that stores, for each method, values calculated based on the start time and end time included in the first execution information table 114 and the second execution information table 115.
  • the measurement result table 116 includes frame number 1161, method information 1162, average execution time 11631 and execution count 11632 of the first program, average execution time 11641 and execution count 11642 of the second program, execution time difference 1165, and execution time ratio 1166. including.
  • the frame number 1161 is a number indicating the order of methods stored in the call stack. Since a row of the measurement result table 116 is added every time a method is measured, the frame number 1161 indicates at least one value.
  • the method information 1162 stores information indicating the measured method.
  • the average execution time 11631 of the first program is an average value of the execution times of each method called multiple times when each method is called multiple times from the calling method.
  • the first execution time is calculated by extracting the method included in the method information 1142 of the first execution information table 114 and subtracting the start time from the end time of each extracted method. Then, a method indicating the same method as the extracted method is further extracted from the method information 1142, and the second execution time is calculated by calculating the execution time of the extracted method.
  • the average execution time 11631 is calculated by calculating the average value of the execution times calculated in this way.
  • the number of executions 11632 of the first program indicates the number of calls from the caller method.
  • the number of methods including both start and end in the start / end information 1153 is obtained by calculating how many methods are included in the first execution information table 114.
  • the average execution time 11641 of the second program and the execution count 11642 of the second program store the same values as the average execution time 11631 and the execution count 11632 of the first program, and are calculated based on the second execution information table 115. Stored value.
  • the execution time difference 1165 is calculated by calculating the total execution time of the method indicated by the method information 1162 in each of the first program 120 and the second program 121 every time the caller method is executed once. Indicates the difference. Specifically, based on the results of the first program 120 and the second program 121, the total execution time during which the method indicated by the method information 1162 is executed is calculated. Then, the total execution time is divided by the number of times the caller method is executed (corresponding to the execution number 11632 indicated by the caller method).
  • the values calculated by the above-described method are generated for the first program 120 and the second program 121.
  • the execution time difference 1165 is calculated by subtracting the value calculated by the second program 121 from the value calculated by the first program 120.
  • the present invention can calculate the difference in method execution time each time the caller method is executed by the execution time difference 1165.
  • a method for calculating the execution time difference 1165 of the method information 1162 “aaa.bbb.MyClass.start (int)” shown in FIG. FIG. 5 shows that the method “aaa.bbb.MyClass.start (int)” is executed by being called from “aaa.bbb.Starter.doStart ()”.
  • the number of times called by the calling method of the method “aaa.bbb.MyClass.start (int)” is 25, as indicated by the number of executions 11632 in FIG.
  • the number of executions 11642 in the second program of the method “aaa.bbb.MyClass.start (int)” and the method “aaa.bbb.Starter.doStart ()” is the number of executions 11632 in the first program. Is the same.
  • the execution time difference 1165 is calculated by the above-described equation.
  • the execution time ratio 1166 calculates the difference between the execution time per method indicated by the method information 1162 between the first program and the second program, and averages the calculated values. And it is a value calculated by dividing the averaged value by the average execution time 11631 of the first program.
  • the execution time ratio 1166 is a ratio of execution time differences between the old and new methods each time the method is executed once.
  • the present invention can calculate the rate of change of the execution time of each method based on the execution time ratio 1166.
  • the execution time ratio 1166 is a value obtained by dividing the value indicated by the execution time difference 1165 by the number of times the method indicated by the method information 1162 is executed each time the call source method is executed once. It is a value divided by the average execution time 11631 of the program.
  • the execution time difference 1165 and the execution time ratio 1166 of the initial method are calculated by assuming that the number of executions of the caller method (that is, the first program or the second program) is 1.
  • FIG. 6 is an explanatory diagram showing a screen 130 when the measurement result of the first embodiment of the present invention is displayed.
  • the screen 130 is a screen for providing the user with the result measured by the present embodiment via the output device 104.
  • the screen 130 includes a display 131, a display 132, and a display 133.
  • the display 131 shows a method name that is a performance difference factor method and a reason that it is a performance difference factor method.
  • the display 131 shown in FIG. 6 shows that the number of times that “aaa.bbb.MyClass.start (int)”, which is the calling method, calls “java.util.Map.get (Object)” is increased.
  • the program 120 is called 200 times, and the second program 121 is called 210 times.
  • the number of times the calling source method calls the method is the number of executions of the called method (the number of executions 11632 of the first program or the number of executions 11642 of the second program in the measurement result table 116). It is calculated by dividing by the number of executions 11632 of the first program or the number of executions 11642 of the second program in the measurement result table 116.
  • the display 132 is a display showing the method called by the hierarchical structure before the performance difference factor method is called. That is, the result of tracing the call stack including the performance difference factor method is shown.
  • the display 132 shows each method of the call stack including the performance difference factor method and the performance difference between the first program 120 and the second program 121 in each method.
  • the display 133 displays information including the measurement result table 116.
  • the display 133 illustrated in FIG. 6 includes information of the measurement result table 116 and caller method information 1133 of the target method table 113.
  • FIG. 7 is a flowchart showing the overall processing of the first embodiment of the present invention.
  • the control unit 106 extracts initial methods of the first program 120 and the second program 121, and among the extracted initial methods, a method common to the first program 120 and the second program 121 is stored in the target method table 113.
  • the target method table 113 stores a method common to the first program 120 and the second program 121 in the target method table 113, thereby measuring the method (measurement target method) for measuring the start time and the end time as the target method.
  • the table 113 140.
  • control unit 106 measures the execution time of the measurement target method stored in the target method table 113 (150). After step 150, the control unit 106 extracts a performance difference factor method based on the measurement result of step 150 (160).
  • control unit 106 determines whether or not a performance difference factor method has been extracted in Step 160 (170). When the performance difference factor method is not extracted, there is no performance difference between the first program 120 and the second program 121, so the control unit 106 ends the process.
  • the control unit 106 stores the method directly called by the performance difference factor method in the target method table 113, and further measures the method directly called by the performance difference factor method (180). ). This is because the method in which the performance difference is most noticeable is extracted from the methods directly called by the performance difference factor method.
  • step 180 the control unit 106 returns to step 150. Thereafter, when it is determined in step 170 that the performance difference factor method is not extracted, the control unit 106 terminates the process to indicate that the performance difference is the greatest in the caller method that called the measurement target method. To do.
  • FIG. 8 is a sequence diagram showing an initial method according to the first embodiment of this invention.
  • the sequence diagram shown in FIG. 8 shows the execution times of thread A 1071, thread B 1072, thread 1073, thread D 1074, and method 201 to method 216.
  • step 140 is during the test execution time from when the test execution unit 107 starts to input specific input values to the first program 120 and the second program 121 until when the specific input values have been input.
  • the call stack of each thread included in the first program 120 and the second program 121 it is a process of extracting the method stored at the highest level.
  • the program 200 is executed from the time when the test execution unit 107 starts to input a specific input value to the first program 120 or the second program 121 to the time when the input of the specific input value is finished.
  • the program 120 or the second program 121 is shown.
  • the program 200 calls the method 201 executed in the thread A 1071.
  • a method 201 called by the program 200 is a method that is always executed, and is a method that is being executed before the program 200 is started. For this reason, the method 201 is not an initial method in the thread A 1071 because it is not a method that causes a difference in performance.
  • the method 201 calls the method 203 executed in the thread B1072.
  • a method 203 called by the method 201 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 203 is an initial method in the thread B1072.
  • the method 203 is stored at the top of the call stack in the thread B 1072.
  • the method 202 and the method 209 in the thread B are not methods executed while the program 200 is executed, and are not initial methods because they do not cause a difference in performance.
  • the method 203 calls the method 204 and the method 205 while the method 203 is being executed.
  • a method 204 and a method 205 called by the method 203 are methods that are executed after the program 200 is started and ended before the program 200 ends. Therefore, the method 204 and the method 205 are methods that may become measurement target methods.
  • the frame numbers of the method 204 and the method 205 are “2”.
  • the method 205 calls a method 206 and a method 207.
  • the frame number of the method 203 is “1”
  • the frame number in the call stack of the method 206 and the method 207 is “3”.
  • method 207 calls method 208.
  • the frame number of the method 203 is “1”
  • the frame number of the method 208 in the call stack is “4”.
  • a method 204, a method 205, a method 206, a method 207, and a method 208 are methods that are executed after the program 200 starts and end before the program 200 ends. Therefore, the method 204, the method 205, the method 206, the method 207, and the method 208 are methods that may become measurement target methods.
  • Method 204 calls method 210 in thread C1073.
  • a method 210 called by the method 204 is a method that is executed even after the program 200 ends. For this reason, the method 210 does not cause a difference in performance, and thus is not an initial method in the thread C1073.
  • Method 210 calls method 211.
  • a method 211 called by the method 210 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 211 is an initial method in the thread C1073.
  • the method 211 is stored above the call stack in the thread C1073. At this time, since the method 210 has already been stored in the call stack of the thread C1073, the frame number of the method 211 is “2”.
  • method 211 calls method 212.
  • the frame number in the call stack of the method 212 is “3”.
  • a method 212 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 212 is a method that may be a measurement target method.
  • Method 211 calls method 214 in thread D1074.
  • a method 214 called by a method 211 is a method that is being executed before the program 200 is started. For this reason, the method 214 does not cause a difference in performance, and thus is not an initial method in the thread D1074.
  • Method 214 calls method 216.
  • a method 216 called by the method 214 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 216 is an initial method.
  • the frame number of the method 216 is “3”.
  • the method 215 of the thread D 1074 is a method that is executed after the method 216 and the method 214 are finished, and is finished before the program 200 is finished. For this reason, it becomes a factor that causes a difference in performance, and thus becomes an initial method.
  • the frame number of the method 216 is “2”.
  • control unit 106 of the present embodiment extracts the initial method described above in step 140.
  • FIG. 9 is a block diagram illustrating a data flow in the process of extracting the initial method according to the first embodiment of this invention.
  • the arrows shown in FIG. 9 indicate the data flow in step 140 shown in FIG. 7, and correspond to the processing shown in FIGS. 10A, 10B, and 10C.
  • FIG. 10A is a flowchart showing a process for executing a test to extract an initial method according to the first embodiment of this invention.
  • the process shown in FIG. 10A is a process included in step 140 shown in FIG. First, after receiving an instruction from the control unit 106 to set the first program 120 in a testable state, the version switching unit 110 switches the first program 120 to a state in which the first program 120 can be executed. 106 is notified (1401).
  • control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the initial method recording unit 117.
  • the initial method recording unit 117 stores “TRUE” in the test execution flag (1402). Note that the initial method recording unit 117 holds an area indicating the test execution flag in the memory 102 in advance.
  • test execution flag When “TRUE” is stored in the test execution flag, the test execution flag indicates that the test execution unit 107 is testing the program. When “FALSE” is stored in the test execution flag, the test execution flag indicates that the test execution unit 107 has not tested the program.
  • the control unit 106 transmits an instruction to start the test of the first program 120 to the test execution unit 107.
  • the test execution unit 107 receives an instruction to start the test of the first program 120 from the control unit 106, the test execution unit 107 starts the test of the first program 120 (1403).
  • the test execution unit 107 inputs a predetermined input value designated in advance by the user to the first program 120 in order to execute the test (1404). After step 1404, the test execution unit 107 ends the test of the first program 120, and then notifies the control unit 106 of the end of the test (1405).
  • control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the initial method recording unit 117.
  • the initial method recording unit 117 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1406).
  • control unit 106 transmits an instruction to end the first program 120 to the version switching unit 110 (1407).
  • the initial method recording unit 117 extracts a method to which a completed flag is added from a call stack for each thread (hereinafter referred to as a method information stack). Information indicating the extracted method is stored in the first method call table 111 (1409).
  • the completed flag is a storage area that is updated in the process shown in FIG.
  • the initial method recording unit 117 repeats Step 1409 for the method information stack of all threads (1408).
  • step 1408 the version switching unit 110 receives an instruction from the control unit 106 to make the second program 121 ready for testing. Thereafter, the version switching unit 110 switches the second program 121 to an executable state, and then notifies the control unit 106 of the end of the switching (1410).
  • control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the initial method recording unit 117.
  • the initial method recording unit 117 stores an instruction to store “TRUE” in the test execution flag from the control unit 106, and then stores “TRUE” in the test execution flag (1411).
  • the control unit 106 transmits an instruction to start the test of the second program 121 to the test execution unit 107.
  • the test execution unit 107 receives an instruction to start the test of the second program 121 from the control unit 106, the test execution unit 107 starts the test of the second program 121 (1412).
  • the test execution unit 107 inputs a predetermined value designated in advance by the user to the second program 121 in order to execute the test (1413). After step 1413, the test execution unit 107 ends the test of the second program 121, and then notifies the control unit 106 of the end of the test (1414).
  • control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the initial method recording unit 117.
  • the initial method recording unit 117 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1415).
  • control unit 106 transmits an instruction to end the second program 121 to the version switching unit 110 (1416).
  • the initial method recording unit 117 extracts the method for which the corresponding completed flag is generated from the method information stack for each thread, and indicates the extracted method.
  • the information is stored in the first method call table 111 (1418).
  • Step 1418 for the method information stack of all threads (1417).
  • control unit 106 extracts a measurement target method (1419), and ends the process illustrated in FIG. 10A.
  • FIG. 10B is a flowchart showing a process of extracting a measurement target method according to the first embodiment of this invention.
  • the process shown in FIG. 10B corresponds to step 1419 shown in FIG. 10A.
  • the control unit 106 transmits an instruction to extract the measurement target method to the method call matching unit 108 after step 1417 in FIG. 10A is completed (1420).
  • the method call matching unit 108 After receiving the instruction to extract the measurement target method from the method call matching unit 108, the method call matching unit 108 reads information stored in the first method call table 111 (1421). Further, the method call matching unit 108 reads information stored in the second method call table 112 (1422).
  • the method call matching unit 108 deletes the data stored in the target method table 113 in order to store the new measurement target method in the target method table 113 (1423).
  • the method call verification unit 108 is common to the first method call table 111 and the second method call table 112 based on the information read from the first method call table 111 and the second method call table 112. The extracted method is extracted, and the extracted method is stored in the target method table 113 (1424).
  • step 1424 the method call matching unit 108 notifies the control unit of the end of extraction of the measurement target method, and ends the process of step 1419 (1425).
  • step 1419 the methods stored in common in the first method call table 111 and the second method call table 112 are extracted as measurement target methods.
  • FIG. 10C is a flowchart showing processing for extracting an initial method according to the first embodiment of this invention.
  • the process shown in FIG. 10C is a process for extracting an initial method from methods called by each program during the execution of the test shown in FIG. 10A.
  • the callback receiving unit 119 receives the callback from the JVMTI 123, and transmits the time when the callback is received and each information included in the callback to the initial method recording unit 117.
  • the callback includes information indicating the method (class name, method name, method identifier, information indicating the calling method, frame number in the call stack, etc.), and whether the method has started or ended. Is included.
  • the initial method recording unit 117 determines whether “TRUE” is stored in the test execution flag (1431). If “FALSE” is stored in the test execution flag, the test has not been executed and it is not necessary to extract the initial method, so the initial method recording unit 117 proceeds to step 1438.
  • the initial method recording unit 117 determines whether the method is started or ended based on the information included in the received callback (1432).
  • step 1433 the method stored in the method information stack becomes an initial method candidate.
  • the initial method recording unit 117 determines whether or not the method information included in the callback is already stored in any of the method information stacks ( 1434).
  • the determination in step 1434 can determine whether or not the method indicated by the callback is a method started while the test is being executed. As described above, the initial method is a method that is started and ended while the test is being executed.
  • step 1434 If it is determined in step 1434 that the method information included in the callback is not stored in any of the method information stacks, the method indicated by the callback is not started while the test is being executed. Since it is a method and not an initial method, the initial method recording unit 117 proceeds to Step 1439.
  • step 1434 If it is determined in step 1434 that the method information included in the callback is stored in any of the method information stacks, the method indicated by the callback is started while the test is being executed. It is a method that can be an initial method. Therefore, the initial method recording unit 117 proceeds to step 1435 after step 1434.
  • the initial method recording unit 117 obtains the method stored at the lowest level of the method information stack determined in step 1434 that the method information included in the callback is stored (POP). Then, information indicating the acquired method is stored in the first method information (1435).
  • the first method information is a storage area for temporarily storing information indicating a method, and is held in advance by the initial method recording unit 117.
  • the initial method recording unit 117 determines whether or not the top method information is the same as the information indicating the method included in the callback (1436).
  • the method stored in the head method information is a lower-level method than the method included in the callback. Therefore, the initial method recording unit 117 returns to step 1435 in order to store a higher-order method in the top method information.
  • step 1436 when the head method information is the same as the information indicating the method included in the callback, the initial method recording unit 117 generates a completed flag corresponding to the head method information (1437).
  • the completed flag is generated correspondingly, the method indicated by the first method information indicates that it has been completed.
  • the initial method recording unit 117 stores (PUSH) again the method information stored in the head method information in the method information stack acquired in step 1435 (1438).
  • the finished method is stored at the lowest position in the method information stack. Also, the process shown in FIG. 10C is executed each time a callback is issued, so that the highest method among the methods started and ended during the test execution remains in the method information stack. For this reason, after the test is finished, the method that is stored in the method information stack and the corresponding finished flag is generated is the initial method.
  • the initial method recording unit 117 ends the process of FIG. 10C and notifies the callback receiving unit 119 of the end of the process (1439).
  • the initial method recording unit 117 After the initial method is stored in the method information stack by the process shown in FIG. 10C, the initial method recording unit 117 generates a completed flag in the method information stack for each thread in steps 1409 and 1418 shown in FIG. 10A.
  • the stored method is stored in the first method call table 111 or the second method call table 112. As a result, the initial method is stored in the first method call table 111 and the second method call table 112.
  • FIG. 11 is a block diagram illustrating a data flow in the process of measuring the execution time of the measurement target method according to the first embodiment of this invention.
  • the arrows shown in FIG. 11 indicate the data flow in step 150 shown in FIG. 7, and correspond to the processing shown in FIGS. 12A and 12B.
  • FIG. 12A is a flowchart illustrating a process for executing a test to measure the execution time according to the first embodiment of this invention.
  • the process shown in FIG. 12A is a process included in step 150 shown in FIG. First, after receiving an instruction from the control unit 106 to set the first program 120 in a testable state, the version switching unit 110 switches the first program 120 to a state in which the first program 120 can be executed. 106 is notified (1501).
  • control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the target method recording unit 118.
  • the target method recording unit 118 receives an instruction to store “TRUE” in the test execution flag from the control unit 106, and then stores “TRUE” in the test execution flag (1502).
  • the target method recording unit 118 holds an area indicating the test execution flag in the memory 102 in advance.
  • the test execution flag held by the initial method recording unit 117 and the test execution flag held by the target method recording unit 118 may be stored in the same area.
  • the control unit 106 transmits an instruction to start the test of the first program 120 to the test execution unit 107.
  • the test execution unit 107 receives an instruction to start the test of the first program 120 from the control unit 106, the test execution unit 107 starts the test of the first program 120 (1503).
  • the test execution unit 107 inputs a predetermined value designated in advance by the user to the first program 120 in order to execute the test (1504). After step 1504, the test execution unit 107 ends the test of the first program 120, and then notifies the control unit 106 of the end of the test (1505).
  • control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the target method recording unit 118.
  • the target method recording unit 118 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1506).
  • control unit 106 transmits an instruction to end the first program 120 to the version switching unit 110 (1507).
  • the version switching unit 110 receives an instruction from the control unit 106 to make the second program 121 ready for testing. Thereafter, the version switching unit 110 switches the second program 121 to an executable state, and then notifies the control unit 106 of the end of the switching (1508).
  • control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the target method recording unit 118.
  • the target method recording unit 118 receives an instruction to store “TRUE” in the test execution flag from the control unit 106, and then stores “TRUE” in the test execution flag (1509).
  • the control unit 106 transmits an instruction to start the test of the second program 121 to the test execution unit 107.
  • the test execution unit 107 receives an instruction to start the test of the second program 121 from the control unit 106, the test execution unit 107 starts the test of the second program 121 (1510).
  • the test execution unit 107 inputs a predetermined value designated in advance by the user to the second program 121 in order to execute the test (1511). After step 1511, the test execution unit 107 ends the test of the second program 121, and then notifies the control unit 106 of the end of the test (1512).
  • control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the target method recording unit 118.
  • the target method recording unit 118 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1513).
  • control unit 106 transmits an instruction to end the second program 121 to the version switching unit 110 (1514), and ends the process illustrated in FIG. 12A.
  • FIG. 12B is a flowchart illustrating processing for measuring the execution time of the measurement target method according to the first embodiment of this invention.
  • the process shown in FIG. 12B shows the process of measuring the execution time of the measurement target method when the first program 120 is executed.
  • the process for measuring the execution time of the measurement target method when the second program 121 is executed is performed by changing the first program 120 to the second program 121 included in the process shown in FIG. This is a process in which the table 111 is changed to the second method call table 112 and the first execution information table 114 is changed to the second execution information table 115.
  • the callback receiving unit 119 receives a callback from the JVMTI 123 during execution of the test of the first program 120, and records the time when the callback is received and each piece of information included in the callback as the target method recording unit 118. (1514).
  • the target method recording unit 118 determines whether “TRUE” is stored in the test execution flag (1515). When “FALSE” is stored in the test execution flag, since the test is not executed and the execution time cannot be measured, the target method recording unit 118 proceeds to step 1526 and the callback reception unit 119 performs processing. Notify the end.
  • Information indicating which of the first program 120 or the second program 121 is being tested may be added to the test execution flag, and the target method recording unit 118 refers to the test execution flag. Thus, it may be determined whether processing is performed using the first execution information table 114 and the first method call table 111 or processing is performed using the second execution information table 115 and the second method call table 112.
  • the target method recording unit 118 determines whether “FALSE” is stored in the method execution flag of the thread in which the method indicated by the callback is executed. (1516).
  • the method executing flag is a storage area that is held in advance by the target method recording unit 118.
  • FALSE When “FALSE” is stored in the method execution flag of a thread, it indicates that the measurement target method is not executed in the thread.
  • TRUE When “TRUE” is stored in the method execution flag of a thread, Indicates that the measurement target method is being executed in the thread.
  • the method executing flag is updated in step 1519 and step 1522 described later.
  • the target method recording unit 118 may acquire the thread in which the method is executed from information indicating the method included in the callback.
  • step 1516 If it is determined in step 1516 that “FALSE” is stored in the method execution flag of the thread in which the method is executed, the callback received in step 1514 indicates the start of the method, so that the target method recording unit 118 moves to step 1517 to store the start time.
  • the target method recording unit 118 determines whether or not the method indicated by the callback is stored in the target method table 113 (1517). If the method indicated by the callback is not stored in the target method table 113, the target method recording unit 118 proceeds to step 1526 because the method indicated by the callback is not a measurement target method.
  • step 1517 If it is determined in step 1517 that the method indicated by the callback is stored in the target method table 113, the target method recording unit 118 is included in the callback because the method indicated by the callback is the measurement target method.
  • Information indicating the method and the time included in the callback (that is, the start time of the method) are stored in the first execution information table 114 (1518).
  • the target method recording unit 118 stores “TRUE” in the method execution flag of the thread in which the method indicated by the callback is executed. Further, the target method recording unit 118 stores the frame number in the call stack of the method included in the callback in the measurement target frame of the thread in which the method is executed (1519).
  • the measurement target frame is a storage area for temporarily storing the frame number held by the target method recording unit 118.
  • the value stored in the measurement target frame indicates the frame number of the measurement target method and is held for each thread.
  • step 1519 the target method recording unit 118 proceeds to step 1526 and notifies the callback receiving unit 119 of the end of processing.
  • step 1516 If it is determined in step 1516 that “TRUE” is stored in the method execution flag of the thread in which the method indicated by the callback is executed, the callback received in step 1514 indicates the end of the method.
  • the target method recording unit 118 proceeds to Step 1520 to store the end time.
  • the target method recording unit 118 determines whether the number stored in the measurement target frame of the thread in which the method indicated by the callback is executed is the same as the frame number of the method indicated by the callback ( 1520).
  • the method indicated by the callback is the method that is started during the execution of the test and the start time is stored in step 1518. . Therefore, the target method recording unit 118 proceeds to Step 1521.
  • the target method recording unit 118 stores information indicating the method included in the callback and the time included in the callback (that is, the end time of the method) in the first execution information table 114 (1521). ). After step 1521, the target method recording unit 118 stores “FALSE” in the method executing flag (1522).
  • step 1522 the target method recording unit 118 proceeds to step 1526 and notifies the callback receiving unit 119 of the end of processing.
  • step 1520 when the number stored in the measurement target frame of the thread in which the method indicated by the callback is executed is different from the frame number of the method indicated by the callback, the method indicated by the callback is not yet completed. A method called by a target method or a method further called from a method called by a measurement target method. Therefore, the target method recording unit 118 proceeds to Step 1523.
  • the target method recording unit 118 determines whether or not the value stored in the measurement target frame is a value obtained by subtracting 1 from the frame number of the method indicated by the callback. That is, it is determined whether or not the method indicated by the callback is a method directly called by the measurement target method (1523).
  • the target method recording unit 118 determines whether or not the callback indicates the start of the method ( 1524). If the callback does not indicate the start of the method but indicates the end in step 1524, the target method recording unit 118 proceeds to step 1626.
  • step 1524 when the callback indicates the start of the method, the method indicated by the callback is the next measurement target method. Therefore, the target method recording unit 118 stores information indicating the method included in the callback as the first method. Store in the call table 111 (1525).
  • the values stored in the first method call table 111 and the second method call table 112 may be deleted before the test shown in FIG. 12A is executed. That is, the first method call table 111 and the second method are stored in the first method call table 111 and the second method call table 112 by the step 1525 so that only methods that may become the next measurement target methods are stored in the first method call table 111 and the second method call table 112.
  • the value indicating the previous initial method or measurement target method included in the call table 112 may be deleted at the start of the test illustrated in FIG. 12A.
  • Step 1526 the target method recording unit 118 proceeds to Step 1526.
  • step 1526 the callback receiver 119 waits until the next callback is received.
  • the target method recording unit 118 measures the execution time of the measurement target method and extracts the next measurement target method.
  • FIG. 13 is a flowchart showing processing for extracting a performance difference factor method according to the first embodiment of this invention.
  • step 160 shown in FIG. Step 160 is executed each time step 150, that is, the process shown in FIG. 12A is performed.
  • the control unit 106 transmits an instruction to extract the performance difference factor method to the performance difference factor method determination unit 109 (1601).
  • the performance difference factor method determination unit 109 receives an instruction to extract a performance difference factor method from the control unit 106, and then holds a temporary storage area in which the variable i and the maximum performance difference factor rate are stored in the memory 102. Then, 1 is stored in the variable i, and 0 is stored in the maximum performance difference factor rate (1602).
  • variable i is a pointer indicating the row of the target method table 113, and the maximum performance difference factor rate indicates the maximum value of the performance difference factor rate described later.
  • step 1602 the performance difference factor method determination unit 109 repeats the processing from step 1604 to step 1613 until the value stored in the variable i exceeds the number of rows in the target method table 113.
  • the performance difference factor method determination unit 109 calculates the average execution time 11631 and the number of executions 11632 of the measurement target method in the first program 120 based on the first execution information table 114.
  • the performance difference factor method determination unit 109 extracts one measurement target method from the target method table 113, and extracts a line in the output device 104 where the value of the method information 1142 indicates the measurement target method. Then, two consecutive lines whose start / end information 1143 indicates “start” and “end” are extracted from the extracted lines. Then, the number of executions 11632 is calculated by calculating the number of sets of two extracted rows.
  • the performance difference factor method determination unit 109 calculates a method execution time per time by calculating a difference between the times 1144 of the two extracted rows. Then, the average execution time 11631 is calculated by calculating the execution time of each of the two sets of extracted rows and calculating the average value of the execution times of all the sets.
  • the performance difference factor method determination unit 109 calculates the average execution time 11641 and the number of executions 11642 of the measurement target method in the second program 121 based on the second execution information table 115 (1605).
  • the specific calculation method of the average execution time 11641 and the execution count 11642 is the same as the calculation method of the average execution time 11631 and the execution count 11632, but the average execution time 11641 and the execution count 11642 are the same as those in the first execution information table 114. Instead, it is calculated based on the second execution information table 115.
  • the performance difference factor method determination unit 109 calculates an execution time difference 1165 and an execution time ratio 1166 (step 1606). Specifically, the performance difference factor method determination unit 109 calculates the total execution time of each group calculated in step 1604 from the value obtained by dividing the total execution time of each group calculated in step 1605 by the execution count 11632. Is subtracted by the number of executions 11642 to calculate the execution time difference 1165.
  • step 1606 the performance difference factor method determination unit 109 further divides the calculated execution time difference 1165 by the execution count 11632, and further divides the divided result by the average execution time 11631, thereby executing the execution time.
  • the ratio 1166 is calculated.
  • the performance difference factor method determination unit 109 calculates the average execution time 11631, the execution count 11632, the average execution time 11641, the execution count 11642, the execution time difference 1165, calculated in step 1604, step 1605, and step 1606.
  • the execution time ratio 1166 is stored in the measurement result table 116 (1607).
  • step 1604 to step 1607 may be performed for all methods included in the target method table 113.
  • the performance difference factor method determination unit 109 executes the method information 1162 of the measurement result table 116 corresponding to the method information 1132 in the row indicated by the variable i of the target method table 113, the execution time difference 1165, and the execution time ratio 1166. Are extracted (1608).
  • the performance difference factor method determination unit 109 determines whether or not the execution time ratio 1166 extracted in step 1608 is equal to or greater than the execution time ratio threshold specified in advance by the user (1609). If the execution time ratio 1166 is not equal to or greater than the execution time ratio threshold, the performance difference between the measurement target methods in the first program 120 and the second program 121 is in a range that is not problematic for the user. Control goes to step 1613.
  • the performance difference between the measurement target methods in the first program 120 and the second program 121 is in a range that causes a problem for the user. Shifts to step 1610.
  • the performance difference factor method determination unit 109 extracts the caller method information 1133 of the row indicated by the variable i from the target method table 113, and executes the method information 1162 corresponding to the extracted caller method information 1133.
  • the time difference 1165 is extracted from the measurement result table 116.
  • the performance difference factor method determination unit 109 stores the extracted execution time difference 1165 in the caller execution time difference (1610).
  • the caller execution time difference is a partial storage area in the memory 102 held by the performance difference factor method determination unit 109.
  • the performance difference factor method determination unit 109 divides the execution time difference 1165 extracted in step 1608 by the caller execution time difference. The divided result is stored in the performance difference factor rate (1611).
  • the performance difference factor rate is a partial storage area in the memory 102 held by the performance difference factor method determination unit 109.
  • the performance difference factor method determination unit 109 determines whether or not the value indicated by the maximum performance difference factor rate is greater than the value indicated by the performance difference factor rate calculated in step 1611 (1612). .
  • the maximum performance difference factor rate is a partial storage area in the memory 102 held by the performance difference factor method determination unit 109.
  • the maximum performance difference factor rate is the value of the performance difference factor rate showing the largest value among the performance difference factor rates calculated by executing Steps 1608 to 1611 in each row indicated by the variable i of the target method table 113. This is an area for storing values.
  • step 1612 if the value indicated by the maximum performance difference factor rate is not greater than the value indicated by the performance difference factor rate calculated in step 1611, the value indicated by the maximum performance difference factor rate is not changed.
  • the unit 109 proceeds to step 1613.
  • step 1612 when the value indicated by the maximum performance difference factor rate is greater than the value indicated by the performance difference factor rate calculated in step 1611, the performance difference factor method determination unit 109 determines that the performance difference factor calculated in step 1611 The maximum performance difference factor rate is updated according to the rate. Also, the performance difference factor method is updated with the value of the method information 1132 in the row of the target method table 113 indicated by the variable i (1613). Through the processing in step 1612 and step 1613, a performance difference factor method is extracted.
  • step 1609 the performance difference factor method determination unit 109 adds 1 to the value stored in the variable i. That is, one row of the target method table 113 for performing the processing shown in steps 1604 to 1613 is moved (1613).
  • the performance difference factor method determination unit 109 After performing the processing of step 1604 to step 1613 for all the rows of the target method table 113, the performance difference factor method determination unit 109 notifies the control unit 106 that the extraction process of the performance difference factor method has ended ( 1614).
  • the performance difference between the first program and the second program is compared by calculating the execution time ratio 1166, but the performance difference is determined by comparing the execution time difference 1165. May be compared. That is, when the number of calls from the caller method is constant, the performance difference may be compared by the execution time difference 1165.
  • step 160 shown in FIG. 7 After the process shown in FIG. 13, that is, step 160 shown in FIG. 7 is completed, if the value is stored in the performance difference factor method, the process shown in FIG. 14, that is, step 180 is executed.
  • FIG. 14 is a flowchart showing processing for setting the measurement target method again according to the first embodiment of this invention.
  • the process shown in FIG. 14 corresponds to the process shown in Step 180 of FIG.
  • control unit 106 transmits an instruction to set the measurement target method again to the method call matching unit 108 (1801).
  • step 1801 the method call verification unit 108 reads the first method call table 111 (1802) and reads the second method call table 112 (1803). Then, the value stored in the target method table 113 is deleted (1804).
  • the method information 1132 stored in step 1525 of FIG. 12B is held in the first method call table 111 and the second method call table 112. That is, information indicating a method that may become the next measurement target method is held.
  • the method call verification unit 108 calls the source method information (1113 and 1122) out of the method information (1112 and 1122) held in common in the first method call table 111 and the second method call table 112.
  • the method indicated by 1123) stores the method information (1112 and 1122) that is the performance difference factor method extracted by the processing shown in FIG. 13 in the target method table 113 (1805).
  • step 1805 the method call matching unit 108 notifies the control unit 106 of the end of resetting the measurement target method (1806).
  • step 150 After the process shown in FIG. 14, the process from step 150 to step 180 shown in FIG. 7 is repeated until no performance difference factor method is extracted.
  • the first embodiment for example, it is possible to compare the performance of a program to which a modification that affects performance is added and a program before the modification. Further, even when there is no significant difference between the execution time of the program before correction and the execution time of the program after correction, it is possible to extract a method having a significant difference in performance by comparing the execution times for each method.
  • the performance difference factor between the two user applications can be analyzed, and if the user application is fixed, the application server Can analyze the performance difference between different versions.
  • the method for measuring the execution time is extracted, the influence on the load due to the measurement of the execution time is reduced, and a method including a process having a problem in performance is detected with high accuracy. can do.
  • FIG. 15 is a block diagram illustrating a configuration of the computer 100 that analyzes a performance difference of a single program according to the second embodiment of this invention.
  • the computer 100 of the second embodiment executes a test of one program and measures the performance of one program. Then, the threshold specified in advance by the user is compared with the measured performance, and a performance difference factor method is extracted.
  • the computer 100 of the second embodiment includes the same memory 102, arithmetic device 101, external storage device 103, output device 104, and input device 105 as the computer 100 of the first embodiment.
  • the JVMTI 313 includes a control unit 106, a test execution unit 107, a method call matching unit 108, a performance difference factor method determination unit 109, a version switching unit 110, an initial method recording unit 117, and a target method recording unit according to the first embodiment.
  • 118, the callback receiver 119, the JavaVM 122, and the JVMTI 123 have the same functions.
  • the target method table 305 and the measurement result table 307 of the second embodiment include the same information as the target method table 113 and the measurement result table 116 of the first embodiment.
  • control unit 300 tests only one measurement target program 311, a method call table 304 corresponding to the first method call table 111 and an execution corresponding to the first execution information table 114 are performed. There is one information table 306 for each.
  • FIG. 16 is an explanatory diagram illustrating a method call table 304 according to the second embodiment of this invention.
  • the method call table 304 includes the same information as the first method call table 111 and the second method call table 112 of the first embodiment.
  • the frame number 3041, the method information 3042, and the caller method information 3043 are the same as those in the first embodiment. This corresponds to the frame number 1111, method information 1112, and caller method information 1113 of the form.
  • FIG. 17 is an explanatory diagram illustrating an execution information table 306 according to the second embodiment of this invention.
  • the execution information table 306 includes the same information as the first execution information table 114 and the second execution information table 115 of the first embodiment, and the frame number 3061, method information 3062, start / end information 3063, and time 3064 are the first information. Corresponds to the frame number 1141, method information 1142, start / end information 1143, and time 1144.
  • FIG. 18 is an explanatory diagram illustrating a measurement result table 307 according to the second embodiment of this invention.
  • the measurement result table 307 includes a frame number 3071, method information 3072, an average execution time 3073, and an execution count 3074.
  • the frame number 3071, the method information 3072, the average execution time 3073, and the execution count 3074 correspond to the frame number 1161, the method information 1162, the average execution time 11631, and the execution count 11632 in the measurement result table 116 of the first embodiment.
  • FIG. 19 is a flowchart showing the overall processing of the second embodiment of the present invention.
  • the control unit 300 first extracts the initial method of the measurement target program 311 and stores the extracted initial method in the target method table 305. As a result, the measurement target method is stored in the target method table 305 (320).
  • step 320 the control unit 300 executes the processing shown in FIGS. 10A to 10C only for the measurement target program 311. Specifically, the control unit 300 executes the test of the measurement target program 311 by causing the test execution unit 301 to execute steps 1402 to 1409 shown in FIG. 10A. When the process illustrated in FIG. 10A is executed, the control unit 300 causes the callback receiver 310 and the initial method recording unit 308 to execute the process illustrated in FIG.
  • step 320 the control unit 300 causes the method call matching unit 302 to execute steps 1421, 1423, and 1424 shown in FIG.
  • step 1424 the method call matching unit 302 stores all the method information 3042 included in the method call table in the target method table 305.
  • step 320 the control unit 300 measures the execution time of the measurement target method (330).
  • step 330 the control unit 300 executes the process shown in FIGS. 12A and 12B only for the measurement target program 311.
  • control unit 300 executes the test of the measurement target program 311 by causing the test execution unit 301 to execute steps 1502 to 1507 in step 330. Then, while the measurement target program 311 is being tested, the target method recording unit 309 executes the process shown in FIG.
  • control unit 300 extracts a method that satisfies a predetermined condition.
  • FIG. 20 is a flowchart showing a process of extracting a method that satisfies a predetermined condition according to the second embodiment of the present invention.
  • the process shown in FIG. 20 corresponds to step 340 shown in FIG.
  • the control unit 300 transmits an instruction to extract a method satisfying a predetermined condition to the performance difference factor method determination unit 303 (3401).
  • the performance difference factor method determination unit 303 holds the variable i in the memory 102 and stores 1 in the variable i (3402).
  • a variable i is a pointer indicating a row of the target method table 305.
  • step 3402 the performance difference factor method determination unit 303 repeats the processing from step 3403 to step 3408 until the value stored in the variable i exceeds the number of rows in the target method table 305 (3402).
  • the performance difference factor method determination unit 303 calculates the average execution time 3073 and the number of executions 3074 of the measurement target method based on the execution information table 306 (3403).
  • the calculation method of the average execution time 3073 and the execution count 3074 is the same as the calculation method of the average execution time 11631 and the execution count 11632 of the first embodiment.
  • the performance difference factor method determination unit 303 stores the calculated average execution time 3073 and execution count 3074 in the measurement result table 307 (3404).
  • the performance difference factor method determination unit 303 determines whether or not a value obtained by multiplying the average execution time 3073 and the number of executions 3074 is larger than a threshold value specified in advance by the user (3405). If the value obtained by multiplying the average execution time 3073 and the number of executions 3074 does not exceed the threshold value designated in advance by the user, there is no problem with the method indicated by the variable i. Transition.
  • step 3405 If it is determined in step 3405 that the value obtained by multiplying the average execution time 3073 and the number of executions 3074 is larger than the threshold value specified in advance by the user, there is a possibility that there is a problem in the method indicated by the variable i.
  • the difference factor method determination unit 303 proceeds to step 3406.
  • the performance difference factor method determination unit 303 determines whether the value obtained by multiplying the average execution time 3073 and the execution count 3074 is greater than the value obtained by multiplying the average execution time 3073 and the execution count 3074 of the maximum execution time method. It is determined (3407).
  • the maximum execution time method is a storage area held in the memory 102 by the performance difference factor method determination unit 303, and among the methods satisfying a predetermined condition, the method having the largest value obtained by multiplying the average execution time 3073 and the execution count 3074 This is an area for storing information indicating.
  • the performance difference factor method determination unit 303 proceeds to Step 3409. .
  • the maximum execution time method is updated by the method indicated by the variable i. (3408).
  • step 3406 After step 3406, step 3407, or step 3408, the performance difference factor method determination unit 303 adds 1 to the value stored in the variable i (3409). As a result, the performance difference factor method determination unit 303 can acquire the next row of the target method table 305.
  • the performance difference factor method determination unit 303 notifies the control unit 300 of the end of the process of extracting a method that satisfies a predetermined condition (3410).
  • the first method call table 111 and the first execution information table 114 are the method call table 304. And the target method recording unit 309.
  • the computer 100 can execute both the first embodiment and the second embodiment. That is, the user designates in advance whether to measure the performance of one program or to compare the performance of two programs, and the control unit 106 or the control unit 300 determines whether the first program is based on the designation from the user. You may choose to perform one embodiment or a second embodiment.
  • the initial method in the first embodiment and the second embodiment described above is automatically extracted by executing the program, it may be specified by the user.
  • FIG. 21 is an explanatory diagram showing a screen for designating an initial method according to the second embodiment of this invention.
  • a screen 400 shown in FIG. 21 is a screen for designating an initial method of the second embodiment, but may be used in the first embodiment.
  • the screen 400 is a screen that the control unit 300 (or the control unit 106) displays to the user via the output device 104.
  • the screen 400 includes a selection area 401, a selection area 402, and an instruction area 403.
  • the user designates information shown on the screen 400 by operating the screen 400 displayed on the output device 104 with the input device 105.
  • the selection area 401 includes a button for selecting a method for extracting an initial method.
  • the control unit 300 or the control unit 106 automatically extracts the initial method in step 320.
  • the control unit 300 or the control unit 106 does not execute step 320 (or 140), and the initial method specified by the user is selected as the target method table 305 (or the target method table 113). To store.
  • the selection area 402 includes a check box for designating an initial method when “manual” is selected in the selection area 401.
  • methods called by programs are displayed in a hierarchical structure. Each method displayed in the selection area 402 can be selected by a check box.
  • the control unit 300 (or the control unit 106) starts the process shown in FIG. 19 (or FIG. 7) based on the information specified on the screen 400.
  • a method (maximum execution time method) that affects performance can be automatically extracted in one program. Further, since the method for measuring the execution time is extracted as in the first embodiment, the maximum execution time method can be accurately extracted regardless of the scale of the program.
  • an initial method can be designated by the user, a part of a program whose execution time is desired to be measured can be measured in a limited manner. For example, when a part of a program is modified and it is not necessary to measure the performance of all programs, the user designates a method that first executes a part of the program whose performance is to be measured as an initial method. As a result, performance can be measured in a short time, and power consumption can be reduced.
  • the execution time and the number of executions of the caller method can be used as the determination condition for the predetermined condition, for example, a ratio with respect to the execution time of the caller method is compared with a predetermined threshold.
  • the performance difference factor method or the maximum execution time method may be extracted.
  • the present invention is applicable to a system that measures the performance of a program in an application server.

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Hardware Design (AREA)
  • Quality & Reliability (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Debugging And Monitoring (AREA)

Abstract

A performance measurement device repeatedly executes a procedure of initially determining at least one first method among methods which are directly called by a program for which performance is to be measured, and by executing the program, measuring execution time of each of the first methods, and then, using the execution time of each of the measured first methods, assessing whether or not at least one second method matching a predetermined condition can be extracted from each of the first methods, and if at least one second method can be extracted, assessing whether or not at least one third method which is directly called from each of the extracted second methods can be extracted, and if at least one third method can be extracted, determining the extracted third method as a first method, measuring the execution time thereof, assessing whether or not a second method can be extracted, assessing whether or not a third method can be extracted, and determining the third method as a first method.

Description

性能測定方法、性能測定装置、及び、性能測定プログラムPerformance measurement method, performance measurement apparatus, and performance measurement program
　本発明は、性能測定方法に関し、特に、プログラムの性能を測定する性能測定方法に関する。 The present invention relates to a performance measurement method, and more particularly to a performance measurement method for measuring the performance of a program.
　複数のメソッドを含むプログラムを用いてシステムを稼動させる場合、期待した処理性能によってシステムが稼動するか否かを判定するため、プログラムの開発者等は、プログラムに特定の入力をし、それに対するシステムの処理性能を測定する。そして、測定の結果、システムが期待した処理性能によって稼動しない場合、開発者等には、プログラムのうち、処理性能に問題がある箇所を特定する必要が生じる。 When a system is operated using a program including a plurality of methods, in order to determine whether or not the system is operated according to expected processing performance, a program developer or the like inputs a specific input to the program, and the system corresponding thereto Measure the processing performance. As a result of the measurement, if the system does not operate according to the expected processing performance, the developer or the like needs to specify a part of the program that has a problem with the processing performance.
　従来用いられる一般的な方法には、あらかじめプログラムに埋め込まれた機能を用いて、実行中のプログラムにログを出力させ、出力されたログによって、プログラムの全体の処理の開始から終了までの実行時刻を取得する方法などがある。しかし、このような方法は、荒い精度でしか実行時間を取得できなかった。 A general method that has been used in the past is to use a function embedded in the program in advance to output a log to the program being executed, and the output time from the start to the end of the entire processing of the program based on the output log There is a way to get it. However, such a method can acquire the execution time only with rough accuracy.
　また、従来の方法には、ＯＳ又は仮想マシンが備えるプロファイリング機能を用いて、個々のメソッドの実行時間を測定するプロファイラと呼ばれるツールを用いる方法がある。この方法によれば、プロファイラを用いることによって、より詳細な精度に、プログラムの実行性能を分析することができる。そして、プログラムを最適化したり、処理性能に問題があった場合に問題を解決することができる。 Further, as a conventional method, there is a method of using a tool called a profiler that measures an execution time of each method by using a profiling function provided in an OS or a virtual machine. According to this method, the execution performance of the program can be analyzed with more detailed accuracy by using the profiler. The problem can be solved when the program is optimized or there is a problem in the processing performance.
　しかし、一般的に、メソッドの個数はプログラムの規模にほぼ比例して増え、メソッドが実行される回数もプログラム全体の実行時間にほぼ比例して増える。このため、大規模なプログラム、及び、起動時間が長いプログラム（例えば、サーバを稼動させるためのプログラムなど）の処理性能を測定する場合、プログラムに含まれるメソッドが大量になるため、プロファイラによる個々のメソッドの実行時間を記録することは、困難であった。 However, in general, the number of methods increases almost in proportion to the scale of the program, and the number of times the method is executed increases in proportion to the execution time of the entire program. For this reason, when measuring the processing performance of a large-scale program and a program with a long startup time (for example, a program for operating a server), the method included in the program becomes large. It was difficult to record the execution time of a method.
　また、性能問題が発生した場合におけるボトルネックとなるメソッドの特定など、性能を測定したいメソッドが特定できていない場合、実行される全てのメソッドの実行時間を測定し、測定された結果を記録しておく必要がある。しかし、一般的にメソッドは、非常に短い時間の間隔で頻繁に呼び出されるため、頻繁に発生する実行時間の測定及び記録自体が、プログラムが稼動するシステムのリソースを消費した。このため、測定自体が性能に影響を及ぼし、正確な測定が困難であった。 Also, if the method whose performance you want to measure cannot be identified, such as identifying the bottleneck when a performance problem occurs, measure the execution time of all the methods to be executed and record the measured results. It is necessary to keep. However, since methods are generally called frequently at very short time intervals, frequent measurement and recording of the execution time itself consumes the resources of the system on which the program runs. For this reason, the measurement itself has an effect on performance, and accurate measurement has been difficult.
　これに対し、従来、１回目の測定において、荒い精度によって全てのメソッドの実行時間を測定し、所定の閾値以上の実行時間がかかったメソッドなど所定の条件に該当するメソッドを主要メソッドとして抽出する方法が開示されている（例えば、特許文献１参照）。そして、特許文献１は、２回目の測定において、前記主要メソッドのみを対象として詳細な精度で実行時間を測定する技術を開示する。 On the other hand, in the first measurement, the execution time of all methods is measured with rough accuracy, and a method that satisfies a predetermined condition such as a method that takes an execution time exceeding a predetermined threshold is extracted as a main method. A method is disclosed (for example, see Patent Document 1). Patent Document 1 discloses a technique for measuring the execution time with detailed accuracy only for the main method in the second measurement.
特開２００８－２１７７２１号公報JP 2008-217721 A
　特許文献１に開示された技術は、特定の入力に対応する全てのメソッドを測定し、その統計情報（平均実行時間、実行回数）を生成する。このため、全体の実行時間が非常に短いプログラムを測定する場合、特許文献１に開示された技術は、正確な実行時間を取得できず、実際はボトルネックであるメソッドが、主要メソッドとして抽出されない場合がある。 The technique disclosed in Patent Document 1 measures all methods corresponding to a specific input and generates statistical information (average execution time, number of executions). For this reason, when measuring a program whose overall execution time is very short, the technique disclosed in Patent Document 1 cannot acquire an accurate execution time, and a method that is actually a bottleneck is not extracted as a main method. There is.
　さらに、特許文献１に開示された技術において用いられる所定の閾値は、どのメソッドにも一定値である。このため、例えば、性能に影響のある修正が加えられたプログラムと、修正される前のプログラムとの二つのプログラム間で性能の比較をする場合において、プログラム全体の実行時間と比較して実行時間に大きな差はないが、修正前と修正後とのメソッドの性能に著しい差が生じているメソッドを特定する、といった用途に、特許文献１に開示された技術は用いることができない。 Furthermore, the predetermined threshold value used in the technique disclosed in Patent Document 1 is a constant value for any method. For this reason, for example, when comparing performance between two programs, a program that has been modified that affects performance and a program that has not been modified, the execution time is compared with the execution time of the entire program. However, the technique disclosed in Patent Document 1 cannot be used for the purpose of identifying a method in which there is a significant difference in the performance of the method before and after the correction.
　本発明はこのような課題に鑑みてなされたものであり、処理性能に影響を与えるメソッドを精度よく特定し、かつ、処理性能を測定するプログラムの規模に関わらず、測定自体による処理性能への影響を低減する測定方法の提供を目的とする。 The present invention has been made in view of such problems, and can accurately specify a method that affects the processing performance, and can improve the processing performance by the measurement itself regardless of the scale of the program for measuring the processing performance. The purpose is to provide a measurement method that reduces the influence.
　また、本発明は、複数のプログラム間で性能に差が生じているメソッドを特定するための性能の測定方法の提供を目的とする。 Also, an object of the present invention is to provide a performance measurement method for specifying a method in which a difference in performance occurs between a plurality of programs.
　本発明の代表的な一例を示せば以下の通りである。すなわち、計算機によってプログラムの性能を測定する性能測定方法であって、前記計算機は、プロセッサと、前記プロセッサによって実行されるプログラムを格納するメモリとを有し、前記方法は、前記計算機が、前記性能が測定されるプログラムによって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを最初に決定する手順と、前記計算機が、前記プログラムを実行することによって、前記決定された各第１のメソッドの実行時間を測定する手順と、前記計算機が、前記測定された各第１のメソッドの実行時間を用いて、所定の条件に一致する少なくとも一つの第２のメソッドが前記各第１のメソッドから抽出可能か否かを判定する手順と、前記少なくとも一つの第２のメソッドが抽出可能である場合、前記計算機が、前記抽出された各第２のメソッドから直接呼ばれる少なくとも一つの第３のメソッドが抽出可能か否かを判定する手順と、前記少なくとも一つの第３のメソッドが抽出可能である場合、前記計算機が、前記抽出された第３のメソッドを、第１のメソッドに決定する手順と、前記計算機が、前記実行時間を測定する手順と、前記第２のメソッドが抽出可能か否かを判定する手順と、前記第３のメソッドが抽出可能か否かを判定する手順と、前記第３のメソッドを第１のメソッドに決定する手順とを繰り返し実行する手順とを含む。 A typical example of the present invention is as follows. That is, a performance measurement method for measuring the performance of a program by a computer, wherein the computer includes a processor and a memory for storing a program executed by the processor, and the method includes: A procedure for first determining at least one first method among methods directly called by a program to be measured, and execution of each determined first method by the computer executing the program Using the time measurement procedure and the execution time of each measured first method, the computer can extract at least one second method that matches a predetermined condition from each first method. If the procedure for determining whether or not the at least one second method can be extracted, A procedure for determining whether or not at least one third method called directly from each extracted second method can be extracted, and when the at least one third method can be extracted, A procedure for determining the extracted third method as a first method; a procedure for the computer to measure the execution time; and a procedure for determining whether or not the second method can be extracted; And a procedure for repeatedly executing a procedure for determining whether or not the third method can be extracted and a procedure for determining the third method as the first method.
　本発明の一実施形態によると、プログラムにおいて処理性能に影響を与えているメソッドを精度よく特定することができる。 According to an embodiment of the present invention, it is possible to accurately identify a method that affects processing performance in a program.
本発明の第１の実施形態の複数のプログラムの性能差を分析する計算機の構成を示すブロック図である。It is a block diagram which shows the structure of the computer which analyzes the performance difference of the some program of the 1st Embodiment of this invention. 本発明の第１の実施形態の対象メソッドテーブルを示す説明図である。It is explanatory drawing which shows the object method table of the 1st Embodiment of this invention. 本発明の第１の実施形態の第１メソッド呼出テーブルを示す説明図である。It is explanatory drawing which shows the 1st method call table of the 1st Embodiment of this invention. 本発明の第１の実施形態の第２メソッド呼出テーブルを示す説明図である。It is explanatory drawing which shows the 2nd method call table of the 1st Embodiment of this invention. 本発明の第１の実施形態の第１実行情報テーブルを示す説明図である。It is explanatory drawing which shows the 1st execution information table of the 1st Embodiment of this invention. 本発明の第１の実施形態の第２実行情報テーブルを示す説明図である。It is explanatory drawing which shows the 2nd execution information table of the 1st Embodiment of this invention. 本発明の第１の実施形態の測定結果テーブルを示す説明図である。It is explanatory drawing which shows the measurement result table of the 1st Embodiment of this invention. 本発明の第１の実施形態の測定結果を表示した場合の画面を示す説明図である。It is explanatory drawing which shows the screen at the time of displaying the measurement result of the 1st Embodiment of this invention. 本発明の第１の実施形態の全体の処理を示すフローチャートである。It is a flowchart which shows the whole process of the 1st Embodiment of this invention. 本発明の第１の実施形態の初期メソッドを示すシーケンス図である。It is a sequence diagram which shows the initial method of the 1st Embodiment of this invention. 本発明の第１の実施形態の初期メソッドを抽出する処理におけるデータの流れを示すブロック図である。It is a block diagram which shows the data flow in the process which extracts the initial method of the 1st Embodiment of this invention. 本発明の第１の実施形態の初期メソッドを抽出するためにテストを実行する処理を示すフローチャートである。It is a flowchart which shows the process which performs a test in order to extract the initial method of the 1st Embodiment of this invention. 本発明の第１の実施形態の測定対象メソッドを抽出する処理を示すフローチャートである。It is a flowchart which shows the process which extracts the measuring object method of the 1st Embodiment of this invention. 本発明の第１の実施形態の初期メソッドを抽出する処理を示すフローチャートである。It is a flowchart which shows the process which extracts the initial method of the 1st Embodiment of this invention. 本発明の第１の実施形態の測定対象メソッドの実行時間を測定する処理におけるデータの流れを示すブロック図である。It is a block diagram which shows the data flow in the process which measures the execution time of the measuring object method of the 1st Embodiment of this invention. 本発明の第１の実施形態の実行時間を測定するためにテストを実行する処理を示すフローチャートである。It is a flowchart which shows the process which performs a test in order to measure the execution time of the 1st Embodiment of this invention. 本発明の第１の実施形態の測定対象メソッドの実行時間を測定する処理を示すフローチャートである。It is a flowchart which shows the process which measures the execution time of the measuring object method of the 1st Embodiment of this invention. 本発明の第１の実施形態の性能差要因メソッドを抽出する処理を示すフローチャートである。It is a flowchart which shows the process which extracts the performance difference factor method of the 1st Embodiment of this invention. 本発明の第１の実施形態の測定対象メソッドを再度設定する処理を示すフローチャートである。It is a flowchart which shows the process which sets the measuring object method of the 1st Embodiment of this invention again. 本発明の第２の実施形態の単一のプログラムの性能差を分析する計算機の構成を示すブロック図である。It is a block diagram which shows the structure of the computer which analyzes the performance difference of the single program of the 2nd Embodiment of this invention. 本発明の第２の実施形態のメソッド呼出テーブルを示す説明図である。It is explanatory drawing which shows the method call table of the 2nd Embodiment of this invention. 本発明の第２の実施形態の実行情報テーブルを示す説明図である。It is explanatory drawing which shows the execution information table of the 2nd Embodiment of this invention. 本発明の第２の実施形態の測定結果テーブルを示す説明図である。It is explanatory drawing which shows the measurement result table of the 2nd Embodiment of this invention. 本発明の第２の実施形態の全体の処理を示すフローチャートである。It is a flowchart which shows the whole process of the 2nd Embodiment of this invention. 本発明の第２の実施形態の所定の条件を満たすメソッドを抽出する処理を示すフローチャートである。It is a flowchart which shows the process which extracts the method which satisfy | fills the predetermined condition of the 2nd Embodiment of this invention. 本発明の第２の実施形態の初期メソッドを指定するための画面を示す説明図である。It is explanatory drawing which shows the screen for designating the initial method of the 2nd Embodiment of this invention.
　以下、本発明について、図面を参照して説明する。 Hereinafter, the present invention will be described with reference to the drawings.
　図１は、本発明の第１の実施形態の複数のプログラムの性能差を分析する計算機１００の構成を示すブロック図である。 FIG. 1 is a block diagram illustrating a configuration of a computer 100 that analyzes a performance difference between a plurality of programs according to the first embodiment of this invention.
　第１の実施形態の計算機１００は、演算装置１０１、メモリ１０２、外部記憶装置１０３、出力装置１０４、及び入力装置１０５を備える。計算機１００は、複数のプログラムの処理性能の差を分析する機能を備える装置である。 The computer 100 according to the first embodiment includes an arithmetic device 101, a memory 102, an external storage device 103, an output device 104, and an input device 105. The computer 100 is a device having a function of analyzing a difference in processing performance of a plurality of programs.
　演算装置１０１は、ＣＰＵなどのプロセッサである。演算装置１０１は、メモリ１０２においてプログラムを実行する。 The arithmetic unit 101 is a processor such as a CPU. The arithmetic device 101 executes a program in the memory 102.
　メモリ１０２は、一時的にデータを記憶するための領域を備える記憶装置である。メモリ１０２は、プログラム及びテーブルを保持する。 The memory 102 is a storage device having an area for temporarily storing data. The memory 102 holds programs and tables.
　外部記憶装置１０３は、ハードディスクドライブなどの記憶装置である。演算装置１０１によって実行されるプログラムに入力される引数等を保持する。 The external storage device 103 is a storage device such as a hard disk drive. An argument or the like input to a program executed by the arithmetic device 101 is held.
　出力装置１０４は、ディスプレイ又はプリンタ等の装置である。計算機１００による処理の結果を、ユーザに提供するための装置である。入力装置１０５は、キーボード又はマウス等の装置である。ユーザは、演算装置１０１によって実行されるプログラムに入力される引数等を、入力装置１０５を介して計算機１００に入力する。 The output device 104 is a device such as a display or a printer. This is an apparatus for providing a result of processing by the computer 100 to a user. The input device 105 is a device such as a keyboard or a mouse. A user inputs an argument or the like input to a program executed by the arithmetic device 101 to the computer 100 via the input device 105.
　メモリ１０２は、制御部１０６、テスト実行部１０７、メソッド呼出照合部１０８、性能差要因メソッド判定部１０９、バージョン切替部１１０、第１メソッド呼出テーブル１１１、第２メソッド呼出テーブル１１２、対象メソッドテーブル１１３、第１実行情報テーブル１１４、第２実行情報テーブル１１５、測定結果テーブル１１６、初期メソッド記録部１１７、対象メソッド記録部１１８、コールバック受信部１１９、第１プログラム１２０、第２プログラム１２１、ＪａｖａＶＭ１２２（Ｊａｖａは登録商標、以下同じ）、及び、ＪＶＭＴＩ１２３を保持する。 The memory 102 includes a control unit 106, a test execution unit 107, a method call verification unit 108, a performance difference factor method determination unit 109, a version switching unit 110, a first method call table 111, a second method call table 112, and a target method table 113. , First execution information table 114, second execution information table 115, measurement result table 116, initial method recording unit 117, target method recording unit 118, callback receiving unit 119, first program 120, second program 121, Java VM 122 ( Java is a registered trademark, and the same shall apply hereinafter) and JVMTI123.
　制御部１０６、テスト実行部１０７、メソッド呼出照合部１０８、性能差要因メソッド判定部１０９、バージョン切替部１１０、初期メソッド記録部１１７、対象メソッド記録部１１８、コールバック受信部１１９、第１プログラム１２０、第２プログラム１２１、ＪａｖａＶＭ１２２、及び、ＪＶＭＴＩ１２３は、プログラムである。第１メソッド呼出テーブル１１１、第２メソッド呼出テーブル１１２、対象メソッドテーブル１１３、第１実行情報テーブル１１４、第２実行情報テーブル１１５、及び、測定結果テーブル１１６は、データが格納されるテーブルである。 Control unit 106, test execution unit 107, method call verification unit 108, performance difference factor method determination unit 109, version switching unit 110, initial method recording unit 117, target method recording unit 118, callback reception unit 119, first program 120 The second program 121, the Java VM 122, and the JVMTI 123 are programs. The first method call table 111, the second method call table 112, the target method table 113, the first execution information table 114, the second execution information table 115, and the measurement result table 116 are tables in which data is stored.
　制御部１０６は、バージョン切替部１１０に第１プログラム１２０又は第２プログラム１２１を実行可能な状態に切り替えさせ、テスト実行部１０７に第１プログラム１２０又は第２プログラム１２１を実行（テスト）させる。また、第１プログラム１２０又は第２プログラム１２１を実行させることによって取得された測定結果の集計を同期させるなど、メモリ１０２において実行されるプログラムを制御する。 The control unit 106 causes the version switching unit 110 to switch to a state where the first program 120 or the second program 121 can be executed, and causes the test execution unit 107 to execute (test) the first program 120 or the second program 121. In addition, the program executed in the memory 102 is controlled, for example, the measurement results obtained by executing the first program 120 or the second program 121 are synchronized.
　テスト実行部１０７は、第１プログラム１２０又は第２プログラム１２１のうち実行可能な状態になっているプログラムに、あらかじめユーザによって指定された所定の値を入力し、第１プログラム１２０又は第２プログラム１２１を実行する。そして、これによって、性能を測定するために必要な負荷をかけて、第１プログラム１２０又は第２プログラム１２１を実行する。 The test execution unit 107 inputs a predetermined value designated in advance by the user to a program that is in an executable state of the first program 120 or the second program 121, and the first program 120 or the second program 121. Execute. Then, the first program 120 or the second program 121 is executed by applying a load necessary for measuring the performance.
　メソッド呼出照合部１０８は、初期メソッド又は性能差要因メソッドによって直接呼ばれたメソッドのうち、所定の条件に一致するメソッドを次の測定対象に設定するため、所定の条件に一致するメソッドを示す情報を対象メソッドテーブル１１３に格納する。 The method call matching unit 108 sets information that indicates a method that matches a predetermined condition in order to set a method that matches a predetermined condition as a next measurement target among methods called directly by the initial method or the performance difference factor method. Are stored in the target method table 113.
　なお、第１の実施形態において初期メソッドとは、第１プログラム１２０又は第２プログラム１２１をテストしている期間中、各スレッドにおいて最初に呼ばれるメソッドである。また、第１の実施形態において性能差要因メソッドとは、第１プログラム１２０と第２プログラム１２１との性能に差があった場合、最も性能の差を発生させる要因となるメソッドである。 In the first embodiment, the initial method is a method that is called first in each thread during the period when the first program 120 or the second program 121 is being tested. In the first embodiment, the performance difference factor method is a method that causes the difference in performance most when there is a difference in performance between the first program 120 and the second program 121.
　性能差要因メソッド判定部１０９は、第１実行情報テーブル１１４及び第２実行情報テーブル１１５に格納されたメソッドの開始時刻とメソッドの終了時刻との差分に基づいて、性能を示す値を算出する。そして、測定結果テーブル１１６に算出された値を格納する。 The performance difference factor method determination unit 109 calculates a value indicating performance based on the difference between the method start time and the method end time stored in the first execution information table 114 and the second execution information table 115. Then, the calculated value is stored in the measurement result table 116.
　具体的には、性能差要因メソッド判定部１０９は、第１プログラム１２０及び第２プログラム１２１における測定対象メソッドの平均実行時間と実行回数とを、第１実行情報テーブル１１４及び第２実行情報テーブル１１５から抽出する。そして、抽出された値に基づいて、第１プログラム１２０と第２プログラム１２１との実行時間差と実行時間比とを算出する。そして、算出された値を測定結果テーブル１１６に格納する。さらに、測定結果テーブル１１６に出力された実行時間差と実行時間比とに基づいて、性能差要因メソッドを抽出する。 Specifically, the performance difference factor method determination unit 109 calculates the average execution time and the number of executions of the measurement target method in the first program 120 and the second program 121 as the first execution information table 114 and the second execution information table 115. Extract from Then, based on the extracted value, an execution time difference and an execution time ratio between the first program 120 and the second program 121 are calculated. The calculated value is stored in the measurement result table 116. Further, a performance difference factor method is extracted based on the execution time difference and the execution time ratio output to the measurement result table 116.
　ここで、測定対象メソッドとは、本実施形態において性能を測定する対象のメソッドである。 Here, the measurement target method is a method whose performance is measured in the present embodiment.
　バージョン切替部１１０は、第１プログラム１２０及び第２プログラム１２１のうち、制御部１０６から指定されたプログラムを起動し、第１プログラム１２０又は第２プログラム１２１を実行可能な状態にする。 The version switching unit 110 activates the program designated by the control unit 106 out of the first program 120 and the second program 121 to make the first program 120 or the second program 121 executable.
　第１プログラム１２０及び第２プログラム１２１は、第１の実施形態の計算機１００によって、性能を測定されるプログラムである。第１の実施形態において、第１プログラム１２０はユーザによる修正前のプログラムであり、第２プログラム１２１はユーザによる修正後のプログラムである。 The first program 120 and the second program 121 are programs whose performance is measured by the computer 100 according to the first embodiment. In the first embodiment, the first program 120 is a program before correction by the user, and the second program 121 is a program after correction by the user.
　本実施形態の第１プログラム１２０及び第２プログラム１２１は、Ｊａｖａ等によって作成されたプログラムであり、ＪａｖａＶＭ（仮想マシン上）１２２において実行される。ＪａｖａＶＭ１２２は、第１プログラム１２０及び第２プログラム１２１によって呼ばれるメソッドが、開始された時及び終了した時において、コールバックを発行するためのプロファイル機能を備える。ＪＶＭＴＩ１２３は、本実施形態のＪａｖａＶＭ１２２に備わり、コールバックを発行するためのプロファイル機能である。 The first program 120 and the second program 121 of this embodiment are programs created by Java or the like, and are executed in the Java VM (on the virtual machine) 122. The Java VM 122 has a profile function for issuing a callback when a method called by the first program 120 and the second program 121 is started and ended. The JVMTI 123 is a profile function for issuing a callback provided in the Java VM 122 of the present embodiment.
　しかし、本発明の第１プログラム１２０及び第２プログラム１２１は、Ｊａｖａ言語によって作成されたプログラムに限られるものではなく、第１プログラム１２０及び第２プログラム１２１によって呼ばれるメソッド（又は、メソッドに相当するサブルーチン）が開始された時及び終了した時において、コールバック（又は、コールバックに相当する信号）を発行する手段を含む言語であれば、いかなる言語によって作成されてもよい。そして、ＪａｖａＶＭ１２２及びＪＶＭＴＩ１２３は、第１プログラム１２０及び第２プログラム１２１の言語に対応した、第１プログラム１２０及び第２プログラム１２１を実行するための環境であれば、いかなる環境でもよい。 However, the first program 120 and the second program 121 of the present invention are not limited to programs created in the Java language, but methods (or subroutines corresponding to methods) called by the first program 120 and the second program 121. ) May be created in any language as long as it includes a means for issuing a callback (or a signal corresponding to the callback) at the time of starting and ending. The Java VM 122 and the JVMTI 123 may be any environment as long as it is an environment for executing the first program 120 and the second program 121 corresponding to the languages of the first program 120 and the second program 121.
　コールバック受信部１１９は、ＪＶＭＴＩ１２３によって発行されるコールバックを受信し、コールバックが受信された時刻を、計算機１００が備える高精度タイマを用いて生成する。そして、生成された時刻と、コールバックに含まれるメソッドを示す情報（クラス名、メソッド名、メソッド識別子、呼び出し元のメソッドを示す情報、及び、コールスタックのフレーム番号など）と、メソッドが開始されたか終了したかを示す情報とを、初期メソッド記録部１１７又は対象メソッド記録部１１８に送信する。 The callback receiving unit 119 receives a callback issued by the JVMTI 123, and generates a time when the callback is received using a high-precision timer provided in the computer 100. Then, the generated time, information indicating the method included in the callback (class name, method name, method identifier, information indicating the calling method, frame number of the call stack, etc.) and the method are started. Information indicating whether or not the process has been completed is transmitted to the initial method recording unit 117 or the target method recording unit 118.
　初期メソッド記録部１１７は、テスト実行部１０７によって第１プログラム１２０及び第２プログラム１２１に値が入力されている間（すなわち、第１プログラム１２０又は第２プログラム１２１がテストされている間）に、コールスタックに格納された最も上位のメソッドを抽出する。そして、抽出されたメソッドを示す情報を、各々第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２に格納する。 The initial method recording unit 117 is in a state where values are input to the first program 120 and the second program 121 by the test execution unit 107 (that is, while the first program 120 or the second program 121 is being tested). Extracts the highest method stored in the call stack. Then, information indicating the extracted method is stored in the first method call table 111 and the second method call table 112, respectively.
　本実施形態において、第１プログラム１２０又は第２プログラム１２１に呼ばれたメソッドは、スレッド毎のコールスタックに格納され、開始される。そして、最も遅くコールスタックに格納され（ＰＵＳＨ）たメソッドは、最も早くコールスタックから取得される（ＰＯＰ）。すなわち、最も遅く開始されたメソッドが、最も早く終了する。 In this embodiment, the method called by the first program 120 or the second program 121 is stored in the call stack for each thread and started. Then, the latest method stored in the call stack (PUSH) is acquired from the call stack earliest (POP). That is, the latest started method finishes earliest.
　本実施形態において、コールスタックに最も早く格納されたメソッドは、最も上位のメソッドと記載され、コールスタックに最も上位のメソッドより遅く格納されたメソッドは、下位のメソッドと記載される。 In this embodiment, the method stored earliest in the call stack is described as the highest method, and the method stored later than the highest method in the call stack is described as a lower method.
　対象メソッド記録部１１８は、テスト実行部１０７によって第１プログラム１２０又は第２プログラム１２１に値が入力されている間（すなわち、第１プログラム１２０又は第２プログラム１２１がテストされている間）に、対象メソッドテーブル１１３に格納されている測定対象メソッドが開始された時刻及び終了した時刻と、メソッドを示す情報とを、コールバック受信部１１９から受信する。そして、コールバック受信部１１９から受信された時刻等の情報を、第１実行情報テーブル１１４又は第２実行情報テーブル１１５に格納する。 The target method recording unit 118 is during the time when a value is input to the first program 120 or the second program 121 by the test execution unit 107 (that is, while the first program 120 or the second program 121 is being tested). The start time and end time of the measurement target method stored in the target method table 113 and information indicating the method are received from the callback receiving unit 119. Information such as the time received from the callback receiver 119 is stored in the first execution information table 114 or the second execution information table 115.
　また、対象メソッド記録部１１８は、第１プログラム１２０又は第２プログラム１２１を測定している間に、測定対象メソッドによって直接呼ばれたメソッドを示す情報を抽出する。そして、抽出されたメソッドを示す情報を、第１メソッド呼出テーブル１１１又は第２メソッド呼出テーブル１１２に格納する。 Further, the target method recording unit 118 extracts information indicating a method directly called by the measurement target method while measuring the first program 120 or the second program 121. Then, information indicating the extracted method is stored in the first method call table 111 or the second method call table 112.
　第１メソッド呼出テーブル１１１は、第１プログラム１２０によって呼ばれたメソッドを示す情報が格納されるテーブルである。第２メソッド呼出テーブル１１２は、第２プログラム１２１によって呼ばれたメソッドを示す情報が格納されるテーブルである。対象メソッドテーブル１１３は、第１プログラム１２０及び第２プログラム１２１によって呼ばれたメソッドのうち、第１プログラム１２０と第２プログラム１２１との間で共通のメソッドを示す情報が格納されるテーブルである。 The first method call table 111 is a table in which information indicating a method called by the first program 120 is stored. The second method call table 112 is a table in which information indicating a method called by the second program 121 is stored. The target method table 113 is a table in which information indicating methods common to the first program 120 and the second program 121 among methods called by the first program 120 and the second program 121 is stored.
　第１実行情報テーブル１１４は、第１プログラム１２０によって呼ばれたメソッドの開始時刻と終了時刻とが格納されるテーブルである。第２実行情報テーブル１１５は、第２プログラム１２１によって呼ばれたメソッドの開始時刻と終了時刻とが格納されるテーブルである。測定結果テーブル１１６は、第１実行情報テーブル１１４及び第２実行情報テーブル１１５に格納される各メソッドの開始時刻と終了時刻とに基づいて算出された、第１プログラム１２０と第２プログラム１２１との性能を比較するための値が格納されるテーブルである。 The first execution information table 114 is a table in which the start time and end time of the method called by the first program 120 are stored. The second execution information table 115 is a table in which the start time and end time of the method called by the second program 121 are stored. The measurement result table 116 includes the first program 120 and the second program 121 calculated based on the start time and end time of each method stored in the first execution information table 114 and the second execution information table 115. It is a table in which the value for comparing performance is stored.
　図２は、本発明の第１の実施形態の対象メソッドテーブル１１３を示す説明図である。 FIG. 2 is an explanatory diagram illustrating the target method table 113 according to the first embodiment of this invention.
　対象メソッドテーブル１１３は、開始時刻と終了時刻とが測定されるメソッドを示す情報が格納される。対象メソッドテーブル１１３に格納される値は、第１プログラム１２０又は第２プログラム１２１によって呼ばれたメソッドが、さらに別のメソッドを呼ぶ場合に更新される。すなわち、測定対象メソッドによって呼ばれたメソッドを、さらに次の測定対象メソッドとするため、対象メソッドテーブル１１３は更新される。 The target method table 113 stores information indicating a method whose start time and end time are measured. The value stored in the target method table 113 is updated when a method called by the first program 120 or the second program 121 calls another method. That is, the target method table 113 is updated in order to set the method called by the measurement target method as the next measurement target method.
　対象メソッドテーブル１１３は、フレーム番号１１３１、メソッド情報１１３２、及び、呼出元メソッド情報１１３３を含む。フレーム番号１１３１は、コールスタックに格納されるメソッドの順番を示す数字である。フレーム番号１１３１の値が小さい程、コールスタックにおいて上位のメソッドであることを示す。 The target method table 113 includes a frame number 1131, method information 1132, and caller method information 1133. The frame number 1131 is a number indicating the order of methods stored in the call stack. A smaller frame number 1131 indicates a higher method in the call stack.
　メソッド情報１１３２は、測定対象メソッドを示す情報が含まれる。メソッド情報１１３２に格納される値は、測定対象メソッドのクラス名とメソッド名とを含み、測定対象メソッドを一意に示す識別子である。 The method information 1132 includes information indicating the measurement target method. The value stored in the method information 1132 is an identifier that uniquely indicates the measurement target method, including the class name and method name of the measurement target method.
　呼出元メソッド情報１１３３は、メソッド情報１１３２に示す測定対象メソッドを呼んだメソッドを示す情報を含む。呼出元メソッド情報１１３３に格納される値は、メソッド情報１１３２に格納される値と同じ表記方法である。すなわち、呼出元メソッド情報１１３３に格納される値は、呼出元メソッドのクラス名とメソッド名とを含み、呼出元メソッドを一意に示す識別子である。 The caller method information 1133 includes information indicating a method that called the measurement target method indicated in the method information 1132. The value stored in the caller method information 1133 has the same notation method as the value stored in the method information 1132. That is, the value stored in the caller method information 1133 is an identifier that uniquely indicates the caller method, including the class name and method name of the caller method.
　図３Ａは、本発明の第１の実施形態の第１メソッド呼出テーブル１１１を示す説明図である。 FIG. 3A is an explanatory diagram illustrating the first method call table 111 according to the first embodiment of this invention.
　第１メソッド呼出テーブル１１１は、第１プログラム１２０が測定される際に、第１プログラム１２０によって呼ばれるメソッドを示す情報が格納される。第１メソッド呼出テーブル１１１は、フレーム番号１１１１、メソッド情報１１１２、及び、呼出元メソッド情報１１１３を含む。フレーム番号１１１１、メソッド情報１１１２、及び、呼出元メソッド情報１１１３は、対象メソッドテーブル１１３のフレーム番号１１３１、メソッド情報１１３２、及び、呼出元メソッド情報１１３３に相当する。 The first method call table 111 stores information indicating a method called by the first program 120 when the first program 120 is measured. The first method call table 111 includes a frame number 1111, method information 1112, and call source method information 1113. The frame number 1111, the method information 1112, and the call source method information 1113 correspond to the frame number 1131, the method information 1132, and the call source method information 1133 in the target method table 113.
　第１プログラム１２０に呼ばれた初期メソッドの開始時刻及び終了時刻を測定されている間、第１メソッド呼出テーブル１１１のメソッド情報１１１２には、初期メソッドによって直接呼ばれるメソッドを示す情報が、対象メソッド記録部１１８によって蓄積される。また、初期メソッドよりも下位の測定対象メソッドの開始時刻及び終了時刻が測定されている間、第１メソッド呼出テーブル１１１のメソッド情報１１１２には、測定対象メソッドによって直接呼ばれるメソッドを示す情報が、対象メソッド記録部１１８によって蓄積される。 While the start time and end time of the initial method called by the first program 120 are being measured, the method information 1112 of the first method call table 111 includes information indicating the method directly called by the initial method. Accumulated by the unit 118. Further, while the start time and end time of the measurement target method lower than the initial method are being measured, the method information 1112 of the first method call table 111 includes information indicating a method directly called by the measurement target method. Accumulated by the method recording unit 118.
　図３Ｂは、本発明の第１の実施形態の第２メソッド呼出テーブル１１２を示す説明図である。 FIG. 3B is an explanatory diagram illustrating the second method call table 112 according to the first embodiment of this invention.
　第２メソッド呼出テーブル１１２は、第２プログラム１２１が測定される際に、第２プログラム１２１によって呼ばれるメソッドを示す情報が格納される。第２メソッド呼出テーブル１１２は、フレーム番号１１２１、メソッド情報１１２２、及び、呼出元メソッド情報１１２３を含む。フレーム番号１１２１、メソッド情報１１２２、及び、呼出元メソッド情報１１２３は、対象メソッドテーブル１１３のフレーム番号１１３１、メソッド情報１１３２、及び、呼出元メソッド情報１１３３に相当する。 The second method call table 112 stores information indicating a method called by the second program 121 when the second program 121 is measured. The second method call table 112 includes a frame number 1121, method information 1122, and call source method information 1123. The frame number 1121, method information 1122, and call source method information 1123 correspond to the frame number 1131, method information 1132, and call source method information 1133 in the target method table 113.
　第２メソッド呼出テーブル１１２に格納される情報は、第１メソッド呼出テーブル１１１に格納される情報と同様であり、第２プログラム１２１が実行されることによって抽出された値が格納される。 The information stored in the second method call table 112 is the same as the information stored in the first method call table 111, and a value extracted by executing the second program 121 is stored.
　図４Ａは、本発明の第１の実施形態の第１実行情報テーブル１１４を示す説明図である。 FIG. 4A is an explanatory diagram illustrating the first execution information table 114 according to the first embodiment of this invention.
　第１実行情報テーブル１１４は、第１プログラム１２０によって呼ばれたメソッドの、開始時刻又は終了時刻を格納するテーブルである。第１実行情報テーブル１１４は、フレーム番号１１４１、メソッド情報１１４２、開始終了情報１１４３、及び時刻１１４４を含む。フレーム番号１１４１及びメソッド情報１１４２は、対象メソッドテーブル１１３のフレーム番号１１３１及びメソッド情報１１３２に相当する。 The first execution information table 114 is a table that stores a start time or an end time of a method called by the first program 120. The first execution information table 114 includes a frame number 1141, method information 1142, start / end information 1143, and time 1144. The frame number 1141 and the method information 1142 correspond to the frame number 1131 and the method information 1132 of the target method table 113.
　開始終了情報１１４３に格納される値は、各行に含まれる時刻１１４４が、メソッド情報１１４２が示すメソッドの開始時刻を示す場合"開始"であり、各行に含まれる時刻１１４４が、メソッド情報１１３２が示すメソッドの終了時刻を示す場合"終了"である。 The value stored in the start / end information 1143 is “start” when the time 1144 included in each row indicates the start time of the method indicated by the method information 1142, and the time 1144 included in each row is indicated by the method information 1132. “End” indicates the end time of the method.
　時刻１１４４は、メソッド情報１１４２が示すメソッドの開始時刻又は終了時刻を示す。図４Ａに示す時刻１１４４は、計算機１００において用いられるシステム時刻によって表記されるが、いかなる表記方法で時刻を示してもよい。 The time 1144 indicates the start time or end time of the method indicated by the method information 1142. Although the time 1144 shown in FIG. 4A is expressed by the system time used in the computer 100, the time may be indicated by any notation method.
　図４Ｂは、本発明の第１の実施形態の第２実行情報テーブル１１５を示す説明図である。 FIG. 4B is an explanatory diagram illustrating the second execution information table 115 according to the first embodiment of this invention.
　第２実行情報テーブル１１５は、第２プログラム１２１によって呼ばれたメソッドの、開始時刻又は終了時刻を格納するテーブルである。第２実行情報テーブル１１５に格納される情報は、第１実行情報テーブル１１４と同様の情報であり、第２プログラム１２１が実行されることによって抽出される値が格納される。 The second execution information table 115 is a table for storing the start time or end time of the method called by the second program 121. The information stored in the second execution information table 115 is the same information as the first execution information table 114, and stores the value extracted by executing the second program 121.
　前述の第１メソッド呼出テーブル１１１、第２メソッド呼出テーブル１１２、対象メソッドテーブル１１３、第１実行情報テーブル１１４及び第２実行情報テーブル１１５において、計算機１００による測定の対象とするメソッドが、次の測定対象メソッドに移行する際、各テーブルが更新され、各テーブルのフレーム番号が更新されてよい。すなわち、第１メソッド呼出テーブル１１１、第２メソッド呼出テーブル１１２、対象メソッドテーブル１１３、第１実行情報テーブル１１４及び第２実行情報テーブル１１５に含まれるフレーム番号は、各測定の時点においてすべて同じ値を示してよい。 In the first method call table 111, the second method call table 112, the target method table 113, the first execution information table 114, and the second execution information table 115, the method to be measured by the computer 100 is the next measurement. When shifting to the target method, each table may be updated, and the frame number of each table may be updated. That is, the frame numbers included in the first method call table 111, the second method call table 112, the target method table 113, the first execution information table 114, and the second execution information table 115 all have the same value at the time of each measurement. May show.
　ただし、測定対象メソッドが識別できれば、各テーブルはいずれの方法によって値を格納してもよく、例えば、測定対象メソッドを示すフレーム番号を変数によって保持し、変数が示すフレーム番号の測定対象メソッドを、計算機１００に測定させるなどの方法を用いてもよい。 However, as long as the measurement target method can be identified, each table may store the value by any method.For example, the frame number indicating the measurement target method is held by a variable, and the measurement target method having the frame number indicated by the variable is A method of causing the computer 100 to measure may be used.
　図５は、本発明の第１の実施形態の測定結果テーブル１１６を示す説明図である。 FIG. 5 is an explanatory diagram illustrating the measurement result table 116 according to the first embodiment of this invention.
　測定結果テーブル１１６は、第１実行情報テーブル１１４及び第２実行情報テーブル１１５に含まれる開始時刻と終了時刻とに基づいて算出された値をメソッドごとに格納するテーブルである。 The measurement result table 116 is a table that stores, for each method, values calculated based on the start time and end time included in the first execution information table 114 and the second execution information table 115.
　測定結果テーブル１１６は、フレーム番号１１６１、メソッド情報１１６２、第１プログラムの平均実行時間１１６３１及び実行回数１１６３２、第２プログラムの平均実行時間１１６４１及び実行回数１１６４２、実行時間差１１６５、並びに、実行時間比１１６６を含む。 The measurement result table 116 includes frame number 1161, method information 1162, average execution time 11631 and execution count 11632 of the first program, average execution time 11641 and execution count 11642 of the second program, execution time difference 1165, and execution time ratio 1166. including.
　フレーム番号１１６１は、コールスタックに格納されるメソッドの順番を示す数字である。測定結果テーブル１１６の行は、メソッドが測定される都度追加されるため、フレーム番号１１６１は、少なくとも一つの値を示す。 The frame number 1161 is a number indicating the order of methods stored in the call stack. Since a row of the measurement result table 116 is added every time a method is measured, the frame number 1161 indicates at least one value.
　メソッド情報１１６２は、測定されたメソッドを示す情報を格納する。 The method information 1162 stores information indicating the measured method.
　第１プログラムの平均実行時間１１６３１は、各メソッドが呼出元メソッドから複数回呼ばれる場合において、複数回呼ばれた各メソッドの実行時間の平均値である。 The average execution time 11631 of the first program is an average value of the execution times of each method called multiple times when each method is called multiple times from the calling method.
　例えば、第１実行情報テーブル１１４のメソッド情報１１４２に含まれるメソッドを抽出し、抽出されたメソッドの各々の終了時刻から開始時刻を減算することによって、１回目の実行時間が算出される。そして、抽出されたメソッドと同じメソッドを示すメソッドを、さらにメソッド情報１１４２から抽出し、さらに抽出されたメソッドの実行時間を算出することによって２回目の実行時間が算出される。このように算出された複数回の実行時間の平均値を算出することによって、平均実行時間１１６３１が算出される。 For example, the first execution time is calculated by extracting the method included in the method information 1142 of the first execution information table 114 and subtracting the start time from the end time of each extracted method. Then, a method indicating the same method as the extracted method is further extracted from the method information 1142, and the second execution time is calculated by calculating the execution time of the extracted method. The average execution time 11631 is calculated by calculating the average value of the execution times calculated in this way.
　第１プログラムの実行回数１１６３２は、呼出元メソッドから呼ばれた回数を示す。第１実行情報テーブル１１４のうち、開始終了情報１１５３に開始及び終了の両方を含むメソッドが、第１実行情報テーブル１１４にいくつ含まれるかを算出することによって、求められる。 The number of executions 11632 of the first program indicates the number of calls from the caller method. In the first execution information table 114, the number of methods including both start and end in the start / end information 1153 is obtained by calculating how many methods are included in the first execution information table 114.
　第２プログラムの平均実行時間１１６４１及び第２プログラムの実行回数１１６４２には、第１プログラムの平均実行時間１１６３１及び実行回数１１６３２と同様な値が格納され、第２実行情報テーブル１１５に基づいて算出された値が格納される。 The average execution time 11641 of the second program and the execution count 11642 of the second program store the same values as the average execution time 11631 and the execution count 11632 of the first program, and are calculated based on the second execution information table 115. Stored value.
　実行時間差１１６５は、呼出元メソッドが１回実行される毎の、メソッド情報１１６２が示すメソッドの実行時間の合計を、第１プログラム１２０と第２プログラム１２１とにおいて算出し、算出された各結果の差を示す。具体的には、第１プログラム１２０及び第２プログラム１２１の結果に基づいて、各々、メソッド情報１１６２が示すメソッドが実行された実行時間の合計を算出する。そして、実行時間の合計を、呼出元メソッドが実行された回数（呼出元メソッドが示す実行回数１１６３２に相当）によって除算する。 The execution time difference 1165 is calculated by calculating the total execution time of the method indicated by the method information 1162 in each of the first program 120 and the second program 121 every time the caller method is executed once. Indicates the difference. Specifically, based on the results of the first program 120 and the second program 121, the total execution time during which the method indicated by the method information 1162 is executed is calculated. Then, the total execution time is divided by the number of times the caller method is executed (corresponding to the execution number 11632 indicated by the caller method).
　そして、前述の方法によって算出された値を、第１プログラム１２０と第２プログラム１２１とについて生成する。そして、第１プログラム１２０の算出された値によって、第２プログラム１２１の算出された値を減算することによって、実行時間差１１６５が算出される。 Then, the values calculated by the above-described method are generated for the first program 120 and the second program 121. Then, the execution time difference 1165 is calculated by subtracting the value calculated by the second program 121 from the value calculated by the first program 120.
　本発明は、実行時間差１１６５によって、呼出元メソッドが１回実行される毎の、メソッドの実行時間の差を算出することができる。 The present invention can calculate the difference in method execution time each time the caller method is executed by the execution time difference 1165.
　例えば、図５に示すメソッド情報１１６２"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"の実行時間差１１６５の算出方法を以下に示す。図５は、メソッド"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"が、"ａａａ．ｂｂｂ．Ｓｔａｒｔｅｒ．ｄｏＳｔａｒｔ（）"から呼ばれることによって実行されることを示す。 For example, a method for calculating the execution time difference 1165 of the method information 1162 “aaa.bbb.MyClass.start (int)” shown in FIG. FIG. 5 shows that the method “aaa.bbb.MyClass.start (int)” is executed by being called from “aaa.bbb.Starter.doStart ()”.
　第１プログラムにおいて実行される場合において、メソッド"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"の呼出元メソッドによって呼ばれる回数は、図５の実行回数１１６３２が示す通り、２５回である。そして、呼出元メソッドである"ａａａ．ｂｂｂ．Ｓｔａｒｔｅｒ．ｄｏＳｔａｒｔ（）"の実行回数１１６３２は５回である。すなわち、図５は、メソッド"ａａａ．ｂｂｂ．Ｓｔａｒｔｅｒ．ｄｏＳｔａｒｔ（）"が１回実行される毎に、メソッド"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"が５回（２５／５＝５）呼ばれることを示す。 When executed in the first program, the number of times called by the calling method of the method “aaa.bbb.MyClass.start (int)” is 25, as indicated by the number of executions 11632 in FIG. The number of executions 11632 of the caller method “aaa.bbb.Starter.doStart ()” is five. That is, FIG. 5 shows that each time the method “aaa.bbb.Starter.doStart ()” is executed once, the method “aaa.bbb.MyClass.start (int)” is executed five times (25/5 = 5). Indicates that it is called.
　なお、図５において、メソッド"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"及びメソッド"ａａａ．ｂｂｂ．Ｓｔａｒｔｅｒ．ｄｏＳｔａｒｔ（）"の、第２プログラムにおける実行回数１１６４２は、第１プログラムにおける実行回数１１６３２と同じである。 In FIG. 5, the number of executions 11642 in the second program of the method “aaa.bbb.MyClass.start (int)” and the method “aaa.bbb.Starter.doStart ()” is the number of executions 11632 in the first program. Is the same.
　そして、実行時間差１１６５は、第１プログラムにおいて呼出元メソッドが１回実行される毎のメソッドの合計実行時間と、第２プログラムにおいて呼出元メソッドが１回実行される毎のメソッドの合計実行時間との差である。このため、メソッド"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"の実行時間差１１６５は、第２プログラムの（平均実行時間１１６４１×（メソッドの実行回数１１６４２／呼出元メソッドの実行回数１１６４２）－第１プログラムの（平均実行時間１１６３１×（メソッドの実行回数１１６３２／呼出元メソッドの実行回数１１６３２）＝１２０７０×（２５／５）－１１４８０×（２５／５）＝２９５０と算出される。 The execution time difference 1165 is the total execution time of the method every time the caller method is executed once in the first program, and the total execution time of the method every time the caller method is executed once in the second program. Is the difference. For this reason, the execution time difference 1165 of the method “aaa.bbb.MyClass.start (int)” is (average execution time 11641 × (method execution count 11642 / calling method execution count 11642) −first program). (Average execution time 11631 × (method execution count 11632 / calling method execution count 11632) = 112070 × (25/5) −11480 × (25/5) = 2950 of the program.
　ここで、呼出元メソッドがメソッドを呼ぶ回数が、第１プログラム１２０と第２プログラム１２１とにおいて相違する場合においても、実行時間差１１６５は、前述の式によって算出される。 Here, even when the number of times the caller method calls the method differs between the first program 120 and the second program 121, the execution time difference 1165 is calculated by the above-described equation.
　実行時間比１１６６は、メソッド情報１１６２が示すメソッドの１回あたりの実行時間の、第１のプログラム及び第２のプログラムとにおける差を算出し、算出された値を平均する。そして、平均された値を、第１プログラムの平均実行時間１１６３１によって除算することによって算出された値である。 The execution time ratio 1166 calculates the difference between the execution time per method indicated by the method information 1162 between the first program and the second program, and averages the calculated values. And it is a value calculated by dividing the averaged value by the average execution time 11631 of the first program.
　すなわち、実行時間比１１６６は、メソッドが１回実行される毎の、新旧のメソッドにおける実行時間の差の比率である。本発明は、実行時間比１１６６によって、各メソッドの実行時間の変化率を算出することができる。 That is, the execution time ratio 1166 is a ratio of execution time differences between the old and new methods each time the method is executed once. The present invention can calculate the rate of change of the execution time of each method based on the execution time ratio 1166.
　具体的には、実行時間比１１６６は、実行時間差１１６５が示す値を、呼出元メソッドが１回実行される毎の、メソッド情報１１６２が示すメソッドの実行回数によって除算した値を、さらに、第１プログラムの平均実行時間１１６３１によって除算した値である。 Specifically, the execution time ratio 1166 is a value obtained by dividing the value indicated by the execution time difference 1165 by the number of times the method indicated by the method information 1162 is executed each time the call source method is executed once. It is a value divided by the average execution time 11631 of the program.
　例えば、図５において"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"は、前述のとおり、呼出元メソッドが１回実行される毎に、５回呼ばれるため、実行時間比１１６６は、（２９５０／５）／１１４８０＝０．０５１と算出される。 For example, in FIG. 5, “aaa.bbb.MyClass.start (int)” is called five times each time the caller method is executed as described above, and therefore the execution time ratio 1166 is (2950/5). ) /11480=0.051.
　なお、測定結果テーブル１１６において、初期メソッドの実行時間差１１６５及び実行時間比１１６６は、呼出元メソッド（すなわち、第１プログラム又は第２プログラム）の実行回数は、１として算出される。 In the measurement result table 116, the execution time difference 1165 and the execution time ratio 1166 of the initial method are calculated by assuming that the number of executions of the caller method (that is, the first program or the second program) is 1.
　図６は、本発明の第１の実施形態の測定結果を表示した場合の画面１３０を示す説明図である。 FIG. 6 is an explanatory diagram showing a screen 130 when the measurement result of the first embodiment of the present invention is displayed.
　画面１３０は、本実施形態によって測定された結果を、出力装置１０４を介してユーザに提供するための画面である。画面１３０は、表示１３１、表示１３２、及び表示１３３を含む。 The screen 130 is a screen for providing the user with the result measured by the present embodiment via the output device 104. The screen 130 includes a display 131, a display 132, and a display 133.
　表示１３１は、性能差要因メソッドであるメソッド名と、性能差要因メソッドである理由とを示す。図６に示す表示１３１は、呼出元メソッドである"ａａａ．ｂｂｂ．ＭｙＣｌａｓｓ．ｓｔａｒｔ（ｉｎｔ）"が、"ｊａｖａ．ｕｔｉｌ．Ｍａｐ．ｇｅｔ（Ｏｂｊｅｃｔ）"を呼ぶ回数が増加しており、第１プログラム１２０においては２００回呼び、第２プログラム１２１においては２１０回呼んでいることを示す。 The display 131 shows a method name that is a performance difference factor method and a reason that it is a performance difference factor method. The display 131 shown in FIG. 6 shows that the number of times that “aaa.bbb.MyClass.start (int)”, which is the calling method, calls “java.util.Map.get (Object)” is increased. The program 120 is called 200 times, and the second program 121 is called 210 times.
　なお、呼出元メソッドがメソッドを呼ぶ回数は、呼ばれたメソッドの実行回数（測定結果テーブル１１６の第１プログラムの実行回数１１６３２又は第２プログラムの実行回数１１６４２）を、呼出元メソッドの実行回数（測定結果テーブル１１６の第１プログラムの実行回数１１６３２又は第２プログラムの実行回数１１６４２）によって除算することによって算出される。 Note that the number of times the calling source method calls the method is the number of executions of the called method (the number of executions 11632 of the first program or the number of executions 11642 of the second program in the measurement result table 116). It is calculated by dividing by the number of executions 11632 of the first program or the number of executions 11642 of the second program in the measurement result table 116.
　表示１３２は、性能差要因メソッドが呼ばれるまでに呼ばれたメソッドを、階層構造によって示した表示である。すなわち、性能差要因メソッドを含むコールスタックをトレースした結果を示す。表示１３２には、性能差要因メソッドを含むコールスタックの各メソッドと、各メソッドにおける第１プログラム１２０と第２プログラム１２１との性能の差を示す。 The display 132 is a display showing the method called by the hierarchical structure before the performance difference factor method is called. That is, the result of tracing the call stack including the performance difference factor method is shown. The display 132 shows each method of the call stack including the performance difference factor method and the performance difference between the first program 120 and the second program 121 in each method.
　表示１３３は、測定結果テーブル１１６を含む情報を表示する。図６に示す表示１３３は、測定結果テーブル１１６の情報と、対象メソッドテーブル１１３の呼出元メソッド情報１１３３とを含む。 The display 133 displays information including the measurement result table 116. The display 133 illustrated in FIG. 6 includes information of the measurement result table 116 and caller method information 1133 of the target method table 113.
　＜処理説明＞
　図７は、本発明の第１の実施形態の全体の処理を示すフローチャートである。 <Description of processing>
FIG. 7 is a flowchart showing the overall processing of the first embodiment of the present invention.
　制御部１０６は、まず、第１プログラム１２０及び第２プログラム１２１の初期メソッドを抽出し、抽出された初期メソッドのうち、第１プログラム１２０及び第２プログラム１２１において共通するメソッドを対象メソッドテーブル１１３に格納する。すなわち、対象メソッドテーブル１１３は、第１プログラム１２０及び第２プログラム１２１において共通するメソッドを対象メソッドテーブル１１３に格納することによって、開始時刻及び終了時刻を測定するメソッド（測定対象メソッド）を、対象メソッドテーブル１１３に格納する（１４０）。 First, the control unit 106 extracts initial methods of the first program 120 and the second program 121, and among the extracted initial methods, a method common to the first program 120 and the second program 121 is stored in the target method table 113. Store. In other words, the target method table 113 stores a method common to the first program 120 and the second program 121 in the target method table 113, thereby measuring the method (measurement target method) for measuring the start time and the end time as the target method. Store in the table 113 (140).
　ステップ１４０の後、制御部１０６は、対象メソッドテーブル１１３に格納された測定対象メソッドの実行時間を測定する（１５０）。ステップ１５０の後、制御部１０６は、ステップ１５０の測定結果に基づいて、性能差要因メソッドを抽出する（１６０）。 After step 140, the control unit 106 measures the execution time of the measurement target method stored in the target method table 113 (150). After step 150, the control unit 106 extracts a performance difference factor method based on the measurement result of step 150 (160).
　そして制御部１０６は、ステップ１６０において性能差要因メソッドが抽出されたか否かを判定する（１７０）。性能差要因メソッドが抽出されない場合、第１プログラム１２０と第２プログラム１２１との間で性能の差がないため、制御部１０６は処理を終了する。 Then, the control unit 106 determines whether or not a performance difference factor method has been extracted in Step 160 (170). When the performance difference factor method is not extracted, there is no performance difference between the first program 120 and the second program 121, so the control unit 106 ends the process.
　性能差要因メソッドが抽出された場合、制御部１０６は、性能差要因メソッドによって直接呼ばれたメソッドを対象メソッドテーブル１１３に格納し、性能差要因メソッドによって直接呼ばれたメソッドをさらに測定させる（１８０）。これは、性能差要因メソッドによって直接呼ばれるメソッドの中から、最も顕著に性能の差が生じているメソッドを抽出するためである。 When the performance difference factor method is extracted, the control unit 106 stores the method directly called by the performance difference factor method in the target method table 113, and further measures the method directly called by the performance difference factor method (180). ). This is because the method in which the performance difference is most noticeable is extracted from the methods directly called by the performance difference factor method.
　ステップ１８０の後、制御部１０６は、ステップ１５０に戻る。その後、ステップ１７０において性能差要因メソッドが抽出されないと判定された場合、測定対象メソッドを呼んだ呼出元メソッドにおいて、最も性能の差が生じていることを示すため、制御部１０６は、処理を終了する。 After step 180, the control unit 106 returns to step 150. Thereafter, when it is determined in step 170 that the performance difference factor method is not extracted, the control unit 106 terminates the process to indicate that the performance difference is the greatest in the caller method that called the measurement target method. To do.
　図８は、本発明の第１の実施形態の初期メソッドを示すシーケンス図である。 FIG. 8 is a sequence diagram showing an initial method according to the first embodiment of this invention.
　図８に示すシーケンス図は、スレッドＡ１０７１、スレッドＢ１０７２、スレッド１０７３、スレッドＤ１０７４、メソッド２０１～メソッド２１６の実行時間を示す。 The sequence diagram shown in FIG. 8 shows the execution times of thread A 1071, thread B 1072, thread 1073, thread D 1074, and method 201 to method 216.
　ステップ１４０は、具体的には、テスト実行部１０７が第１プログラム１２０及び第２プログラム１２１へ特定の入力値を入力し始める時点から、特定の入力値を入力し終える時点までのテスト実行時間中に、第１プログラム１２０及び第２プログラム１２１に含まれる各スレッドのコールスタックにおいて、最も上位に格納されるメソッドを抽出する処理である。 Specifically, step 140 is during the test execution time from when the test execution unit 107 starts to input specific input values to the first program 120 and the second program 121 until when the specific input values have been input. In addition, in the call stack of each thread included in the first program 120 and the second program 121, it is a process of extracting the method stored at the highest level.
　図８において、プログラム２００は、テスト実行部１０７が第１プログラム１２０又は第２プログラム１２１へ特定の入力値を入力し始める時から、特定の入力値を入力し終える時までに実行される第１プログラム１２０又は第２プログラム１２１を示す。 In FIG. 8, the program 200 is executed from the time when the test execution unit 107 starts to input a specific input value to the first program 120 or the second program 121 to the time when the input of the specific input value is finished. The program 120 or the second program 121 is shown.
　プログラム２００は、スレッドＡ１０７１において実行されるメソッド２０１を呼ぶ。図８において、プログラム２００によって呼ばれるメソッド２０１は、常に実行されるメソッドであり、プログラム２００が開始する前から実行されているメソッドである。このため、メソッド２０１は、性能の差が生じるメソッドではないため、スレッドＡ１０７１における初期メソッドではない。 The program 200 calls the method 201 executed in the thread A 1071. In FIG. 8, a method 201 called by the program 200 is a method that is always executed, and is a method that is being executed before the program 200 is started. For this reason, the method 201 is not an initial method in the thread A 1071 because it is not a method that causes a difference in performance.
　プログラム２００によって呼ばれた後、メソッド２０１は、スレッドＢ１０７２において実行されるメソッド２０３を呼ぶ。図８において、メソッド２０１によって呼ばれるメソッド２０３は、プログラム２００が開始した後に実行され、プログラム２００が終了する前に終了するメソッドである。このため、メソッド２０３は、スレッドＢ１０７２における初期メソッドである。ここで、メソッド２０３は、スレッドＢ１０７２におけるコールスタックの最も上位に格納される。 After being called by the program 200, the method 201 calls the method 203 executed in the thread B1072. In FIG. 8, a method 203 called by the method 201 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 203 is an initial method in the thread B1072. Here, the method 203 is stored at the top of the call stack in the thread B 1072.
　なお、スレッドＢにおけるメソッド２０２及びメソッド２０９は、いずれもプログラム２００が実行される間に実行されるメソッドではなく、性能の差を生じる要因とならないため、初期メソッドではない。 Note that the method 202 and the method 209 in the thread B are not methods executed while the program 200 is executed, and are not initial methods because they do not cause a difference in performance.
　メソッド２０３は、メソッド２０３が実行されている間に、メソッド２０４及びメソッド２０５を呼ぶ。図８において、メソッド２０３によって呼ばれるメソッド２０４及びメソッド２０５は、プログラム２００が開始した後に実行され、プログラム２００が終了する前に終了するメソッドである。このため、メソッド２０４及びメソッド２０５は、測定対象メソッドとなる可能性があるメソッドである。 The method 203 calls the method 204 and the method 205 while the method 203 is being executed. In FIG. 8, a method 204 and a method 205 called by the method 203 are methods that are executed after the program 200 is started and ended before the program 200 ends. Therefore, the method 204 and the method 205 are methods that may become measurement target methods.
　本実施形態において、初期メソッドであるメソッド２０３の、コールスタックにおけるフレーム番号が"１"である場合、メソッド２０４及びメソッド２０５のフレーム番号は"２"である。 In this embodiment, when the frame number in the call stack of the method 203 as the initial method is “1”, the frame numbers of the method 204 and the method 205 are “2”.
　さらに、メソッド２０５は、メソッド２０６及びメソッド２０７を呼ぶ。ここで、メソッド２０３のフレーム番号が"１"である場合、メソッド２０６及びメソッド２０７の、コールスタックにおけるフレーム番号は"３"である。 Furthermore, the method 205 calls a method 206 and a method 207. Here, when the frame number of the method 203 is “1”, the frame number in the call stack of the method 206 and the method 207 is “3”.
　さらに、メソッド２０７はメソッド２０８を呼ぶ。ここで、メソッド２０３のフレーム番号が"１"である場合、コールスタックにおけるメソッド２０８のフレーム番号は"４"である。 Furthermore, method 207 calls method 208. Here, when the frame number of the method 203 is “1”, the frame number of the method 208 in the call stack is “4”.
　図８において、メソッド２０４、メソッド２０５、メソッド２０６、メソッド２０７及びメソッド２０８は、プログラム２００が開始した後に実行され、プログラム２００が終了する前に終了するメソッドである。このため、メソッド２０４、メソッド２０５、メソッド２０６、メソッド２０７及びメソッド２０８は、測定対象メソッドとなる可能性があるメソッドである。 In FIG. 8, a method 204, a method 205, a method 206, a method 207, and a method 208 are methods that are executed after the program 200 starts and end before the program 200 ends. Therefore, the method 204, the method 205, the method 206, the method 207, and the method 208 are methods that may become measurement target methods.
　メソッド２０４は、スレッドＣ１０７３におけるメソッド２１０を呼ぶ。図８において、メソッド２０４によって呼ばれるメソッド２１０は、プログラム２００が終了した後も実行されるメソッドである。このため、メソッド２１０は、性能の差を生じる要因とならないため、スレッドＣ１０７３における初期メソッドではない。 Method 204 calls method 210 in thread C1073. In FIG. 8, a method 210 called by the method 204 is a method that is executed even after the program 200 ends. For this reason, the method 210 does not cause a difference in performance, and thus is not an initial method in the thread C1073.
　メソッド２１０は、メソッド２１１を呼ぶ。図８において、メソッド２１０によって呼ばれるメソッド２１１は、プログラム２００が開始した後に実行され、プログラム２００が終了する前に終了するメソッドである。このため、メソッド２１１は、スレッドＣ１０７３における初期メソッドである。ここで、メソッド２１１は、スレッドＣ１０７３におけるコールスタックの上位に格納される。この時、スレッドＣ１０７３のコールスタックには、メソッド２１０が既に格納されているため、メソッド２１１のフレーム番号は"２"である。 Method 210 calls method 211. In FIG. 8, a method 211 called by the method 210 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 211 is an initial method in the thread C1073. Here, the method 211 is stored above the call stack in the thread C1073. At this time, since the method 210 has already been stored in the call stack of the thread C1073, the frame number of the method 211 is “2”.
　さらに、メソッド２１１はメソッド２１２を呼ぶ。ここで、メソッド２１１のフレーム番号が"２"である場合、メソッド２１２のコールスタックにおけるフレーム番号は"３"である。 Furthermore, method 211 calls method 212. Here, when the frame number of the method 211 is “2”, the frame number in the call stack of the method 212 is “3”.
　図８において、メソッド２１２は、プログラム２００が開始した後に実行され、プログラム２００が終了する前に終了するメソッドである。このため、メソッド２１２は、測定対象メソッドとなる可能性があるメソッドである。 Referring to FIG. 8, a method 212 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 212 is a method that may be a measurement target method.
　メソッド２１１は、スレッドＤ１０７４におけるメソッド２１４を呼ぶ。図８において、メソッド２１１によって呼ばれるメソッド２１４は、プログラム２００が開始する前から実行されているメソッドである。このため、メソッド２１４は、性能の差を生じる要因とならないため、スレッドＤ１０７４における初期メソッドではない。 Method 211 calls method 214 in thread D1074. In FIG. 8, a method 214 called by a method 211 is a method that is being executed before the program 200 is started. For this reason, the method 214 does not cause a difference in performance, and thus is not an initial method in the thread D1074.
　メソッド２１４は、メソッド２１６を呼ぶ。図８において、メソッド２１４によって呼ばれるメソッド２１６は、プログラム２００が開始した後に実行され、プログラム２００が終了する前に終了するメソッドである。このためメソッド２１６は、初期メソッドである。ここで、スレッドＤ１０７４のコールスタックには、メソッド２１３及びメソッド２１４が既に格納されているため、メソッド２１６のフレーム番号は"３"である。 Method 214 calls method 216. In FIG. 8, a method 216 called by the method 214 is a method that is executed after the program 200 starts and ends before the program 200 ends. Therefore, the method 216 is an initial method. Here, since the method 213 and the method 214 are already stored in the call stack of the thread D1074, the frame number of the method 216 is “3”.
　スレッドＤ１０７４のメソッド２１５は、メソッド２１６及びメソッド２１４が終了した後に実行され、プログラム２００が終了する前に終了するメソッドである。このため、性能の差を生じる要因となるため、初期メソッドとなる。ここで、スレッドＤ１０７４のコールスタックには、メソッド２１３が既に格納されているため、メソッド２１６のフレーム番号は"２"である。 The method 215 of the thread D 1074 is a method that is executed after the method 216 and the method 214 are finished, and is finished before the program 200 is finished. For this reason, it becomes a factor that causes a difference in performance, and thus becomes an initial method. Here, since the method 213 is already stored in the call stack of the thread D1074, the frame number of the method 216 is “2”.
　本実施形態の制御部１０６は、前述の初期メソッドをステップ１４０において抽出する。 The control unit 106 of the present embodiment extracts the initial method described above in step 140.
　図９は、本発明の第１の実施形態の初期メソッドを抽出する処理におけるデータの流れを示すブロック図である。 FIG. 9 is a block diagram illustrating a data flow in the process of extracting the initial method according to the first embodiment of this invention.
　図９に示す矢印は、図７に示すステップ１４０におけるデータの流れを示し、図１０Ａ、図１０Ｂ、及び、図１０Ｃに示す処理に対応する。 The arrows shown in FIG. 9 indicate the data flow in step 140 shown in FIG. 7, and correspond to the processing shown in FIGS. 10A, 10B, and 10C.
　図１０Ａは、本発明の第１の実施形態の初期メソッドを抽出するためにテストを実行する処理を示すフローチャートである。 FIG. 10A is a flowchart showing a process for executing a test to extract an initial method according to the first embodiment of this invention.
　図１０Ａに示す処理は、図７に示すステップ１４０に含まれる処理である。まず、第１プログラム１２０をテストが可能な状態にする指示を制御部１０６から受信した後、バージョン切替部１１０は、第１プログラム１２０を実行可能な状態に切り替え、その後、切り替えの終了を制御部１０６に通知する（１４０１）。 The process shown in FIG. 10A is a process included in step 140 shown in FIG. First, after receiving an instruction from the control unit 106 to set the first program 120 in a testable state, the version switching unit 110 switches the first program 120 to a state in which the first program 120 can be executed. 106 is notified (1401).
　ステップ１４０１の後、制御部１０６は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を初期メソッド記録部１１７に送信する。初期メソッド記録部１１７は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＴＲＵＥ"を格納する（１４０２）。なお、初期メソッド記録部１１７は、メモリ１０２の中に、テスト実行中フラグを示す領域をあらかじめ保持する。 After step 1401, the control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the initial method recording unit 117. After receiving an instruction to store “TRUE” in the test execution flag from the control unit 106, the initial method recording unit 117 stores “TRUE” in the test execution flag (1402). Note that the initial method recording unit 117 holds an area indicating the test execution flag in the memory 102 in advance.
　テスト実行中フラグに"ＴＲＵＥ"が格納される場合、テスト実行中フラグは、テスト実行部１０７がプログラムをテストしていることを示す。テスト実行中フラグに"ＦＡＬＳＥ"が格納される場合、テスト実行中フラグは、テスト実行部１０７がプログラムをテストしていないことを示す。 When “TRUE” is stored in the test execution flag, the test execution flag indicates that the test execution unit 107 is testing the program. When “FALSE” is stored in the test execution flag, the test execution flag indicates that the test execution unit 107 has not tested the program.
　ステップ１４０２の後、制御部１０６は、第１プログラム１２０のテストを開始する指示をテスト実行部１０７に送信する。テスト実行部１０７は、第１プログラム１２０のテストを開始する指示を制御部１０６から受信すると、第１プログラム１２０のテストを開始する（１４０３）。 After step 1402, the control unit 106 transmits an instruction to start the test of the first program 120 to the test execution unit 107. When the test execution unit 107 receives an instruction to start the test of the first program 120 from the control unit 106, the test execution unit 107 starts the test of the first program 120 (1403).
　ステップ１４０３の後、テスト実行部１０７は、テストを実行するため、あらかじめユーザによって指定された所定の入力値を第１プログラム１２０に入力する（１４０４）。ステップ１４０４の後、テスト実行部１０７は、第１プログラム１２０のテストを終了し、その後、テストの終了を制御部１０６に通知する（１４０５）。 After step 1403, the test execution unit 107 inputs a predetermined input value designated in advance by the user to the first program 120 in order to execute the test (1404). After step 1404, the test execution unit 107 ends the test of the first program 120, and then notifies the control unit 106 of the end of the test (1405).
　ステップ１４０５の後、制御部１０６は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を初期メソッド記録部１１７に送信する。初期メソッド記録部１１７は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＦＡＬＳＥ"を格納する（１４０６）。 After step 1405, the control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the initial method recording unit 117. The initial method recording unit 117 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1406).
　ステップ１４０６の後、制御部１０６は、第１プログラム１２０を終了させる指示をバージョン切替部１１０に送信する（１４０７）。 After step 1406, the control unit 106 transmits an instruction to end the first program 120 to the version switching unit 110 (1407).
　テスト実行中フラグに"ＦＡＬＳＥ"を格納した後、初期メソッド記録部１１７は、スレッドごとのコールスタック（以下、メソッド情報スタックと記載）のうち、終了済みフラグが付加されているメソッドを抽出し、抽出されたメソッドを示す情報を、第１メソッド呼出テーブル１１１に格納する（１４０９）。終了済みフラグは、後述する図１０Ｃに示す処理において更新される記憶領域である。 After storing “FALSE” in the test execution flag, the initial method recording unit 117 extracts a method to which a completed flag is added from a call stack for each thread (hereinafter referred to as a method information stack). Information indicating the extracted method is stored in the first method call table 111 (1409). The completed flag is a storage area that is updated in the process shown in FIG.
　そして、初期メソッド記録部１１７は、すべてのスレッドのメソッド情報スタックについて、ステップ１４０９を繰り返す（１４０８）。 The initial method recording unit 117 repeats Step 1409 for the method information stack of all threads (1408).
　ステップ１４０８が終了した後、バージョン切替部１１０は、第２プログラム１２１をテストが可能な状態にする指示を制御部１０６から受信する。その後、バージョン切替部１１０は、第２プログラム１２１を実行可能な状態に切り替え、その後、切り替えの終了を制御部１０６に通知する（１４１０）。 After step 1408 is completed, the version switching unit 110 receives an instruction from the control unit 106 to make the second program 121 ready for testing. Thereafter, the version switching unit 110 switches the second program 121 to an executable state, and then notifies the control unit 106 of the end of the switching (1410).
　ステップ１４１０の後、制御部１０６は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を初期メソッド記録部１１７に送信する。初期メソッド記録部１１７は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＴＲＵＥ"を格納する（１４１１）。 After step 1410, the control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the initial method recording unit 117. The initial method recording unit 117 stores an instruction to store “TRUE” in the test execution flag from the control unit 106, and then stores “TRUE” in the test execution flag (1411).
　ステップ１４１１の後、制御部１０６は、第２プログラム１２１のテストを開始する指示をテスト実行部１０７に送信する。テスト実行部１０７は、第２プログラム１２１のテストを開始する指示を制御部１０６から受信すると、第２プログラム１２１のテストを開始する（１４１２）。 After step 1411, the control unit 106 transmits an instruction to start the test of the second program 121 to the test execution unit 107. When the test execution unit 107 receives an instruction to start the test of the second program 121 from the control unit 106, the test execution unit 107 starts the test of the second program 121 (1412).
　ステップ１４１２の後、テスト実行部１０７は、テストを実行するため、あらかじめユーザによって指定された所定の値を第２プログラム１２１に入力する（１４１３）。ステップ１４１３の後、テスト実行部１０７は、第２プログラム１２１のテストを終了し、その後、テストの終了を制御部１０６に通知する（１４１４）。 After step 1412, the test execution unit 107 inputs a predetermined value designated in advance by the user to the second program 121 in order to execute the test (1413). After step 1413, the test execution unit 107 ends the test of the second program 121, and then notifies the control unit 106 of the end of the test (1414).
　ステップ１４１４の後、制御部１０６は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を初期メソッド記録部１１７に送信する。初期メソッド記録部１１７は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＦＡＬＳＥ"を格納する（１４１５）。 After step 1414, the control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the initial method recording unit 117. The initial method recording unit 117 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1415).
　ステップ１４１５の後、制御部１０６は、第２プログラム１２１を終了させる指示をバージョン切替部１１０に送信する（１４１６）。 After step 1415, the control unit 106 transmits an instruction to end the second program 121 to the version switching unit 110 (1416).
　テスト実行中フラグに"ＦＡＬＳＥ"を格納した後、初期メソッド記録部１１７は、スレッドごとのメソッド情報スタックのうち、対応する終了済みフラグが生成されているメソッドを抽出し、抽出されたメソッドを示す情報を、第１メソッド呼出テーブル１１１に格納する（１４１８）。 After storing “FALSE” in the test execution flag, the initial method recording unit 117 extracts the method for which the corresponding completed flag is generated from the method information stack for each thread, and indicates the extracted method. The information is stored in the first method call table 111 (1418).
　そして、初期メソッド記録部１１７は、すべてのスレッドのメソッド情報スタックについて、ステップ１４１８を繰り返す（１４１７）。 Then, the initial method recording unit 117 repeats Step 1418 for the method information stack of all threads (1417).
　ステップ１４１７の後、制御部１０６は、測定対象メソッドを抽出し（１４１９）、図１０Ａに示す処理を終了する。 After step 1417, the control unit 106 extracts a measurement target method (1419), and ends the process illustrated in FIG. 10A.
　図１０Ｂは、本発明の第１の実施形態の測定対象メソッドを抽出する処理を示すフローチャートである。 FIG. 10B is a flowchart showing a process of extracting a measurement target method according to the first embodiment of this invention.
　図１０Ｂに示す処理は、図１０Ａに示すステップ１４１９に相当する。 The process shown in FIG. 10B corresponds to step 1419 shown in FIG. 10A.
　制御部１０６は、図１０Ａのステップ１４１７が終了した後、測定対象メソッドを抽出する指示をメソッド呼出照合部１０８に送信する（１４２０）。測定対象メソッドを抽出する指示をメソッド呼出照合部１０８から受信した後、メソッド呼出照合部１０８は、第１メソッド呼出テーブル１１１に格納される情報を読み込む（１４２１）。また、メソッド呼出照合部１０８は、第２メソッド呼出テーブル１１２に格納される情報を読み込む（１４２２）。 The control unit 106 transmits an instruction to extract the measurement target method to the method call matching unit 108 after step 1417 in FIG. 10A is completed (1420). After receiving the instruction to extract the measurement target method from the method call matching unit 108, the method call matching unit 108 reads information stored in the first method call table 111 (1421). Further, the method call matching unit 108 reads information stored in the second method call table 112 (1422).
　ステップ１４２１及びステップ１４２２の後、メソッド呼出照合部１０８は、新たな測定対象メソッドを対象メソッドテーブル１１３に格納するため、対象メソッドテーブル１１３に格納されていたデータを削除する（１４２３）。ステップ１４２３の後、メソッド呼出照合部１０８は、第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２から読み込んだ情報に基づいて、第１メソッド呼出テーブル１１１と第２メソッド呼出テーブル１１２とに共通して格納されるメソッドを抽出し、抽出されたメソッドを対象メソッドテーブル１１３に格納する（１４２４）。 After step 1421 and step 1422, the method call matching unit 108 deletes the data stored in the target method table 113 in order to store the new measurement target method in the target method table 113 (1423). After step 1423, the method call verification unit 108 is common to the first method call table 111 and the second method call table 112 based on the information read from the first method call table 111 and the second method call table 112. The extracted method is extracted, and the extracted method is stored in the target method table 113 (1424).
　ステップ１４２４の後、メソッド呼出照合部１０８は、制御部に測定対象メソッドの抽出の終了を通知し、ステップ１４１９の処理を終了する（１４２５）。 After step 1424, the method call matching unit 108 notifies the control unit of the end of extraction of the measurement target method, and ends the process of step 1419 (1425).
　ステップ１４１９の処理によって、第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２に共通して格納されたメソッドが測定対象メソッドとして抽出される。 By the processing in step 1419, the methods stored in common in the first method call table 111 and the second method call table 112 are extracted as measurement target methods.
　図１０Ｃは、本発明の第１の実施形態の初期メソッドを抽出する処理を示すフローチャートである。 FIG. 10C is a flowchart showing processing for extracting an initial method according to the first embodiment of this invention.
　図１０Ｃに示す処理は、図１０Ａに示すテストの実行中に、各プログラムから呼ばれるメソッドの中から初期メソッドを抽出するための処理である。 The process shown in FIG. 10C is a process for extracting an initial method from methods called by each program during the execution of the test shown in FIG. 10A.
　コールバック受信部１１９は、ＪＶＭＴＩ１２３からコールバックを受信し、コールバックが受信された時刻と、コールバックに含まれる各情報を初期メソッド記録部１１７に送信する。コールバックには、前述の通り、メソッドを示す情報（クラス名、メソッド名、メソッド識別子、呼び出し元のメソッドを示す情報、及び、コールスタックにおけるフレーム番号など）と、メソッドが開始されたか終了したかを示す情報とが含まれる。 The callback receiving unit 119 receives the callback from the JVMTI 123, and transmits the time when the callback is received and each information included in the callback to the initial method recording unit 117. As described above, the callback includes information indicating the method (class name, method name, method identifier, information indicating the calling method, frame number in the call stack, etc.), and whether the method has started or ended. Is included.
　ステップ１４３０の後、初期メソッド記録部１１７は、テスト実行中フラグに"ＴＲＵＥ"が格納されているか否かを判定する（１４３１）。テスト実行中フラグに"ＦＡＬＳＥ"が格納されている場合、テストが実行されておらず、初期メソッドを抽出する必要がないため、初期メソッド記録部１１７はステップ１４３８に移行する。 After step 1430, the initial method recording unit 117 determines whether “TRUE” is stored in the test execution flag (1431). If “FALSE” is stored in the test execution flag, the test has not been executed and it is not necessary to extract the initial method, so the initial method recording unit 117 proceeds to step 1438.
　テスト実行中フラグに"ＴＲＵＥ"が格納されている場合、初期メソッド記録部１１７は、受信されたコールバックに含まれる情報に基づいて、メソッドが開始されたか終了したかを判定する（１４３２）。 When “TRUE” is stored in the test execution flag, the initial method recording unit 117 determines whether the method is started or ended based on the information included in the received callback (1432).
　メソッドが開始されたと判定された場合、スレッド毎のメソッド情報スタックにコールバックに含まれるメソッドを示す情報を格納する（ＰＵＳＨ）（１４３３）。ステップ１４３３によって、メソッド情報スタックに格納されたメソッドは、初期メソッドの候補になる。 When it is determined that the method is started, information indicating the method included in the callback is stored in the method information stack for each thread (PUSH) (1433). In step 1433, the method stored in the method information stack becomes an initial method candidate.
　ステップ１４３２においてメソッドが終了したと判定された場合、初期メソッド記録部１１７は、コールバックに含まれるメソッドの情報が、既にメソッド情報スタックのうちのいずれかに格納されているか否かを判定する（１４３４）。ステップ１４３４の判定によって、コールバックが示すメソッドが、テストが実行されている間に開始されたメソッドであるか否かを判定できる。なお、前述の通り、初期メソッドは、テストが実行されている間に、開始され、終了したメソッドである。 If it is determined in step 1432 that the method has ended, the initial method recording unit 117 determines whether or not the method information included in the callback is already stored in any of the method information stacks ( 1434). The determination in step 1434 can determine whether or not the method indicated by the callback is a method started while the test is being executed. As described above, the initial method is a method that is started and ended while the test is being executed.
　ステップ１４３４において、コールバックに含まれるメソッドの情報が、メソッド情報スタックのいずれにも格納されていないと判定された場合、コールバックが示すメソッドは、テストが実行されている間に開始されていないメソッドであり、初期メソッドにはならないため、初期メソッド記録部１１７は、ステップ１４３９に移行する。 If it is determined in step 1434 that the method information included in the callback is not stored in any of the method information stacks, the method indicated by the callback is not started while the test is being executed. Since it is a method and not an initial method, the initial method recording unit 117 proceeds to Step 1439.
　ステップ１４３４において、コールバックに含まれるメソッドの情報が、メソッド情報スタックのうちのいずれかに格納されていると判定された場合、コールバックが示すメソッドは、テストが実行されている間に開始されたメソッドであり、初期メソッドになる可能性があるメソッドである。このため初期メソッド記録部１１７は、ステップ１４３４の後、ステップ１４３５に移行する。 If it is determined in step 1434 that the method information included in the callback is stored in any of the method information stacks, the method indicated by the callback is started while the test is being executed. It is a method that can be an initial method. Therefore, the initial method recording unit 117 proceeds to step 1435 after step 1434.
　初期メソッド記録部１１７は、ステップ１４３４においてコールバックに含まれるメソッドの情報が格納されていると判定されたメソッド情報スタックの、最も下位に格納されたメソッドを取得する（ＰＯＰ）。そして、取得されたメソッドを示す情報を、先頭のメソッド情報に格納する（１４３５）。先頭のメソッド情報は、メソッドを示す情報を一時的に格納するための記憶領域であり、初期メソッド記録部１１７によってあらかじめ保持される。 The initial method recording unit 117 obtains the method stored at the lowest level of the method information stack determined in step 1434 that the method information included in the callback is stored (POP). Then, information indicating the acquired method is stored in the first method information (1435). The first method information is a storage area for temporarily storing information indicating a method, and is held in advance by the initial method recording unit 117.
　ステップ１４３５の後、初期メソッド記録部１１７は、先頭のメソッド情報が、コールバックに含まれるメソッドを示す情報と同じであるか否かを判定する（１４３６）。先頭のメソッド情報が、コールバックに含まれるメソッドを示す情報と異なる場合、先頭のメソッド情報に格納されたメソッドは、コールバックに含まれるメソッドよりも下位のメソッドである。このため、初期メソッド記録部１１７は、さらに上位のメソッドを先頭のメソッド情報に格納するため、ステップ１４３５に戻る。 After step 1435, the initial method recording unit 117 determines whether or not the top method information is the same as the information indicating the method included in the callback (1436). When the head method information is different from the information indicating the method included in the callback, the method stored in the head method information is a lower-level method than the method included in the callback. Therefore, the initial method recording unit 117 returns to step 1435 in order to store a higher-order method in the top method information.
　ステップ１４３６において、先頭のメソッド情報が、コールバックに含まれるメソッドを示す情報と同じである場合、初期メソッド記録部１１７は、先頭のメソッド情報に対応する終了済みフラグを生成する（１４３７）。終了済みフラグが対応して生成されている場合、先頭のメソッド情報が示すメソッドは、終了していることを示す。 In step 1436, when the head method information is the same as the information indicating the method included in the callback, the initial method recording unit 117 generates a completed flag corresponding to the head method information (1437). When the completed flag is generated correspondingly, the method indicated by the first method information indicates that it has been completed.
　ステップ１４３７の後、初期メソッド記録部１１７は、先頭のメソッド情報に格納されるメソッド情報を、ステップ１４３５において取得されたメソッド情報スタックに再度格納（ＰＵＳＨ）する（１４３８）。 After step 1437, the initial method recording unit 117 stores (PUSH) again the method information stored in the head method information in the method information stack acquired in step 1435 (1438).
　ステップ１４３８の処理によって、メソッド情報スタックのうち最も下位には、終了済みのメソッドが格納される。また、図１０Ｃに示す処理が、コールバックが発行された都度実行されることによって、テスト実行中に開始され、終了したメソッドのうち、最も上位のメソッドがメソッド情報スタックに残る。このため、テストが終了した後において、メソッド情報スタックに格納され、対応する終了済みフラグが生成されているメソッドは、初期メソッドである。 By the processing of step 1438, the finished method is stored at the lowest position in the method information stack. Also, the process shown in FIG. 10C is executed each time a callback is issued, so that the highest method among the methods started and ended during the test execution remains in the method information stack. For this reason, after the test is finished, the method that is stored in the method information stack and the corresponding finished flag is generated is the initial method.
　ステップ１４３８の後、初期メソッド記録部１１７は、図１０Ｃの処理を終了し、コールバック受信部１１９に処理の終了を通知する（１４３９）。 After step 1438, the initial method recording unit 117 ends the process of FIG. 10C and notifies the callback receiving unit 119 of the end of the process (1439).
　図１０Ｃに示す処理によってメソッド情報スタックに初期メソッドが格納された後、図１０Ａに示すステップ１４０９及びステップ１４１８によって、初期メソッド記録部１１７は、スレッドごとのメソッド情報スタックのうち、終了済みフラグが生成されているメソッドを、第１メソッド呼出テーブル１１１又は第２メソッド呼出テーブル１１２に格納する。これによって、第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２に初期メソッドが格納される。 After the initial method is stored in the method information stack by the process shown in FIG. 10C, the initial method recording unit 117 generates a completed flag in the method information stack for each thread in steps 1409 and 1418 shown in FIG. 10A. The stored method is stored in the first method call table 111 or the second method call table 112. As a result, the initial method is stored in the first method call table 111 and the second method call table 112.
　図１１は、本発明の第１の実施形態の測定対象メソッドの実行時間を測定する処理におけるデータの流れを示すブロック図である。 FIG. 11 is a block diagram illustrating a data flow in the process of measuring the execution time of the measurement target method according to the first embodiment of this invention.
　図１１に示す矢印は、図７に示すステップ１５０におけるデータの流れを示し、図１２Ａ、及び、図１２Ｂに示す処理に対応する。 The arrows shown in FIG. 11 indicate the data flow in step 150 shown in FIG. 7, and correspond to the processing shown in FIGS. 12A and 12B.
　図１２Ａは、本発明の第１の実施形態の実行時間を測定するためにテストを実行する処理を示すフローチャートである。 FIG. 12A is a flowchart illustrating a process for executing a test to measure the execution time according to the first embodiment of this invention.
　図１２Ａに示す処理は、図７に示すステップ１５０に含まれる処理である。まず、第１プログラム１２０をテストが可能な状態にする指示を制御部１０６から受信した後、バージョン切替部１１０は、第１プログラム１２０を実行可能な状態に切り替え、その後、切り替えの終了を制御部１０６に通知する（１５０１）。 The process shown in FIG. 12A is a process included in step 150 shown in FIG. First, after receiving an instruction from the control unit 106 to set the first program 120 in a testable state, the version switching unit 110 switches the first program 120 to a state in which the first program 120 can be executed. 106 is notified (1501).
　ステップ１５０１の後、制御部１０６は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を対象メソッド記録部１１８に送信する。対象メソッド記録部１１８は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＴＲＵＥ"を格納する（１５０２）。 After step 1501, the control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the target method recording unit 118. The target method recording unit 118 receives an instruction to store “TRUE” in the test execution flag from the control unit 106, and then stores “TRUE” in the test execution flag (1502).
　なお、対象メソッド記録部１１８は、メモリ１０２の中に、テスト実行中フラグを示す領域をあらかじめ保持する。初期メソッド記録部１１７が保持するテスト実行中フラグと対象メソッド記録部１１８が保持するテスト実行中フラグは、同じ領域に格納されてもよい。 Note that the target method recording unit 118 holds an area indicating the test execution flag in the memory 102 in advance. The test execution flag held by the initial method recording unit 117 and the test execution flag held by the target method recording unit 118 may be stored in the same area.
　ステップ１５０２の後、制御部１０６は、第１プログラム１２０のテストを開始する指示をテスト実行部１０７に送信する。テスト実行部１０７は、第１プログラム１２０のテストを開始する指示を制御部１０６から受信すると、第１プログラム１２０のテストを開始する（１５０３）。 After step 1502, the control unit 106 transmits an instruction to start the test of the first program 120 to the test execution unit 107. When the test execution unit 107 receives an instruction to start the test of the first program 120 from the control unit 106, the test execution unit 107 starts the test of the first program 120 (1503).
　ステップ１５０３の後、テスト実行部１０７は、テストを実行するため、あらかじめユーザによって指定された所定の値を第１プログラム１２０に入力する（１５０４）。ステップ１５０４の後、テスト実行部１０７は、第１プログラム１２０のテストを終了し、その後、テストの終了を制御部１０６に通知する（１５０５）。 After step 1503, the test execution unit 107 inputs a predetermined value designated in advance by the user to the first program 120 in order to execute the test (1504). After step 1504, the test execution unit 107 ends the test of the first program 120, and then notifies the control unit 106 of the end of the test (1505).
　ステップ１５０５の後、制御部１０６は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を対象メソッド記録部１１８に送信する。対象メソッド記録部１１８は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＦＡＬＳＥ"を格納する（１５０６）。 After step 1505, the control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the target method recording unit 118. The target method recording unit 118 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1506).
　ステップ１５０６の後、制御部１０６は、第１プログラム１２０を終了させる指示をバージョン切替部１１０に送信する（１５０７）。 After step 1506, the control unit 106 transmits an instruction to end the first program 120 to the version switching unit 110 (1507).
　ステップ１５０７の後、バージョン切替部１１０は、第２プログラム１２１をテストが可能な状態にする指示を制御部１０６から受信する。その後、バージョン切替部１１０は、第２プログラム１２１を実行可能な状態に切り替え、その後、切り替えの終了を制御部１０６に通知する（１５０８）。 After step 1507, the version switching unit 110 receives an instruction from the control unit 106 to make the second program 121 ready for testing. Thereafter, the version switching unit 110 switches the second program 121 to an executable state, and then notifies the control unit 106 of the end of the switching (1508).
　ステップ１５０８の後、制御部１０６は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を対象メソッド記録部１１８に送信する。対象メソッド記録部１１８は、テスト実行中フラグに"ＴＲＵＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＴＲＵＥ"を格納する（１５０９）。 After step 1508, the control unit 106 transmits an instruction to store “TRUE” in the test execution flag to the target method recording unit 118. The target method recording unit 118 receives an instruction to store “TRUE” in the test execution flag from the control unit 106, and then stores “TRUE” in the test execution flag (1509).
　ステップ１５０９の後、制御部１０６は、第２プログラム１２１のテストを開始する指示をテスト実行部１０７に送信する。テスト実行部１０７は、第２プログラム１２１のテストを開始する指示を制御部１０６から受信すると、第２プログラム１２１のテストを開始する（１５１０）。 After step 1509, the control unit 106 transmits an instruction to start the test of the second program 121 to the test execution unit 107. When the test execution unit 107 receives an instruction to start the test of the second program 121 from the control unit 106, the test execution unit 107 starts the test of the second program 121 (1510).
　ステップ１５１０の後、テスト実行部１０７は、テストを実行するため、あらかじめユーザによって指定された所定の値を第２プログラム１２１に入力する（１５１１）。ステップ１５１１の後、テスト実行部１０７は、第２プログラム１２１のテストを終了し、その後、テストの終了を制御部１０６に通知する（１５１２）。 After step 1510, the test execution unit 107 inputs a predetermined value designated in advance by the user to the second program 121 in order to execute the test (1511). After step 1511, the test execution unit 107 ends the test of the second program 121, and then notifies the control unit 106 of the end of the test (1512).
　ステップ１５１２の後、制御部１０６は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を対象メソッド記録部１１８に送信する。対象メソッド記録部１１８は、テスト実行中フラグに"ＦＡＬＳＥ"を格納する指示を制御部１０６から受信した後、テスト実行中フラグに"ＦＡＬＳＥ"を格納する（１５１３）。 After step 1512, the control unit 106 transmits an instruction to store “FALSE” in the test execution flag to the target method recording unit 118. The target method recording unit 118 receives an instruction to store “FALSE” in the test execution flag from the control unit 106, and then stores “FALSE” in the test execution flag (1513).
　ステップ１５１３の後、制御部１０６は、第２プログラム１２１を終了させる指示をバージョン切替部１１０に送信し（１５１４）、図１２Ａに示す処理を終了する。 After step 1513, the control unit 106 transmits an instruction to end the second program 121 to the version switching unit 110 (1514), and ends the process illustrated in FIG. 12A.
　図１２Ｂは、本発明の第１の実施形態の測定対象メソッドの実行時間を測定する処理を示すフローチャートである。 FIG. 12B is a flowchart illustrating processing for measuring the execution time of the measurement target method according to the first embodiment of this invention.
　図１２Ｂに示す処理は、第１プログラム１２０が実行された際の測定対象メソッドの実行時間を測定する処理を示す。なお、第２プログラム１２１が実行された際の測定対象メソッドの実行時間を測定する処理は、図１２Ｂに示す処理に含まれる、第１プログラム１２０を第２プログラム１２１に変更し、第１メソッド呼出テーブル１１１を第２メソッド呼出テーブル１１２に変更し、第１実行情報テーブル１１４を第２実行情報テーブル１１５に変更した処理である。 The process shown in FIG. 12B shows the process of measuring the execution time of the measurement target method when the first program 120 is executed. The process for measuring the execution time of the measurement target method when the second program 121 is executed is performed by changing the first program 120 to the second program 121 included in the process shown in FIG. This is a process in which the table 111 is changed to the second method call table 112 and the first execution information table 114 is changed to the second execution information table 115.
　まず、コールバック受信部１１９は、第１プログラム１２０のテストを実行中に、ＪＶＭＴＩ１２３からコールバックを受信し、コールバックが受信された時刻と、コールバックに含まれる各情報を対象メソッド記録部１１８に送信する（１５１４）。 First, the callback receiving unit 119 receives a callback from the JVMTI 123 during execution of the test of the first program 120, and records the time when the callback is received and each piece of information included in the callback as the target method recording unit 118. (1514).
　対象メソッド記録部１１８は、テスト実行中フラグに"ＴＲＵＥ"が格納されているか否かを判定する（１５１５）。テスト実行中フラグに"ＦＡＬＳＥ"が格納されている場合、テストが実行されていないため、実行時間を測定できないため、対象メソッド記録部１１８はステップ１５２６に移行し、コールバック受信部１１９に処理の終了を通知する。 The target method recording unit 118 determines whether “TRUE” is stored in the test execution flag (1515). When “FALSE” is stored in the test execution flag, since the test is not executed and the execution time cannot be measured, the target method recording unit 118 proceeds to step 1526 and the callback reception unit 119 performs processing. Notify the end.
　なお、テスト実行中フラグには、第１プログラム１２０又は第２プログラム１２１のいずれがテストされているかを示す情報が付加されてもよく、対象メソッド記録部１１８は、テスト実行中フラグを参照することによって、第１実行情報テーブル１１４及び第１メソッド呼出テーブル１１１を用いて処理するか、第２実行情報テーブル１１５及び第２メソッド呼出テーブル１１２を用いて処理するかを判定してもよい。 Information indicating which of the first program 120 or the second program 121 is being tested may be added to the test execution flag, and the target method recording unit 118 refers to the test execution flag. Thus, it may be determined whether processing is performed using the first execution information table 114 and the first method call table 111 or processing is performed using the second execution information table 115 and the second method call table 112.
　テスト実行中フラグに"ＴＲＵＥ"が格納されている場合、対象メソッド記録部１１８は、コールバックが示すメソッドが実行されたスレッドのメソッド実行中フラグに"ＦＡＬＳＥ"が格納されているか否かを判定する（１５１６）。 When “TRUE” is stored in the test execution flag, the target method recording unit 118 determines whether “FALSE” is stored in the method execution flag of the thread in which the method indicated by the callback is executed. (1516).
　メソッド実行中フラグは、対象メソッド記録部１１８によってあらかじめ保持される記憶領域である。あるスレッドのメソッド実行中フラグに"ＦＡＬＳＥ"が格納される場合、そのスレッドにおいて測定対象メソッドが実行されていないことを示し、あるスレッドのメソッド実行中フラグに"ＴＲＵＥ"が格納される場合、そのスレッドにおいて測定対象メソッドが実行されていることを示す。メソッド実行中フラグは、後述するステップ１５１９、及び、ステップ１５２２において更新される。 The method executing flag is a storage area that is held in advance by the target method recording unit 118. When “FALSE” is stored in the method execution flag of a thread, it indicates that the measurement target method is not executed in the thread. When “TRUE” is stored in the method execution flag of a thread, Indicates that the measurement target method is being executed in the thread. The method executing flag is updated in step 1519 and step 1522 described later.
　なお、対象メソッド記録部１１８は、メソッドが実行されるスレッドを、コールバックに含まれるメソッドを示す情報から取得してもよい。 Note that the target method recording unit 118 may acquire the thread in which the method is executed from information indicating the method included in the callback.
　ステップ１５１６において、メソッドが実行されるスレッドのメソッド実行中フラグに"ＦＡＬＳＥ"が格納されていると判定された場合、ステップ１５１４において受信されたコールバックはメソッドの開始を示すため、対象メソッド記録部１１８は、開始時刻を格納するためステップ１５１７に移行する。 If it is determined in step 1516 that “FALSE” is stored in the method execution flag of the thread in which the method is executed, the callback received in step 1514 indicates the start of the method, so that the target method recording unit 118 moves to step 1517 to store the start time.
　ステップ１５１６の後、対象メソッド記録部１１８は、コールバックが示すメソッドが、対象メソッドテーブル１１３に格納されているか否かを判定する（１５１７）。コールバックが示すメソッドが、対象メソッドテーブル１１３に格納されていない場合、コールバックが示すメソッドは測定対象メソッドではないため、対象メソッド記録部１１８は、ステップ１５２６に移行する。 After step 1516, the target method recording unit 118 determines whether or not the method indicated by the callback is stored in the target method table 113 (1517). If the method indicated by the callback is not stored in the target method table 113, the target method recording unit 118 proceeds to step 1526 because the method indicated by the callback is not a measurement target method.
　ステップ１５１７において、コールバックが示すメソッドが対象メソッドテーブル１１３に格納されていると判定された場合、コールバックが示すメソッドは測定対象メソッドであるため、対象メソッド記録部１１８は、コールバックに含まれるメソッドを示す情報と、コールバックに含まれる時刻（すなわち、メソッドの開始時刻）とを第１実行情報テーブル１１４に格納する（１５１８）。 If it is determined in step 1517 that the method indicated by the callback is stored in the target method table 113, the target method recording unit 118 is included in the callback because the method indicated by the callback is the measurement target method. Information indicating the method and the time included in the callback (that is, the start time of the method) are stored in the first execution information table 114 (1518).
　ステップ１５１８の後、対象メソッド記録部１１８は、コールバックが示すメソッドが実行されるスレッドのメソッド実行中フラグに、"ＴＲＵＥ"を格納する。また、対象メソッド記録部１１８は、コールバックに含まれる、メソッドのコールスタックにおけるフレーム番号を、メソッドが実行されるスレッドの測定対象フレームに格納する（１５１９）。 After step 1518, the target method recording unit 118 stores “TRUE” in the method execution flag of the thread in which the method indicated by the callback is executed. Further, the target method recording unit 118 stores the frame number in the call stack of the method included in the callback in the measurement target frame of the thread in which the method is executed (1519).
　測定対象フレームは、対象メソッド記録部１１８によって保持される、フレーム番号を一時的に格納するための記憶領域である。測定対象フレームに格納される値は、測定対象メソッドのフレーム番号を示し、スレッドごとに保持される。 The measurement target frame is a storage area for temporarily storing the frame number held by the target method recording unit 118. The value stored in the measurement target frame indicates the frame number of the measurement target method and is held for each thread.
　ステップ１５１９の後、対象メソッド記録部１１８は、ステップ１５２６に移行し、コールバック受信部１１９に処理の終了を通知する。 After step 1519, the target method recording unit 118 proceeds to step 1526 and notifies the callback receiving unit 119 of the end of processing.
　ステップ１５１６において、コールバックが示すメソッドが実行されるスレッドのメソッド実行中フラグに"ＴＲＵＥ"が格納されていると判定された場合、ステップ１５１４において受信されたコールバックはメソッドの終了を示すため、対象メソッド記録部１１８は、終了時刻を格納するためステップ１５２０に移行する。 If it is determined in step 1516 that “TRUE” is stored in the method execution flag of the thread in which the method indicated by the callback is executed, the callback received in step 1514 indicates the end of the method. The target method recording unit 118 proceeds to Step 1520 to store the end time.
　対象メソッド記録部１１８は、ステップ１５１６の後、コールバックが示すメソッドが実行されるスレッドの測定対象フレームに格納される番号が、コールバックが示すメソッドのフレーム番号と同じか否かを判定する（１５２０）。 After step 1516, the target method recording unit 118 determines whether the number stored in the measurement target frame of the thread in which the method indicated by the callback is executed is the same as the frame number of the method indicated by the callback ( 1520).
　測定対象フレームに格納される番号が、コールバックが示すメソッドのフレーム番号と同じである場合、コールバックが示すメソッドは、テスト実行中に開始され、ステップ１５１８において開始時刻を格納されたメソッドである。このため、対象メソッド記録部１１８は、ステップ１５２１に移行する。 When the number stored in the measurement target frame is the same as the frame number of the method indicated by the callback, the method indicated by the callback is the method that is started during the execution of the test and the start time is stored in step 1518. . Therefore, the target method recording unit 118 proceeds to Step 1521.
　ステップ１５２０の後、対象メソッド記録部１１８は、コールバックに含まれるメソッドを示す情報と、コールバックに含まれる時刻（すなわち、メソッドの終了時刻）とを第１実行情報テーブル１１４に格納する（１５２１）。ステップ１５２１の後、対象メソッド記録部１１８は、メソッド実行中フラグに"ＦＡＬＳＥ"を格納する（１５２２）。 After step 1520, the target method recording unit 118 stores information indicating the method included in the callback and the time included in the callback (that is, the end time of the method) in the first execution information table 114 (1521). ). After step 1521, the target method recording unit 118 stores “FALSE” in the method executing flag (1522).
　ステップ１５２２の後、対象メソッド記録部１１８は、ステップ１５２６に移行し、コールバック受信部１１９に処理の終了を通知する。 After step 1522, the target method recording unit 118 proceeds to step 1526 and notifies the callback receiving unit 119 of the end of processing.
　ステップ１５２０において、コールバックが示すメソッドが実行されるスレッドの測定対象フレームに格納される番号が、コールバックが示すメソッドのフレーム番号と異なる場合、コールバックが示すメソッドは、まだ終了していない測定対象メソッドによって呼ばれたメソッド、又は、測定対象メソッドによって呼ばれたメソッドからさらに呼ばれたメソッドである。このため、対象メソッド記録部１１８は、ステップ１５２３に移行する。 In step 1520, when the number stored in the measurement target frame of the thread in which the method indicated by the callback is executed is different from the frame number of the method indicated by the callback, the method indicated by the callback is not yet completed. A method called by a target method or a method further called from a method called by a measurement target method. Therefore, the target method recording unit 118 proceeds to Step 1523.
　ステップ１５２０の後、対象メソッド記録部１１８は、測定対象フレームに格納される値が、コールバックが示すメソッドのフレーム番号から１を減算した値か否かを判定する。すなわち、コールバックが示すメソッドが、測定対象メソッドによって直接呼ばれたメソッドであるか否かを判定する（１５２３）。 After step 1520, the target method recording unit 118 determines whether or not the value stored in the measurement target frame is a value obtained by subtracting 1 from the frame number of the method indicated by the callback. That is, it is determined whether or not the method indicated by the callback is a method directly called by the measurement target method (1523).
　測定対象フレームに格納される値が、コールバックが示すメソッドのフレーム番号から１を減算した値ではない場合、コールバックが示すメソッドは次の測定対象メソッドではないため、対象メソッド記録部１１８は、ステップ１５２６に移行する。 When the value stored in the measurement target frame is not a value obtained by subtracting 1 from the frame number of the method indicated by the callback, the method indicated by the callback is not the next measurement target method. Control goes to step 1526.
　測定対象フレームに格納される値が、コールバックが示すメソッドのフレーム番号から１を減算した値である場合、対象メソッド記録部１１８は、コールバックがメソッドの開始を示すか否かを判定する（１５２４）。ステップ１５２４において、コールバックがメソッドの開始を示さず、終了を示す場合、対象メソッド記録部１１８は、ステップ１６２６に移行する。 When the value stored in the measurement target frame is a value obtained by subtracting 1 from the frame number of the method indicated by the callback, the target method recording unit 118 determines whether or not the callback indicates the start of the method ( 1524). If the callback does not indicate the start of the method but indicates the end in step 1524, the target method recording unit 118 proceeds to step 1626.
　ステップ１５２４において、コールバックがメソッドの開始を示す場合、コールバックが示すメソッドは次の測定対象メソッドであるため、対象メソッド記録部１１８は、コールバックに含まれるメソッドを示す情報を、第１メソッド呼出テーブル１１１に格納する（１５２５）。 In step 1524, when the callback indicates the start of the method, the method indicated by the callback is the next measurement target method. Therefore, the target method recording unit 118 stores information indicating the method included in the callback as the first method. Store in the call table 111 (1525).
　ここで、第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２に格納された値は、図１２Ａに示すテストが実行される前に削除されていてもよい。すなわち、ステップ１５２５によって、次の測定対象メソッドになる可能性があるメソッドのみが第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２に格納されるように、第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２に含まれる前回の初期メソッド又は測定対象メソッドを示す値は、図１２Ａに示すテスト開始時に削除されてもよい。 Here, the values stored in the first method call table 111 and the second method call table 112 may be deleted before the test shown in FIG. 12A is executed. That is, the first method call table 111 and the second method are stored in the first method call table 111 and the second method call table 112 by the step 1525 so that only methods that may become the next measurement target methods are stored in the first method call table 111 and the second method call table 112. The value indicating the previous initial method or measurement target method included in the call table 112 may be deleted at the start of the test illustrated in FIG. 12A.
　その後、対象メソッド記録部１１８は、ステップ１５２６に移行する。 Thereafter, the target method recording unit 118 proceeds to Step 1526.
　ステップ１５２６において、コールバック受信部１１９は、次のコールバックの受信まで待機する。 In step 1526, the callback receiver 119 waits until the next callback is received.
　図１２Ｂに示す処理によって、対象メソッド記録部１１８は、測定対象メソッドの実行時間を測定し、次の測定対象メソッドを抽出する。 12B, the target method recording unit 118 measures the execution time of the measurement target method and extracts the next measurement target method.
　図１３は、本発明の第１の実施形態の性能差要因メソッドを抽出する処理を示すフローチャートである。 FIG. 13 is a flowchart showing processing for extracting a performance difference factor method according to the first embodiment of this invention.
　図１３に示す処理は、図７に示すステップ１６０に相当する。ステップ１５０、すなわち、図１２Ａに示す処理が行われる毎に、ステップ１６０は実行される。 The process shown in FIG. 13 corresponds to step 160 shown in FIG. Step 160 is executed each time step 150, that is, the process shown in FIG. 12A is performed.
　制御部１０６は、性能差要因メソッドを抽出する指示を性能差要因メソッド判定部１０９に送信する（１６０１）。 The control unit 106 transmits an instruction to extract the performance difference factor method to the performance difference factor method determination unit 109 (1601).
　性能差要因メソッド判定部１０９は、性能差要因メソッドを抽出する指示を制御部１０６から受信した後、変数ｉと最大性能差要因率とを格納する一時的な記憶領域をメモリ１０２に保持する。そして、変数ｉに、１を格納し、最大性能差要因率に、０を格納する（１６０２）。 The performance difference factor method determination unit 109 receives an instruction to extract a performance difference factor method from the control unit 106, and then holds a temporary storage area in which the variable i and the maximum performance difference factor rate are stored in the memory 102. Then, 1 is stored in the variable i, and 0 is stored in the maximum performance difference factor rate (1602).
　変数ｉは、対象メソッドテーブル１１３の行を示すポインタであり、最大性能差要因率は、後述する性能差要因率の最大値を示す。 The variable i is a pointer indicating the row of the target method table 113, and the maximum performance difference factor rate indicates the maximum value of the performance difference factor rate described later.
　ステップ１６０２の後、性能差要因メソッド判定部１０９は、ステップ１６０４からステップ１６１３までの処理を、変数ｉに格納された値が対象メソッドテーブル１１３の行数を超えるまで、繰り返す。 After step 1602, the performance difference factor method determination unit 109 repeats the processing from step 1604 to step 1613 until the value stored in the variable i exceeds the number of rows in the target method table 113.
　ステップ１６０２の後、性能差要因メソッド判定部１０９は、第１プログラム１２０における測定対象メソッドの平均実行時間１１６３１と実行回数１１６３２とを第１実行情報テーブル１１４に基づいて、算出する。 After step 1602, the performance difference factor method determination unit 109 calculates the average execution time 11631 and the number of executions 11632 of the measurement target method in the first program 120 based on the first execution information table 114.
　具体的には、性能差要因メソッド判定部１０９は、対象メソッドテーブル１１３から測定対象メソッドを一つ抽出し、出力装置１０４のうち、メソッド情報１１４２の値が測定対象メソッドを示す行を抽出する。そして、抽出された行からさらに、開始終了情報１１４３が"開始"と"終了"とを各々示す連続した二つの行を抽出する。そして、抽出された二つの行の組の数を算出することによって、実行回数１１６３２を算出する。 Specifically, the performance difference factor method determination unit 109 extracts one measurement target method from the target method table 113, and extracts a line in the output device 104 where the value of the method information 1142 indicates the measurement target method. Then, two consecutive lines whose start / end information 1143 indicates “start” and “end” are extracted from the extracted lines. Then, the number of executions 11632 is calculated by calculating the number of sets of two extracted rows.
　また性能差要因メソッド判定部１０９は、ステップ１６０４において、抽出された二つの行の時刻１１４４の差を算出することによって、１回あたりのメソッドの実行時間を算出する。そして、抽出された二つの行の組の各々の実行時間を算出し、すべての組の実行時間の平均値を算出することによって、平均実行時間１１６３１を算出する。 Further, in step 1604, the performance difference factor method determination unit 109 calculates a method execution time per time by calculating a difference between the times 1144 of the two extracted rows. Then, the average execution time 11631 is calculated by calculating the execution time of each of the two sets of extracted rows and calculating the average value of the execution times of all the sets.
　ステップ１６０４の後、性能差要因メソッド判定部１０９は、第２プログラム１２１における測定対象メソッドの平均実行時間１１６４１と実行回数１１６４２とを第２実行情報テーブル１１５に基づいて、算出する（１６０５）。平均実行時間１１６４１と実行回数１１６４２の具体的な算出方法は、平均実行時間１１６３１と実行回数１１６３２の算出方法と同様であるが、平均実行時間１１６４１と実行回数１１６４２は、第１実行情報テーブル１１４の代わりに第２実行情報テーブル１１５に基づいて算出される。 After step 1604, the performance difference factor method determination unit 109 calculates the average execution time 11641 and the number of executions 11642 of the measurement target method in the second program 121 based on the second execution information table 115 (1605). The specific calculation method of the average execution time 11641 and the execution count 11642 is the same as the calculation method of the average execution time 11631 and the execution count 11632, but the average execution time 11641 and the execution count 11642 are the same as those in the first execution information table 114. Instead, it is calculated based on the second execution information table 115.
　ステップ１６０５の後、性能差要因メソッド判定部１０９は、実行時間差１１６５及び実行時間比１１６６を算出する（ステップ１６０６）。具体的には、性能差要因メソッド判定部１０９は、ステップ１６０５において算出された各組の実行時間の合計を実行回数１１６３２によって除算した値から、ステップ１６０４において算出された各組の実行時間の合計を実行回数１１６４２によって除算した値を減算することによって、実行時間差１１６５を算出する。 After step 1605, the performance difference factor method determination unit 109 calculates an execution time difference 1165 and an execution time ratio 1166 (step 1606). Specifically, the performance difference factor method determination unit 109 calculates the total execution time of each group calculated in step 1604 from the value obtained by dividing the total execution time of each group calculated in step 1605 by the execution count 11632. Is subtracted by the number of executions 11642 to calculate the execution time difference 1165.
　また、性能差要因メソッド判定部１０９は、ステップ１６０６において、算出された実行時間差１１６５を、さらに実行回数１１６３２によって除算し、除算された結果を、さらに平均実行時間１１６３１によって除算することによって、実行時間比１１６６を算出する。 In step 1606, the performance difference factor method determination unit 109 further divides the calculated execution time difference 1165 by the execution count 11632, and further divides the divided result by the average execution time 11631, thereby executing the execution time. The ratio 1166 is calculated.
　ステップ１６０６の後、性能差要因メソッド判定部１０９は、ステップ１６０４、ステップ１６０５、及び、ステップ１６０６において算出された平均実行時間１１６３１、実行回数１１６３２、平均実行時間１１６４１、実行回数１１６４２、実行時間差１１６５、及び、実行時間比１１６６を、測定結果テーブル１１６に格納する（１６０７）。 After step 1606, the performance difference factor method determination unit 109 calculates the average execution time 11631, the execution count 11632, the average execution time 11641, the execution count 11642, the execution time difference 1165, calculated in step 1604, step 1605, and step 1606. The execution time ratio 1166 is stored in the measurement result table 116 (1607).
　なお、ステップ１６０４からステップ１６０７の処理は、対象メソッドテーブル１１３に含まれる全てのメソッドに対して実施されてもよい。 Note that the processing from step 1604 to step 1607 may be performed for all methods included in the target method table 113.
　ステップ１６０７の後、性能差要因メソッド判定部１０９は、対象メソッドテーブル１１３の変数ｉが示す行のメソッド情報１１３２に対応する測定結果テーブル１１６のメソッド情報１１６２と、実行時間差１１６５と、実行時間比１１６６とを抽出する（１６０８）。 After step 1607, the performance difference factor method determination unit 109 executes the method information 1162 of the measurement result table 116 corresponding to the method information 1132 in the row indicated by the variable i of the target method table 113, the execution time difference 1165, and the execution time ratio 1166. Are extracted (1608).
　ステップ１６０８の後、性能差要因メソッド判定部１０９は、ステップ１６０８において抽出された実行時間比１１６６が、あらかじめユーザによって指定された実行時間比閾値以上であるか否かを判定する（１６０９）。実行時間比１１６６が実行時間比閾値以上でない場合、第１プログラム１２０と第２プログラム１２１とにおける測定対象メソッドの性能差は、ユーザにとって問題ない範囲であるため、性能差要因メソッド判定部１０９は、ステップ１６１３に移行する。 After step 1608, the performance difference factor method determination unit 109 determines whether or not the execution time ratio 1166 extracted in step 1608 is equal to or greater than the execution time ratio threshold specified in advance by the user (1609). If the execution time ratio 1166 is not equal to or greater than the execution time ratio threshold, the performance difference between the measurement target methods in the first program 120 and the second program 121 is in a range that is not problematic for the user. Control goes to step 1613.
　実行時間比１１６６が実行時間比閾値以上である場合、第１プログラム１２０と第２プログラム１２１とにおける測定対象メソッドの性能差は、ユーザにとって問題がある範囲であるため、性能差要因メソッド判定部１０９は、ステップ１６１０に移行する。 When the execution time ratio 1166 is equal to or greater than the execution time ratio threshold, the performance difference between the measurement target methods in the first program 120 and the second program 121 is in a range that causes a problem for the user. Shifts to step 1610.
　ステップ１６０９の後、性能差要因メソッド判定部１０９は、変数ｉが示す行の呼出元メソッド情報１１３３を対象メソッドテーブル１１３から抽出し、抽出された呼出元メソッド情報１１３３に対応するメソッド情報１１６２の実行時間差１１６５を、測定結果テーブル１１６から抽出する。そして、性能差要因メソッド判定部１０９は、抽出された実行時間差１１６５を、呼出元実行時間差に格納する（１６１０）。呼出元実行時間差とは、性能差要因メソッド判定部１０９によって保持される、メモリ１０２の中の一部の記憶領域である。 After step 1609, the performance difference factor method determination unit 109 extracts the caller method information 1133 of the row indicated by the variable i from the target method table 113, and executes the method information 1162 corresponding to the extracted caller method information 1133. The time difference 1165 is extracted from the measurement result table 116. Then, the performance difference factor method determination unit 109 stores the extracted execution time difference 1165 in the caller execution time difference (1610). The caller execution time difference is a partial storage area in the memory 102 held by the performance difference factor method determination unit 109.
　ステップ１６１０の後、性能差要因メソッド判定部１０９は、ステップ１６０８において抽出された実行時間差１１６５を、呼出元実行時間差によって除算する。そして、除算された結果を、性能差要因率に格納する（１６１１）。 After step 1610, the performance difference factor method determination unit 109 divides the execution time difference 1165 extracted in step 1608 by the caller execution time difference. The divided result is stored in the performance difference factor rate (1611).
　性能差要因率とは、性能差要因メソッド判定部１０９によって保持される、メモリ１０２の中の一部の記憶領域である。性能差要因率に格納される値が大きい程、呼出元メソッドの性能差から変化が大きいことを示す。すなわち、呼出元メソッドの性能差よりも、性能差が大きいメソッドは、第１プログラム１２０と第２プログラム１２１との性能差を生じさせるメソッドである可能性が高いため、性能差要因率が大きいほど、性能差を生じさせるメソッドである可能性が高い。 The performance difference factor rate is a partial storage area in the memory 102 held by the performance difference factor method determination unit 109. The larger the value stored in the performance difference factor rate, the greater the change from the performance difference of the caller method. That is, a method having a larger performance difference than that of the caller method is likely to be a method that causes a performance difference between the first program 120 and the second program 121. The method is likely to cause a performance difference.
　ステップ１６１１の後、性能差要因メソッド判定部１０９は、ステップ１６１１において算出された性能差要因率が示す値よりも、最大性能差要因率が示す値のほうが大きいか否かを判定する（１６１２）。 After step 1611, the performance difference factor method determination unit 109 determines whether or not the value indicated by the maximum performance difference factor rate is greater than the value indicated by the performance difference factor rate calculated in step 1611 (1612). .
　最大性能差要因率とは、性能差要因メソッド判定部１０９によって保持される、メモリ１０２の中の一部の記憶領域である。最大性能差要因率は、対象メソッドテーブル１１３の変数ｉが示す各行に、ステップ１６０８～ステップ１６１１を実行することによって算出される性能差要因率の中で、最も大きい値を示す性能差要因率の値を格納するための領域である。 The maximum performance difference factor rate is a partial storage area in the memory 102 held by the performance difference factor method determination unit 109. The maximum performance difference factor rate is the value of the performance difference factor rate showing the largest value among the performance difference factor rates calculated by executing Steps 1608 to 1611 in each row indicated by the variable i of the target method table 113. This is an area for storing values.
　ステップ１６１２において、ステップ１６１１において算出された性能差要因率が示す値より、最大性能差要因率が示す値は大きくない場合、最大性能差要因率が示す値は変更されないため、性能差要因メソッド判定部１０９は、ステップ１６１３に移行する。 In step 1612, if the value indicated by the maximum performance difference factor rate is not greater than the value indicated by the performance difference factor rate calculated in step 1611, the value indicated by the maximum performance difference factor rate is not changed. The unit 109 proceeds to step 1613.
　ステップ１６１２において、ステップ１６１１において算出された性能差要因率が示す値よりも、最大性能差要因率が示す値が大きい場合、性能差要因メソッド判定部１０９は、ステップ１６１１において算出された性能差要因率によって最大性能差要因率を更新する。また、変数ｉが示す対象メソッドテーブル１１３の行のメソッド情報１１３２の値によって、性能差要因メソッドを更新する（１６１３）。ステップ１６１２及びステップ１６１３の処理によって、性能差要因メソッドが抽出される。 In step 1612, when the value indicated by the maximum performance difference factor rate is greater than the value indicated by the performance difference factor rate calculated in step 1611, the performance difference factor method determination unit 109 determines that the performance difference factor calculated in step 1611 The maximum performance difference factor rate is updated according to the rate. Also, the performance difference factor method is updated with the value of the method information 1132 in the row of the target method table 113 indicated by the variable i (1613). Through the processing in step 1612 and step 1613, a performance difference factor method is extracted.
　ステップ１６０９、ステップ１６１２、又は、ステップ１６１３の後、性能差要因メソッド判定部１０９は、変数ｉに格納される値に、１を加算する。すなわち、ステップ１６０４～ステップ１６１３に示す処理を行う対象メソッドテーブル１１３の行を、一つ移動させる（１６１３）。 After step 1609, step 1612, or step 1613, the performance difference factor method determination unit 109 adds 1 to the value stored in the variable i. That is, one row of the target method table 113 for performing the processing shown in steps 1604 to 1613 is moved (1613).
　対象メソッドテーブル１１３のすべての行に、ステップ１６０４～ステップ１６１３の処理を行った後、性能差要因メソッド判定部１０９は、性能差要因メソッドの抽出処理が終了したことを制御部１０６に通知する（１６１４）。 After performing the processing of step 1604 to step 1613 for all the rows of the target method table 113, the performance difference factor method determination unit 109 notifies the control unit 106 that the extraction process of the performance difference factor method has ended ( 1614).
　なお、図１３のステップ１６０９に示す処理において、実行時間比１１６６を算出することによって、第１のプログラムと第２のプログラムとの性能差を比較したが、実行時間差１１６５を比較することによって性能差を比較してもよい。すなわち、呼出元メソッドからの呼出回数が一定している場合などは、実行時間差１１６５によって性能差を比較してもよい。 In the processing shown in Step 1609 of FIG. 13, the performance difference between the first program and the second program is compared by calculating the execution time ratio 1166, but the performance difference is determined by comparing the execution time difference 1165. May be compared. That is, when the number of calls from the caller method is constant, the performance difference may be compared by the execution time difference 1165.
　図１３のに示す処理、すなわち、図７に示すステップ１６０が終了した後、性能差要因メソッドに値が格納されている場合、図１４に示す処理、すなわち、ステップ１８０が実行される。 After the process shown in FIG. 13, that is, step 160 shown in FIG. 7 is completed, if the value is stored in the performance difference factor method, the process shown in FIG. 14, that is, step 180 is executed.
　図１４は、本発明の第１の実施形態の測定対象メソッドを再度設定する処理を示すフローチャートである。 FIG. 14 is a flowchart showing processing for setting the measurement target method again according to the first embodiment of this invention.
　図１４に示す処理は、図７のステップ１８０に示す処理に相当する。 The process shown in FIG. 14 corresponds to the process shown in Step 180 of FIG.
　ステップ１７０の後、制御部１０６は、測定対象メソッドを再度設定する指示を、メソッド呼出照合部１０８に送信する（１８０１）。 After step 170, the control unit 106 transmits an instruction to set the measurement target method again to the method call matching unit 108 (1801).
　ステップ１８０１の後、メソッド呼出照合部１０８は、第１メソッド呼出テーブル１１１を読み込み（１８０２）、第２メソッド呼出テーブル１１２を読み込む（１８０３）。そして、対象メソッドテーブル１１３に格納された値を削除する（１８０４）。 After step 1801, the method call verification unit 108 reads the first method call table 111 (1802) and reads the second method call table 112 (1803). Then, the value stored in the target method table 113 is deleted (1804).
　ここで、第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２には、図１２Ｂのステップ１５２５において格納されたメソッド情報１１３２が保持される。すなわち、次回の測定対象メソッドになる可能性があるメソッドを示す情報が保持される。 Here, the method information 1132 stored in step 1525 of FIG. 12B is held in the first method call table 111 and the second method call table 112. That is, information indicating a method that may become the next measurement target method is held.
　ステップ１８０４の後、メソッド呼出照合部１０８は、第１メソッド呼出テーブル１１１と第２メソッド呼出テーブル１１２とに共通して保持されるメソッド情報（１１１２及び１１２２）のうち、呼出元メソッド情報（１１１３及び１１２３）が示すメソッドが、図１３に示す処理によって抽出された性能差要因メソッドであるメソッド情報（１１１２及び１１２２）を、対象メソッドテーブル１１３に格納する（１８０５）。 After step 1804, the method call verification unit 108 calls the source method information (1113 and 1122) out of the method information (1112 and 1122) held in common in the first method call table 111 and the second method call table 112. The method indicated by 1123) stores the method information (1112 and 1122) that is the performance difference factor method extracted by the processing shown in FIG. 13 in the target method table 113 (1805).
　ステップ１８０５の後、メソッド呼出照合部１０８は、測定対象メソッドの再設定の終了を制御部１０６に通知する（１８０６）。 After step 1805, the method call matching unit 108 notifies the control unit 106 of the end of resetting the measurement target method (1806).
　図１４に示す処理の後、図７に示すステップ１５０からステップ１８０までの処理が、性能差要因メソッドが抽出されなくなるまで、繰り返される。 After the process shown in FIG. 14, the process from step 150 to step 180 shown in FIG. 7 is repeated until no performance difference factor method is extracted.
　第１の実施形態によれば、例えば、性能に影響のある修正が加えられたプログラムと、修正前のプログラムとの性能を比較することができる。また、修正前のプログラムの実行時間と修正後のプログラムの実行時間とに大きな差がない場合も、メソッドごとに実行時間を比較することによって、性能に著しい差が生じているメソッドを抽出できる。 According to the first embodiment, for example, it is possible to compare the performance of a program to which a modification that affects performance is added and a program before the modification. Further, even when there is no significant difference between the execution time of the program before correction and the execution time of the program after correction, it is possible to extract a method having a significant difference in performance by comparing the execution times for each method.
　また第１の実施形態によれば、例えば、アプリケーションサーバの分野においては、アプリケーションサーバのバージョンを固定すれば、２つのユーザアプリケーション間の性能差要因を分析でき、ユーザアプリケーションを固定すれば、アプリケーションサーバのバージョン間の性能差要因を分析できる。 According to the first embodiment, for example, in the field of application servers, if the version of the application server is fixed, the performance difference factor between the two user applications can be analyzed, and if the user application is fixed, the application server Can analyze the performance difference between different versions.
　さらに、第１の実施形態によれば、実行時間を測定するメソッドを抽出するため、実行時間の測定による負荷への影響が軽減され、性能に問題のある処理が含まれるメソッドを高い精度で検出することができる。 Furthermore, according to the first embodiment, since the method for measuring the execution time is extracted, the influence on the load due to the measurement of the execution time is reduced, and a method including a process having a problem in performance is detected with high accuracy. can do.
　図１５は、本発明の第２の実施形態の単一のプログラムの性能差を分析する計算機１００の構成を示すブロック図である。 FIG. 15 is a block diagram illustrating a configuration of the computer 100 that analyzes a performance difference of a single program according to the second embodiment of this invention.
　第２の実施形態の計算機１００は、一つのプログラムのテストを実行し、一つのプログラムにおける性能を測定する。そして、ユーザによってあらかじめ指定された閾値と、測定された性能を比較し、性能差要因メソッドを抽出する。 The computer 100 of the second embodiment executes a test of one program and measures the performance of one program. Then, the threshold specified in advance by the user is compared with the measured performance, and a performance difference factor method is extracted.
　第２の実施形態の計算機１００は、第１の実施形態の計算機１００と同じ、メモリ１０２、演算装置１０１、外部記憶装置１０３、出力装置１０４、及び、入力装置１０５を備える。 The computer 100 of the second embodiment includes the same memory 102, arithmetic device 101, external storage device 103, output device 104, and input device 105 as the computer 100 of the first embodiment.
　また、第２の実施形態の制御部３００、テスト実行部３０１、メソッド呼出照合部３０２、性能差要因メソッド判定部３０３、初期メソッド記録部３０８、対象メソッド記録部３０９、コールバック受信部３１０、ＪａｖａＶＭ３１２、及び、ＪＶＭＴＩ３１３は、第１の実施形態の制御部１０６、テスト実行部１０７、メソッド呼出照合部１０８、性能差要因メソッド判定部１０９、バージョン切替部１１０、初期メソッド記録部１１７、対象メソッド記録部１１８、コールバック受信部１１９、ＪａｖａＶＭ１２２、及び、ＪＶＭＴＩ１２３と各々同様な機能を持つ。また、第２の実施形態の対象メソッドテーブル３０５及び測定結果テーブル３０７は、第１の実施形態の対象メソッドテーブル１１３、測定結果テーブル１１６と同じ情報を含む。 In addition, the control unit 300, the test execution unit 301, the method call verification unit 302, the performance difference factor method determination unit 303, the initial method recording unit 308, the target method recording unit 309, the callback receiving unit 310, and the Java VM 312 according to the second embodiment. The JVMTI 313 includes a control unit 106, a test execution unit 107, a method call matching unit 108, a performance difference factor method determination unit 109, a version switching unit 110, an initial method recording unit 117, and a target method recording unit according to the first embodiment. 118, the callback receiver 119, the JavaVM 122, and the JVMTI 123 have the same functions. In addition, the target method table 305 and the measurement result table 307 of the second embodiment include the same information as the target method table 113 and the measurement result table 116 of the first embodiment.
　しかし、第２の実施形態の制御部３００は、一つの測定対象プログラム３１１のみをテストするため、第１メソッド呼出テーブル１１１に対応するメソッド呼出テーブル３０４と、第１実行情報テーブル１１４に対応する実行情報テーブル３０６は、各々一つである。 However, since the control unit 300 according to the second embodiment tests only one measurement target program 311, a method call table 304 corresponding to the first method call table 111 and an execution corresponding to the first execution information table 114 are performed. There is one information table 306 for each.
　図１６は、本発明の第２の実施形態のメソッド呼出テーブル３０４を示す説明図である。 FIG. 16 is an explanatory diagram illustrating a method call table 304 according to the second embodiment of this invention.
　メソッド呼出テーブル３０４は、第１の実施形態の第１メソッド呼出テーブル１１１及び第２メソッド呼出テーブル１１２と同じ情報を含み、フレーム番号３０４１、メソッド情報３０４２及び呼出元メソッド情報３０４３は、第１の実施形態のフレーム番号１１１１、メソッド情報１１１２及び呼出元メソッド情報１１１３に対応する。 The method call table 304 includes the same information as the first method call table 111 and the second method call table 112 of the first embodiment. The frame number 3041, the method information 3042, and the caller method information 3043 are the same as those in the first embodiment. This corresponds to the frame number 1111, method information 1112, and caller method information 1113 of the form.
　図１７は、本発明の第２の実施形態の実行情報テーブル３０６を示す説明図である。 FIG. 17 is an explanatory diagram illustrating an execution information table 306 according to the second embodiment of this invention.
　実行情報テーブル３０６は、第１の実施形態の第１実行情報テーブル１１４及び第２実行情報テーブル１１５と同じ情報を含み、フレーム番号３０６１、メソッド情報３０６２、開始終了情報３０６３及び時刻３０６４は、第１の実施形態のフレーム番号１１４１、メソッド情報１１４２、開始終了情報１１４３及び時刻１１４４に対応する。 The execution information table 306 includes the same information as the first execution information table 114 and the second execution information table 115 of the first embodiment, and the frame number 3061, method information 3062, start / end information 3063, and time 3064 are the first information. Corresponds to the frame number 1141, method information 1142, start / end information 1143, and time 1144.
　図１８は、本発明の第２の実施形態の測定結果テーブル３０７を示す説明図である。 FIG. 18 is an explanatory diagram illustrating a measurement result table 307 according to the second embodiment of this invention.
　測定結果テーブル３０７は、フレーム番号３０７１、メソッド情報３０７２、平均実行時間３０７３及び実行回数３０７４を含む。フレーム番号３０７１、メソッド情報３０７２、平均実行時間３０７３及び実行回数３０７４は、第１の実施形態の測定結果テーブル１１６のフレーム番号１１６１、メソッド情報１１６２、平均実行時間１１６３１及び実行回数１１６３２に対応する。 The measurement result table 307 includes a frame number 3071, method information 3072, an average execution time 3073, and an execution count 3074. The frame number 3071, the method information 3072, the average execution time 3073, and the execution count 3074 correspond to the frame number 1161, the method information 1162, the average execution time 11631, and the execution count 11632 in the measurement result table 116 of the first embodiment.
　以下に、第１の実施形態の処理と異なる処理について説明する。 Hereinafter, processing different from the processing of the first embodiment will be described.
　図１９は、本発明の第２の実施形態の全体の処理を示すフローチャートである。 FIG. 19 is a flowchart showing the overall processing of the second embodiment of the present invention.
　制御部３００は、まず、測定対象プログラム３１１の初期メソッドを抽出し、抽出された初期メソッドを対象メソッドテーブル３０５に格納する。これによって、測定対象メソッドが対象メソッドテーブル３０５に格納される（３２０）。 The control unit 300 first extracts the initial method of the measurement target program 311 and stores the extracted initial method in the target method table 305. As a result, the measurement target method is stored in the target method table 305 (320).
　制御部３００は、ステップ３２０において、図１０Ａ～図１０Ｃに示す処理を、測定対象プログラム３１１のみに実行する。具体的には、制御部３００は、図１０Ａに示すステップ１４０２からステップ１４０９までをテスト実行部３０１に実行させることによって、測定対象プログラム３１１のテストを実行する。図１０Ａに示す処理が実行される際、制御部３００は、コールバック受信部３１０及び初期メソッド記録部３０８によって、図１０Ｃに示す処理を測定対象プログラム３１１のみに実行する。 In step 320, the control unit 300 executes the processing shown in FIGS. 10A to 10C only for the measurement target program 311. Specifically, the control unit 300 executes the test of the measurement target program 311 by causing the test execution unit 301 to execute steps 1402 to 1409 shown in FIG. 10A. When the process illustrated in FIG. 10A is executed, the control unit 300 causes the callback receiver 310 and the initial method recording unit 308 to execute the process illustrated in FIG.
　さらにステップ３２０において制御部３００は、メソッド呼出照合部３０２によって、図１０Ｂに示すステップ１４２１、ステップ１４２３及びステップ１４２４を測定対象プログラム３１１に実行する。ここで、ステップ１４２４において、メソッド呼出照合部３０２は、メソッド呼出テーブルに含まれるメソッド情報３０４２を、すべて対象メソッドテーブル３０５に格納する。 Further, in step 320, the control unit 300 causes the method call matching unit 302 to execute steps 1421, 1423, and 1424 shown in FIG. Here, in step 1424, the method call matching unit 302 stores all the method information 3042 included in the method call table in the target method table 305.
　ステップ３２０の後、制御部３００は、測定対象メソッドの実行時間を測定する（３３０）。制御部３００は、ステップ３３０において、図１２Ａ及び図１２Ｂに示す処理を測定対象プログラム３１１にのみ実行する。 After step 320, the control unit 300 measures the execution time of the measurement target method (330). In step 330, the control unit 300 executes the process shown in FIGS. 12A and 12B only for the measurement target program 311.
　具体的には、制御部３００は、ステップ３３０において、ステップ１５０２からステップ１５０７までをテスト実行部３０１に実行させることによって、測定対象プログラム３１１のテストを実行する。そして、測定対象プログラム３１１をテストしている間に、対象メソッド記録部３０９によって、図１２Ｂに示す処理を測定対象プログラム３１１に実行する。 Specifically, the control unit 300 executes the test of the measurement target program 311 by causing the test execution unit 301 to execute steps 1502 to 1507 in step 330. Then, while the measurement target program 311 is being tested, the target method recording unit 309 executes the process shown in FIG.
　ステップ３３０の後、制御部３００は、所定の条件を満たすメソッドを抽出する。 After step 330, the control unit 300 extracts a method that satisfies a predetermined condition.
　図２０は、本発明の第２の実施形態の所定の条件を満たすメソッドを抽出する処理を示すフローチャートである。 FIG. 20 is a flowchart showing a process of extracting a method that satisfies a predetermined condition according to the second embodiment of the present invention.
　図２０に示す処理は、図１９に示すステップ３４０に相当する。 The process shown in FIG. 20 corresponds to step 340 shown in FIG.
　制御部３００は、所定の条件を満たすメソッドを抽出する指示を性能差要因メソッド判定部３０３に送信する（３４０１）。 The control unit 300 transmits an instruction to extract a method satisfying a predetermined condition to the performance difference factor method determination unit 303 (3401).
　ステップ３４０１の後、性能差要因メソッド判定部３０３は、変数ｉをメモリ１０２に保持し、変数ｉに１を格納する（３４０２）。変数ｉは、対象メソッドテーブル３０５の行を示すポインタである。 After step 3401, the performance difference factor method determination unit 303 holds the variable i in the memory 102 and stores 1 in the variable i (3402). A variable i is a pointer indicating a row of the target method table 305.
　ステップ３４０２の後、性能差要因メソッド判定部３０３は、ステップ３４０３からステップ３４０８までの処理を、変数ｉに格納された値が対象メソッドテーブル３０５の行数を超えるまで、繰り返す（３４０２）。 After step 3402, the performance difference factor method determination unit 303 repeats the processing from step 3403 to step 3408 until the value stored in the variable i exceeds the number of rows in the target method table 305 (3402).
　ステップ１６０２の後、性能差要因メソッド判定部３０３は、測定対象メソッドの平均実行時間３０７３と実行回数３０７４とを実行情報テーブル３０６に基づいて、算出する（３４０３）。平均実行時間３０７３と実行回数３０７４との算出方法は、第１の実施形態の平均実行時間１１６３１及び実行回数１１６３２の算出方法と同じである。 After step 1602, the performance difference factor method determination unit 303 calculates the average execution time 3073 and the number of executions 3074 of the measurement target method based on the execution information table 306 (3403). The calculation method of the average execution time 3073 and the execution count 3074 is the same as the calculation method of the average execution time 11631 and the execution count 11632 of the first embodiment.
　ステップ３４０３の後、性能差要因メソッド判定部３０３は、算出された平均実行時間３０７３と実行回数３０７４とを、測定結果テーブル３０７に格納する（３４０４）。 After step 3403, the performance difference factor method determination unit 303 stores the calculated average execution time 3073 and execution count 3074 in the measurement result table 307 (3404).
　ステップ３４０４の後、性能差要因メソッド判定部３０３は、平均実行時間３０７３及び実行回数３０７４を乗じた値が、ユーザによってあらかじめ指定された閾値よりも大きいか否かを判定する（３４０５）。平均実行時間３０７３及び実行回数３０７４を乗じた値が、ユーザによってあらかじめ指定された閾値を超えない場合、変数ｉが示すメソッドには問題がないため、性能差要因メソッド判定部３０３は、ステップ３４０８に移行する。 After step 3404, the performance difference factor method determination unit 303 determines whether or not a value obtained by multiplying the average execution time 3073 and the number of executions 3074 is larger than a threshold value specified in advance by the user (3405). If the value obtained by multiplying the average execution time 3073 and the number of executions 3074 does not exceed the threshold value designated in advance by the user, there is no problem with the method indicated by the variable i. Transition.
　ステップ３４０５において、平均実行時間３０７３及び実行回数３０７４を乗じた値が、ユーザによってあらかじめ指定された閾値よりも大きいと判定された場合、変数ｉが示すメソッドに問題がある可能性があるため、性能差要因メソッド判定部３０３は、ステップ３４０６に移行する。 If it is determined in step 3405 that the value obtained by multiplying the average execution time 3073 and the number of executions 3074 is larger than the threshold value specified in advance by the user, there is a possibility that there is a problem in the method indicated by the variable i. The difference factor method determination unit 303 proceeds to step 3406.
　ステップ３４０５の後、性能差要因メソッド判定部３０３は、平均実行時間３０７３及び実行回数３０７４を乗じた値が、最大実行時間メソッドの平均実行時間３０７３及び実行回数３０７４を乗じた値よりも大きいか否かを判定する（３４０７）。最大実行時間メソッドは、性能差要因メソッド判定部３０３によってメモリ１０２に保持される記憶領域であり、所定の条件を満たすメソッドのうち、平均実行時間３０７３及び実行回数３０７４を乗じた値が最も大きいメソッドを示す情報を格納するための領域である。 After step 3405, the performance difference factor method determination unit 303 determines whether the value obtained by multiplying the average execution time 3073 and the execution count 3074 is greater than the value obtained by multiplying the average execution time 3073 and the execution count 3074 of the maximum execution time method. It is determined (3407). The maximum execution time method is a storage area held in the memory 102 by the performance difference factor method determination unit 303, and among the methods satisfying a predetermined condition, the method having the largest value obtained by multiplying the average execution time 3073 and the execution count 3074 This is an area for storing information indicating.
　平均実行時間３０７３及び実行回数３０７４を乗じた値が、最大実行時間メソッドの平均実行時間３０７３及び実行回数３０７４を乗じた値を超えない場合、性能差要因メソッド判定部３０３は、ステップ３４０９に移行する。 If the value obtained by multiplying the average execution time 3073 and the number of executions 3074 does not exceed the value obtained by multiplying the average execution time 3073 and the number of executions 3074 of the maximum execution time method, the performance difference factor method determination unit 303 proceeds to Step 3409. .
　平均実行時間３０７３及び実行回数３０７４を乗じた値が、最大実行時間メソッドの平均実行時間３０７３及び実行回数３０７４を乗じた値よりも大きい場合、変数ｉが示すメソッドによって、最大実行時間メソッドを更新する（３４０８）。 When the value obtained by multiplying the average execution time 3073 and the execution count 3074 is larger than the value obtained by multiplying the average execution time 3073 and the execution count 3074 of the maximum execution time method, the maximum execution time method is updated by the method indicated by the variable i. (3408).
　ステップ３４０６、ステップ３４０７、又は、ステップ３４０８の後、性能差要因メソッド判定部３０３は、変数ｉに格納される値に、１を加算する（３４０９）。これによって、性能差要因メソッド判定部３０３は、対象メソッドテーブル３０５の次の行を取得することができる。 After step 3406, step 3407, or step 3408, the performance difference factor method determination unit 303 adds 1 to the value stored in the variable i (3409). As a result, the performance difference factor method determination unit 303 can acquire the next row of the target method table 305.
　ステップ３４０３の繰り返し処理が終了した後、性能差要因メソッド判定部３０３は、所定の条件を満たすメソッドを抽出する処理の終了を、制御部３００に通知する（３４１０）。 After the repetition process of step 3403 is completed, the performance difference factor method determination unit 303 notifies the control unit 300 of the end of the process of extracting a method that satisfies a predetermined condition (3410).
　なお、前述の第２の実施形態において、図１０Ａ～図１０Ｃ、図１２Ａ、図１２Ｂ及び図１４に示す処理のうち、第１メソッド呼出テーブル１１１及び第１実行情報テーブル１１４は、メソッド呼出テーブル３０４及び対象メソッド記録部３０９に置き換える。 In the second embodiment described above, of the processes shown in FIGS. 10A to 10C, 12A, 12B, and 14, the first method call table 111 and the first execution information table 114 are the method call table 304. And the target method recording unit 309.
　また、本実施形態の計算機１００は、第１の実施形態と第２の実施形態との両方を実行することが可能である。すなわち、ユーザが、一つのプログラムの性能を測定するか、又は、二つのプログラムの性能を比較するかをあらかじめ指定し、制御部１０６又は制御部３００が、ユーザからの指定をに基づいて、第１の実施形態を実行するか、又は、第２の実施形態を実行するかを選択してよい。 In addition, the computer 100 according to the present embodiment can execute both the first embodiment and the second embodiment. That is, the user designates in advance whether to measure the performance of one program or to compare the performance of two programs, and the control unit 106 or the control unit 300 determines whether the first program is based on the designation from the user. You may choose to perform one embodiment or a second embodiment.
　また、前述の第１の実施形態及び第２の実施形態における初期メソッドは、プログラムを実行することによって自動的に抽出されたが、ユーザによって指定されてもよい。 In addition, although the initial method in the first embodiment and the second embodiment described above is automatically extracted by executing the program, it may be specified by the user.
　図２１は、本発明の第２の実施形態の初期メソッドを指定するための画面を示す説明図である。 FIG. 21 is an explanatory diagram showing a screen for designating an initial method according to the second embodiment of this invention.
　図２１に示す画面４００は、第２の実施形態の初期メソッドを指定するための画面であるが、第１の実施形態において用いられてもよい。 A screen 400 shown in FIG. 21 is a screen for designating an initial method of the second embodiment, but may be used in the first embodiment.
　画面４００は、制御部３００（又は制御部１０６）が出力装置１０４を介してユーザに表示する画面である。画面４００は、選択領域４０１、選択領域４０２、指示領域４０３を含む。ユーザは、出力装置１０４に表示された画面４００を、入力装置１０５によって操作することによって、画面４００に示された情報を指定する。 The screen 400 is a screen that the control unit 300 (or the control unit 106) displays to the user via the output device 104. The screen 400 includes a selection area 401, a selection area 402, and an instruction area 403. The user designates information shown on the screen 400 by operating the screen 400 displayed on the output device 104 with the input device 105.
　選択領域４０１は、初期メソッドを抽出する方法を選択するためのボタン等を含む。図２１に示す選択領域４０１において、"自動"ボタンが選択された場合、制御部３００（又は制御部１０６）は、ステップ３２０によって、初期メソッドを自動的に抽出する。"手動"ボタンが選択された場合、制御部３００（又は制御部１０６は、ステップ３２０（又は１４０）を実行せず、ユーザによって指定された初期メソッドを対象メソッドテーブル３０５（又は対象メソッドテーブル１１３）に格納する。 The selection area 401 includes a button for selecting a method for extracting an initial method. When the “automatic” button is selected in the selection area 401 shown in FIG. 21, the control unit 300 (or the control unit 106) automatically extracts the initial method in step 320. When the “manual” button is selected, the control unit 300 (or the control unit 106 does not execute step 320 (or 140), and the initial method specified by the user is selected as the target method table 305 (or the target method table 113). To store.
　選択領域４０２は、選択領域４０１において"手動"が選択された場合において、初期メソッドを指定するためのチェックボックス等を含む。図２１に示す選択領域４０２には、プログラムによって呼ばれるメソッドが階層構造によって表示される。選択領域４０２に表示される各メソッドは、チェックボックスによって選択可能である。 The selection area 402 includes a check box for designating an initial method when “manual” is selected in the selection area 401. In the selection area 402 shown in FIG. 21, methods called by programs are displayed in a hierarchical structure. Each method displayed in the selection area 402 can be selected by a check box.
　ユーザは、選択領域４０１及び選択領域４０２において、初期メソッドを抽出するための情報を指定すると、指示領域４０３に含まれるボタン等を操作する。その後、制御部３００（又は制御部１０６）は、画面４００によって指定された情報に基づいて、図１９（又は図７）に示す処理を開始する。 When the user designates information for extracting an initial method in the selection area 401 and the selection area 402, the user operates a button or the like included in the instruction area 403. Thereafter, the control unit 300 (or the control unit 106) starts the process shown in FIG. 19 (or FIG. 7) based on the information specified on the screen 400.
　第２の実施形態によれば、一つのプログラムにおいて、性能に影響があるメソッド（最大実行時間メソッド）を自動的に抽出することができる。また、第１の実施形態と同じく、実行時間を測定するメソッドを抽出するため、プログラムの規模によらず、精度よく最大実行時間メソッドを抽出することができる。 According to the second embodiment, a method (maximum execution time method) that affects performance can be automatically extracted in one program. Further, since the method for measuring the execution time is extracted as in the first embodiment, the maximum execution time method can be accurately extracted regardless of the scale of the program.
　さらに、本実施形態は、ユーザによって初期メソッドを指定することが可能であるため、実行時間を測定したいプログラムの一部を、限定的に測定することが可能である。例えば、プログラムの一部が修正され、すべてのプログラムの性能を測定する必要がない場合、ユーザは、性能を測定したいプログラムの一部を最初に実行するメソッドを、初期メソッドに指定する。これによって、短時間で性能を測定することが可能であり、消費電力を低減できる。 Furthermore, in this embodiment, since an initial method can be designated by the user, a part of a program whose execution time is desired to be measured can be measured in a limited manner. For example, when a part of a program is modified and it is not necessary to measure the performance of all programs, the user designates a method that first executes a part of the program whose performance is to be measured as an initial method. As a result, performance can be measured in a short time, and power consumption can be reduced.
　なお、本発明は、前記所定の条件に、呼び出し元メソッドの実行時間及び実行回数を判定条件に用いることもできるため、例えば、呼び出し元メソッドの実行時間に対する割合と所定の閾値とを比較することによって、性能差要因メソッド又は最大実行時間メソッドを抽出してもよい。 In the present invention, since the execution time and the number of executions of the caller method can be used as the determination condition for the predetermined condition, for example, a ratio with respect to the execution time of the caller method is compared with a predetermined threshold. The performance difference factor method or the maximum execution time method may be extracted.
　以上、本発明の実施に関わる性能分析装置、方法、及びプログラムについて説明したが、本発明はこのような具体的構成に限定されるものではなく、添付した請求の範囲の趣旨内における様々な変更及び同等の構成を含むものである。 The performance analysis apparatus, method, and program related to the implementation of the present invention have been described above, but the present invention is not limited to such a specific configuration, and various modifications within the scope of the appended claims. And equivalent configurations.
　本発明は、アプリケーションサーバにおいてプログラムの性能を測定するシステムに適用可能である。 The present invention is applicable to a system that measures the performance of a program in an application server.

Claims (18)

  1. 　計算機によってプログラムの性能を測定する性能測定方法であって、
    　前記計算機は、プロセッサと、前記プロセッサによって実行されるプログラムを格納するメモリとを有し、
    　前記方法は、
    　前記計算機が、前記性能が測定されるプログラムによって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを最初に決定する手順と、
    　前記計算機が、前記プログラムを実行することによって、前記決定された各第１のメソッドの実行時間を測定する手順と、
    　前記計算機が、前記測定された各第１のメソッドの実行時間を用いて、所定の条件に一致する少なくとも一つの第２のメソッドが前記各第１のメソッドから抽出可能か否かを判定する手順と、
    　前記少なくとも一つの第２のメソッドが抽出可能である場合、前記計算機が、前記抽出された各第２のメソッドから直接呼ばれる少なくとも一つの第３のメソッドが抽出可能か否かを判定する手順と、
    　前記少なくとも一つの第３のメソッドが抽出可能である場合、前記計算機が、前記抽出された第３のメソッドを、第１のメソッドに決定する手順と、
    　前記計算機が、前記実行時間を測定する手順と、前記第２のメソッドが抽出可能か否かを判定する手順と、前記第３のメソッドが抽出可能か否かを判定する手順と、前記第３のメソッドを第１のメソッドに決定する手順とを繰り返し実行する手順とを含むことを特徴とする性能測定方法。     A performance measurement method for measuring the performance of a program by a computer,
    The computer includes a processor and a memory that stores a program executed by the processor.
    The method
    A step in which the computer first determines at least one first method among methods called directly by the program whose performance is measured;
    A step in which the computer measures the execution time of each determined first method by executing the program;
    The computer determines whether or not at least one second method that matches a predetermined condition can be extracted from each first method by using the measured execution time of each first method. When,
    If the at least one second method is extractable, the computer determines whether at least one third method called directly from each extracted second method can be extracted;
    If the at least one third method is extractable, the computer determines the extracted third method as the first method;
    A procedure for the computer to measure the execution time; a procedure for determining whether the second method can be extracted; a procedure for determining whether the third method can be extracted; And a step of repeatedly executing a step of determining the method as the first method.
  2. 　請求項１に記載の性能測定方法であって、
    　前記性能が測定されるプログラムは、第１のプログラム及び第２のプログラムを含み、
    　前記少なくとも一つの第１のメソッドを最初に決定する手順は、前記計算機が、前記第１のプログラム及び第２のプログラムの両方によって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを決定する手順を含み、
    　前記実行時間を測定する手順は、前記計算機が、前記各第１のメソッドの前記第１のプログラムにおける第１の実行時間と、前記各第１のメソッドの前記第２のプログラムにおける第２の実行時間とを測定する手順を含み、
    　前記第２のメソッドが抽出可能か否かを判定する手順は、前記計算機が、前記測定された各第１の実行時間と各第２の実行時間とを用いて、前記所定の条件に一致する第２のメソッドが前記第１のメソッドから抽出可能か否かを判定する手順を含むことを特徴とする性能測定方法。     The performance measurement method according to claim 1,
    The program whose performance is measured includes a first program and a second program,
    The procedure of first determining the at least one first method is a procedure in which the computer determines at least one first method among methods directly called by both the first program and the second program. Including
    The procedure for measuring the execution time is such that the computer executes a first execution time in the first program of each first method and a second execution in the second program of each first method. Including procedures for measuring time and
    The procedure for determining whether or not the second method can be extracted matches the predetermined condition using the measured first execution time and the second execution time. A performance measurement method comprising a step of determining whether or not a second method can be extracted from the first method.
  3. 　請求項２に記載の性能測定方法であって、
    　前記少なくとも一つの第１のメソッドを最初に決定する手順は、
    　前記計算機が、前記第１のプログラムの開始後に開始し、前記第１のプログラムの終了前に終了するメソッドのうち、呼び出し元が前記第１のプログラムの開始前に開始されたか、または前記第１のプログラムの終了後に終了されたメソッドである、少なくとも一つの第４のメソッドを抽出する手順と、
    　前記計算機が、前記第２のプログラムの開始後に開始し、前記第２のプログラムの終了前に終了するメソッドのうち、呼び出し元が前記第２のプログラムの開始前に開始されたか、または前記第２のプログラムの終了後に終了されたメソッドである、少なくとも一つの第５のメソッドを抽出する手順と、
    　前記計算機が、前記第４のメソッドと前記第５のメソッドとのうち、同じメソッドを各第１のメソッドとして決定する手順とを含むことを特徴とする性能測定方法。     The performance measurement method according to claim 2,
    The procedure for initially determining the at least one first method comprises:
    Of the methods that the computer starts after the start of the first program and ends before the end of the first program, the caller has started before the start of the first program, or the first A procedure for extracting at least one fourth method, which is a method terminated after completion of the program;
    Among the methods that the computer starts after the start of the second program and ends before the end of the second program, the caller is started before the start of the second program, or the second A procedure for extracting at least one fifth method, which is a method terminated after completion of the program;
    The performance measurement method characterized by including the procedure in which the said computer determines the same method as each 1st method among the said 4th method and the said 5th method.
  4. 　請求項２又は３に記載の性能測定方法であって、
    　前記少なくとも一つの第２のメソッドが抽出可能か否かを判定する手順は、
    　前記計算機が、前記第１のメソッドを直接呼んだ第６のメソッドが、前記第１のプログラムにおいて実行された第１の実行回数と、前記第１のメソッドを直接呼んだ前記各第６のメソッドが前記第２のプログラムにおいて実行された第２の実行回数とを保持する手順と、
    　前記計算機が、前記第６のメソッドから呼ばれる第１のメソッドの、前記第１のプログラムにおける第１の合計実行時間と、前記第２のプログラムにおける第２の合計実行時間とを保持する手順と、
    　前記計算機が、前記第２の合計実行時間を前記第２の実行回数によって除算した結果から、前記第１の合計実行時間を前記第１の実行回数によって除算した結果を、減算することによって、実行時間差を算出する手順と、
    　前記計算機が、前記算出された実行時間差が、所定の条件に一致する前記第１のメソッドが第２のメソッドとして抽出可能か否かを判定する手順とを、含むことを特徴とする性能測定方法。     The performance measurement method according to claim 2 or 3,
    The procedure for determining whether or not the at least one second method can be extracted includes:
    The sixth method in which the computer directly calls the first method is the first number of executions executed in the first program, and each sixth method in which the first method is directly called. Holding a second execution count executed in the second program,
    A procedure in which the computer holds a first total execution time in the first program and a second total execution time in the second program of the first method called from the sixth method;
    Execution by subtracting the result of dividing the first total execution time by the first number of executions from the result of dividing the second total execution time by the second number of executions. The procedure for calculating the time difference;
    And a step of determining whether or not the computer can extract the first method whose calculated execution time difference matches a predetermined condition as a second method. .
  5. 　請求項４に記載の性能測定方法であって、
    　前記実行時間差によって第２のメソッドが抽出可能か否かを判定する手順は、
    　前記計算機が、前記第１のメソッドが、前記第１のプログラムにおいて実行された第３の実行回数と、１回あたりの第１の平均実行時間とを保持する手順と、
    　前記計算機が、前記算出された実行時間差を、前記第３の実行回数と、前記第１の平均実行時間とによって除算することによって、実行時間比を算出する手順と、
    　前記計算機が、前記算出された実行時間比と、所定の閾値とを比較する手順と、
    　前記比較の結果、前記計算機が、前記算出された実行時間比が、所定の閾値より大きい前記第１のメソッドが第２のメソッドとして抽出可能か否かを判定する手順とを、含むことを特徴とする性能測定方法。     The performance measurement method according to claim 4,
    The procedure for determining whether the second method can be extracted based on the execution time difference is as follows:
    A step in which the computer holds the third execution number of times that the first method is executed in the first program and a first average execution time per time;
    A step of calculating an execution time ratio by dividing the calculated execution time difference by the third execution count and the first average execution time;
    A step in which the computer compares the calculated execution time ratio with a predetermined threshold;
    As a result of the comparison, the computer includes a procedure for determining whether or not the first method having the calculated execution time ratio larger than a predetermined threshold can be extracted as a second method. A performance measurement method.
  6. 　請求項１から５のいずれか一つに記載の性能測定方法であって、
    　前記計算機は、ユーザが少なくとも一つのメソッドを指定するための入力装置を備え、
    　前記少なくとも一つの第１のメソッドを最初に決定する手順は、前記入力装置を介して、前記ユーザから各メソッドが指定された場合、前記計算機が、前記指定されたメソッドを、前記各第１のメソッドに決定する手順を含むことを特徴とする性能測定方法。     A performance measurement method according to any one of claims 1 to 5,
    The computer includes an input device for a user to specify at least one method,
    The procedure for first determining the at least one first method is as follows. When each method is designated by the user via the input device, the computer designates the designated method as the first method. A method for measuring performance, comprising a procedure for determining a method.
  7. 　請求項１から６のいずれか一つに記載の性能測定方法であって、
    　前記各第１のメソッドの実行時間を測定する手順は、
    　前記計算機が、前記各第１のメソッドが開始した時刻と、前記各第１のメソッドが終了した時刻とを取得する手順と、
    　前記計算機が、前記取得された各第１のメソッドが終了した時刻から、前記取得された各第１のメソッドが開始した時刻を減算することによって、前記各実行時間を算出する手順とを含むことを特徴とする性能測定方法。     A performance measurement method according to any one of claims 1 to 6,
    The procedure for measuring the execution time of each first method is as follows:
    A procedure for the computer to obtain a time at which each first method starts and a time at which each first method ends;
    And a step of calculating each execution time by subtracting a time at which each acquired first method starts from a time at which each acquired first method ends. A method for measuring performance.
  8. 　プログラムの性能を測定する性能測定装置であって、
    　前記性能測定装置は、
    　プロセッサと、前記プロセッサによって実行されるプログラムを格納するメモリとを有し、
    　前記性能が測定されるプログラムによって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを最初に決定し、
    　前記プログラムを実行することによって、前記決定された各第１のメソッドの実行時間を測定し、
    　前記測定された各第１のメソッドの実行時間を用いて、所定の条件に一致する少なくとも一つの第２のメソッドが前記各第１のメソッドから抽出可能か否かを判定し、
    　前記少なくとも一つの第２のメソッドが抽出可能である場合、前記抽出された各第２のメソッドから直接呼ばれる少なくとも一つの第３のメソッドが抽出可能か否かを判定し、
    　前記少なくとも一つの第３のメソッドが抽出可能である場合、前記抽出された第３のメソッドを、第１のメソッドに決定し、
    　前記実行時間を測定し、前記第２のメソッドが抽出可能か否かを判定し、前記第３のメソッドが抽出可能か否かを判定し、前記第３のメソッドを第１のメソッドに決定する手順を繰り返し実行することを特徴とする性能測定装置。     A performance measuring device for measuring the performance of a program,
    The performance measuring device is
    A processor and a memory for storing a program executed by the processor;
    First determining at least one first method among methods called directly by the program whose performance is to be measured;
    Measuring the execution time of each determined first method by executing the program;
    Using the measured execution time of each first method to determine whether at least one second method matching a predetermined condition can be extracted from each first method;
    If the at least one second method is extractable, determine whether at least one third method called directly from each extracted second method is extractable;
    If the at least one third method is extractable, the extracted third method is determined as a first method;
    Measure the execution time, determine whether the second method can be extracted, determine whether the third method can be extracted, and determine the third method as the first method A performance measuring apparatus characterized by repeatedly executing a procedure.
  9. 　請求項８に記載の性能測定装置であって、
    　前記性能が測定されるプログラムは、第１のプログラム及び第２のプログラムを含み、
    　前記性能測定装置は、
    　前記第１のプログラム及び第２のプログラムの両方によって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを決定し、
    　前記各第１のメソッドの前記第１のプログラムにおける第１の実行時間と、前記各第１のメソッドの前記第２のプログラムにおける第２の実行時間とを測定し、
    　前記測定された各第１の実行時間と各第２の実行時間とを用いて、前記所定の条件に一致する第２のメソッドが前記第１のメソッドから抽出可能か否かを判定することを特徴とする性能測定装置。     The performance measuring device according to claim 8,
    The program whose performance is measured includes a first program and a second program,
    The performance measuring device is
    Determining at least one first method among methods called directly by both the first program and the second program;
    Measuring a first execution time in the first program of each first method and a second execution time in the second program of each first method;
    Determining whether a second method that matches the predetermined condition can be extracted from the first method by using the measured first execution time and the second execution time; A characteristic performance measuring device.
  10. 　請求項９に記載の性能測定装置であって、
    　前記性能測定装置は、
    　前記第１のプログラムの開始後に開始し、前記第１のプログラムの終了前に終了するメソッドのうち、呼び出し元が前記第１のプログラムの開始前に開始されたか、または前記第１のプログラムの終了後に終了されたメソッドである、少なくとも一つの第４のメソッドを抽出し、
    　前記第２のプログラムの開始後に開始し、前記第２のプログラムの終了前に終了するメソッドのうち、呼び出し元が前記第２のプログラムの開始前に開始されたか、または前記第２のプログラムの終了後に終了されたメソッドである、少なくとも一つの第５のメソッドを抽出し、
    　前記第４のメソッドと前記第５のメソッドとのうち、同じメソッドを各第１のメソッドとして最初に決定する手順とを含むことを特徴とする性能測定装置。     The performance measuring device according to claim 9,
    The performance measuring device is
    Of the methods that start after the first program starts and end before the first program ends, the caller started before the first program starts or the first program ends Extract at least one fourth method that is later terminated,
    Of the methods that start after the start of the second program and end before the end of the second program, the caller started before the start of the second program, or the end of the second program Extract at least one fifth method, which is a method that was terminated later,
    A performance measuring apparatus comprising: a step of first determining the same method as each first method out of the fourth method and the fifth method.
  11. 　請求項９又は１０に記載の性能測定装置であって、
    　前記性能測定装置は、
    　前記第１のメソッドを直接呼んだ第６のメソッドが、前記第１のプログラムにおいて実行された第１の実行回数と、前記第１のメソッドを直接呼んだ前記各第６のメソッドが前記第２のプログラムにおいて実行された第２の実行回数とを保持し、
    　前記第６のメソッドから呼ばれる第１のメソッドの、前記第１のプログラムにおける第１の合計実行時間と、前記第２のプログラムにおける第２の合計実行時間とを保持し、
    　前記第２の合計実行時間を前記第２の実行回数によって除算した結果から、前記第１の合計実行時間を前記第１の実行回数によって除算した結果を、減算することによって、実行時間差を算出し、
    　前記算出された実行時間差が、所定の条件に一致する前記第１のメソッドが第２のメソッドとして抽出可能か否かを判定することを特徴とする性能測定装置。     The performance measuring device according to claim 9 or 10,
    The performance measuring device is
    The sixth method that directly calls the first method is the first number of executions executed in the first program, and each sixth method that directly calls the first method is the second method. And the second execution count executed in the program of
    Holding a first total execution time in the first program and a second total execution time in the second program of the first method called from the sixth method;
    An execution time difference is calculated by subtracting a result of dividing the first total execution time by the first number of executions from a result of dividing the second total execution time by the second number of executions. ,
    A performance measurement apparatus that determines whether or not the calculated first execution method whose execution time difference matches a predetermined condition can be extracted as a second method.
  12. 　請求項１１に記載の性能測定装置であって、
    　前記性能測定装置は、
    　前記第１のメソッドが、前記第１のプログラムにおいて実行された第３の実行回数と、１回あたりの第１の平均実行時間とを保持し、
    　前記算出された実行時間差を、前記第３の実行回数と、前記第１の平均実行時間とによって除算することによって、実行時間比を算出し、
    　前記算出された実行時間比と、所定の閾値とを比較し、
    　前記比較の結果、前記算出された実行時間比が、所定の閾値より大きい前記第１のメソッドが第２のメソッドとして抽出可能か否かを判定することを特徴とする性能測定装置。     The performance measuring device according to claim 11,
    The performance measuring device is
    The first method holds a third execution number of times executed in the first program and a first average execution time per time,
    An execution time ratio is calculated by dividing the calculated execution time difference by the third execution count and the first average execution time,
    Comparing the calculated execution time ratio with a predetermined threshold;
    As a result of the comparison, it is determined whether or not the first method having the calculated execution time ratio larger than a predetermined threshold can be extracted as a second method.
  13. 　請求項８から１２のいずれか一つに記載の性能測定装置であって、
    　前記性能測定装置は、
    　ユーザが少なくとも一つのメソッドを指定するための入力装置を備え、
    　前記入力装置を介して、前記ユーザから各メソッドが指定された場合、前記指定されたメソッドを、前記各第１のメソッドに決定することを特徴とする性能測定装置。     The performance measuring device according to any one of claims 8 to 12,
    The performance measuring device is
    An input device for the user to specify at least one method;
    When each method is designated by the user via the input device, the designated method is determined as the first method.
  14. 　請求項８から１３のいずれか一つに記載の性能測定装置であって、
    　前記性能測定装置は、
    　前記前記各第１のメソッドが開始した時刻と、前記各第１のメソッドが終了した時刻とを取得し、
    　前記取得された各第１のメソッドが終了した時刻から、前記取得された各第１のメソッドが開始した時刻を減算することによって、前記各実行時間を算出することを特徴とする性能測定装置。     The performance measuring device according to any one of claims 8 to 13,
    The performance measuring device is
    Obtaining a time at which each of the first methods starts and a time at which each of the first methods ends;
    The performance measuring apparatus according to claim 1, wherein the execution time is calculated by subtracting a time at which each acquired first method starts from a time at which each acquired first method ends.
  15. 　計算機にプログラムの性能を測定させる性能測定プログラムであって、
    　前記計算機は、プロセッサと、前記プロセッサによって実行されるプログラムを格納するメモリとを有し、
    　前記性能測定プログラムは、
    　前記計算機に、前記性能が測定されるプログラムによって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを最初に決定させる手順と、
    　前記計算機に、前記プログラムを実行することによって、前記決定された各第１のメソッドの実行時間を測定させる手順と、
    　前記計算機に、前記測定された各第１のメソッドの実行時間を用いて、所定の条件に一致する少なくとも一つの第２のメソッドが前記各第１のメソッドから抽出可能か否かを判定させる手順と、
    　前記少なくとも一つの第２のメソッドが抽出可能である場合、前記計算機に、前記抽出された各第２のメソッドから直接呼ばれる少なくとも一つの第３のメソッドが抽出可能か否かを判定させる手順と、
    　前記少なくとも一つの第３のメソッドが抽出可能である場合、前記計算機に、前記抽出された第３のメソッドを、第１のメソッドに決定させる手順と、
    　前記計算機に、前記実行時間を測定させる手順と、前記第２のメソッドが抽出可能か否かを判定させる手順と、前記第３のメソッドが抽出可能か否かを判定させる手順と、前記第３のメソッドを第１のメソッドに決定させる手順とを繰り返し実行させるための性能測定プログラム。     A performance measurement program that causes a computer to measure the performance of a program,
    The computer includes a processor and a memory that stores a program executed by the processor.
    The performance measurement program is
    A step of causing the computer to first determine at least one first method among methods called directly by the program whose performance is measured;
    A step of causing the computer to measure the execution time of each of the determined first methods by executing the program;
    Procedure for causing the computer to determine whether or not at least one second method that matches a predetermined condition can be extracted from each first method by using the measured execution time of each first method. When,
    If the at least one second method is extractable, causing the computer to determine whether or not at least one third method called directly from each extracted second method can be extracted;
    If the at least one third method is extractable, causing the computer to determine the extracted third method as the first method;
    A step for causing the computer to measure the execution time; a step for determining whether the second method can be extracted; a step for determining whether the third method can be extracted; A performance measurement program for repeatedly executing a procedure for causing the first method to be determined as the first method.
  16. 　請求項１５に記載の性能測定プログラムであって、
    　前記性能が測定されるプログラムは、第１のプログラム及び第２のプログラムを含み、
    　前記少なくとも一つの第１のメソッドを最初に決定させる手順において、前記計算機に、前記第１のプログラム及び第２のプログラムの両方によって直接呼ばれるメソッドのうち、少なくとも一つの第１のメソッドを決定させる手順と、
    　前記実行時間を測定させる手順において、前記計算機に、前記各第１のメソッドの前記第１のプログラムにおける第１の実行時間と、前記各第１のメソッドの前記第２のプログラムにおける第２の実行時間とを測定させる手順と、
    　前記第２のメソッドが抽出可能か否かを判定させる手順において、前記計算機に、前記測定された各第１の実行時間と各第２の実行時間とを用いて、前記所定の条件に一致する第２のメソッドが前記第１のメソッドから抽出可能か否かを判定させる手順とを実行させるための性能測定プログラム。     The performance measurement program according to claim 15,
    The program whose performance is measured includes a first program and a second program,
    A step of causing the computer to determine at least one first method among methods called directly by both the first program and the second program in the step of first determining the at least one first method; When,
    In the procedure for measuring the execution time, the computer causes the computer to execute a first execution time of the first method in the first program and a second execution of the first method in the second program. Procedures to measure time,
    In the procedure for determining whether or not the second method can be extracted, the computer uses the measured first execution time and the second execution time to meet the predetermined condition. A performance measurement program for executing a procedure for determining whether a second method can be extracted from the first method.
  17. 　請求項１６に記載の性能測定プログラムであって、
    　前記少なくとも一つの第１のメソッドを最初に決定させる手順において、
    　前記計算機に、前記第１のプログラムの開始後に開始し、前記第１のプログラムの終了前に終了するメソッドのうち、呼び出し元が前記第１のプログラムの開始前に開始されたか、または前記第１のプログラムの終了後に終了されたメソッドである、少なくとも一つの第４のメソッドを抽出させる手順と、
    　前記計算機に、前記第２のプログラムの開始後に開始し、前記第２のプログラムの終了前に終了するメソッドのうち、呼び出し元が前記第２のプログラムの開始前に開始されたか、または前記第２のプログラムの終了後に終了されたメソッドである、少なくとも一つの第５のメソッドを抽出させる手順と、
    　前記計算機に、前記第４のメソッドと前記第５のメソッドとのうち、同じメソッドを各第１のメソッドとして決定させる手順とを実行させるための性能測定プログラム。     A performance measurement program according to claim 16, comprising:
    In the step of first determining the at least one first method,
    Of the methods that start after the first program starts and end before the end of the first program, the caller has started before the start of the first program, or the first A procedure for extracting at least one fourth method, which is a method terminated after completion of the program;
    Of the methods that start after the start of the second program and end before the end of the second program, the caller is started before the start of the second program, or the second A procedure for extracting at least one fifth method, which is a method terminated after completion of the program;
    A performance measurement program for causing the computer to execute a procedure for determining the same method as each first method out of the fourth method and the fifth method.
  18. 　請求項１５から１７のいずれか一つに記載の性能測定プログラムであって、
    　前記各第１のメソッドの実行時間を測定させる手順において、
    　前記計算機に、前記各第１のメソッドが開始した時刻と、前記各第１のメソッドが終了した時刻とを取得させる手順と、
    　前記計算機に、前記取得された各第１のメソッドが終了した時刻から、前記取得された各第１のメソッドが開始した時刻を減算することによって、前記各実行時間を算出させる手順とを実行させるための性能測定プログラム。     A performance measurement program according to any one of claims 15 to 17,
    In the procedure for measuring the execution time of each first method,
    A procedure for causing the computer to acquire a time at which each first method starts and a time at which each first method ends;
    Causing the calculator to calculate each execution time by subtracting the time at which each acquired first method starts from the time at which each acquired first method ends. Performance measurement program for.
PCT/JP2010/069335 2010-10-29 2010-10-29 Performance measurement method, performance measurement device, and performance measurement program WO2012056569A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/JP2010/069335 WO2012056569A1 (en) 2010-10-29 2010-10-29 Performance measurement method, performance measurement device, and performance measurement program

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2010/069335 WO2012056569A1 (en) 2010-10-29 2010-10-29 Performance measurement method, performance measurement device, and performance measurement program

Publications (1)

Publication Number Publication Date
WO2012056569A1 true WO2012056569A1 (en) 2012-05-03

Family

ID=45993323

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2010/069335 WO2012056569A1 (en) 2010-10-29 2010-10-29 Performance measurement method, performance measurement device, and performance measurement program

Country Status (1)

Country Link
WO (1) WO2012056569A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2014170500A (en) * 2013-03-05 2014-09-18 Toshiba Tec Corp Information processor and program
US10740358B2 (en) 2013-04-11 2020-08-11 Oracle International Corporation Knowledge-intensive data processing system

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007249495A (en) * 2006-03-15 2007-09-27 Nec Corp Software verification method, information processor and program
JP2008217721A (en) * 2007-03-07 2008-09-18 Ntt Data Corp Program profiling system, program profiling method, and program
JP2010198133A (en) * 2009-02-23 2010-09-09 Internatl Business Mach Corp <Ibm> Method of selecting function for performance evaluation modeling

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007249495A (en) * 2006-03-15 2007-09-27 Nec Corp Software verification method, information processor and program
JP2008217721A (en) * 2007-03-07 2008-09-18 Ntt Data Corp Program profiling system, program profiling method, and program
JP2010198133A (en) * 2009-02-23 2010-09-09 Internatl Business Mach Corp <Ibm> Method of selecting function for performance evaluation modeling

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2014170500A (en) * 2013-03-05 2014-09-18 Toshiba Tec Corp Information processor and program
US10740358B2 (en) 2013-04-11 2020-08-11 Oracle International Corporation Knowledge-intensive data processing system
US11468098B2 (en) 2013-04-11 2022-10-11 Oracle International Corporation Knowledge-intensive data processing system

Similar Documents

Publication Publication Date Title
US10127129B2 (en) Non-invasive time-based profiling tool
US7853929B2 (en) Dynamically profiling consumption of CPU time in Java methods with respect to method line numbers while executing in a Java virtual machine
US7315795B2 (en) Performance analyzing method using hardware
US8572581B2 (en) Measurement and reporting of performance event rates
JP5669630B2 (en) Test case generation method, program and system
US7673287B2 (en) Testing usability of a software program
US8850394B2 (en) Debugging system and method
US9400732B2 (en) Performance regression testing in presence of noise
EP2476059B1 (en) A method and apparatus for determining processor performance
KR100921514B1 (en) A Software Development Apparatus for Providing Performance Prediction and A method thereof
US8584095B2 (en) Test support system, method and computer program product, which optimize test scenarios to minimize total test time
JP6090327B2 (en) Bottleneck detection apparatus, method and program
US20130332913A1 (en) Indirect Software Performance Analysis
US20090083716A1 (en) Profiling method and program
JP6691295B2 (en) Information processing apparatus, program execution state display method, and program execution state display program
JP4879782B2 (en) Program profiling method and program
Fournier et al. Analyzing blocking to debug performance problems on multi-core systems
JP2017167930A (en) Information processing device, power measurement method and power measurement program
JP6995146B2 (en) Performance analysis of adaptive applications
WO2012056569A1 (en) Performance measurement method, performance measurement device, and performance measurement program
CN103838662A (en) Dynamic library profiling method and dynamic library profiling system
JP6544054B2 (en) INFORMATION PROCESSING APPARATUS, EXECUTION INFORMATION RECORDING PROGRAM, AND EXECUTION INFORMATION RECORDING METHOD
JP2016514326A (en) Method and system for analyzing a trace timeline of computer system activity
JP5056479B2 (en) Competitive partial processing detection method, apparatus and computer program
JP2005215816A (en) Performance profiling method using hardware monitor

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10858957

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 10858957

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: JP