System and method for measuring code segment performance

• US 7,143,396 B2
• Filed: 11/06/2002
• Issued: 11/28/2006
• Est. Priority Date: 11/06/2002
• Status: Active Grant

• Current Assignee: Oracle America, Inc. (Oracle Corporation)
• Original Assignee: Sun Microsystems Incorporated (Oracle Corporation)
• Inventors: Suresh, Charles
• Primary Examiner(s): INGBERG, TODD D
• Application Number:
  US10/288,967
• Publication Number:
  US 20040088699A1
• Time in Patent Office: 1,483 Days
• Field of Search: 717/130, 717/131, 717/158, 714/37, 714/35, 714/39
• US Class Current: 717/130
• CPC Class Codes: G06F 11/3419 by assessing time
• CPC Class Codes: G06F 11/3466 Performance evaluation by tracing or monitoring
• CPC Class Codes: G06F 2201/805 Real-time
• CPC Class Codes: G06F 2201/865 Monitoring of software
• CPC Class Codes: G06F 2201/88 Monitoring involving counting

Abstract

A system and methods are provided for inserting probe points into an executing program, and measuring the time spent traversing code paths from one probe point to any other probe point or some other performance metric (e.g., instructions executed, cache misses, memory addresses accessed). One method is implemented by inserting N probes. Each probe has a corresponding function configured to: retrieve the identifier and timestamp of the previous probe executed, calculate the time spent traversing the path from the previous probe to the current probe, and update a matrix of N×N elements, wherein each element corresponds to a path from one probe to another probe. After completion of the program, this matrix is useful for identifying code paths that are bottlenecks and hence candidates for optimization.

Claims

1. A method of monitoring performance of a program executing in a computer, comprising:

installing a uniquely identifiable probe at each of N probe points (N≧

2) by;

copying an original instruction at the probe point to a corresponding probe function;

modifying the original instruction path to branch to said probe function prior to the original instruction; and

configuring said probe function to return to the original instruction path; and

during execution of a first probe function corresponding to a first probe point;

updating a first timestamp associated with the first probe point with a current timestamp; and

for each of N-1 other probe points;

retrieving another timestamp associated with another probe point that preceded the first probe point during execution of the program; and

updating a probe path matrix on a computer readable medium to reflect the time difference between said current timestamp and said retrieved timestamp.

2. The method of claim 1, further comprising:

maintaining a timestamp array comprising an element for each of said N probe points;

wherein each element of said timestamp array is configured to store a timestamp of the most recent execution of the probe function corresponding to the associated probe point.

3. The method of claim 2, wherein said updating a first timestamp comprises:

storing said current time stamp in an element of said timestamp array associated with the first probe point.

4. The method of claim 2, wherein said retrieving another timestamp comprises:

retrieving said other timestamp from an element of said timestamp array associated with the other probe point.

5. The method of claim 2, wherein:

the computer comprises multiple processors; and

each element of said timestamp array is further configured to identify the processor on which the probe function corresponding to the associated probe point was most recently executed.

6. The method of claim 1, further comprising:

initializing said probe path matrix for N×

N elements, wherein each element corresponds to a path from one probe point to another probe point, and is configured to store;

a total number of times said path is traversed during said execution of the program; and

a cumulative time spent traversing the path said total number of times.

7. The method of claim 6, wherein said updating a probe path matrix comprises:

identifying an element of said probe path matrix corresponding to the path from said other probe point to said first probe point; and

within said identified element;

incrementing said total number of times said path from said other probe point to said first probe point has been traversed; and

increasing said cumulative time spent traversing said path from said other probe point to said first probe point by said time difference.

8. The method of claim 1, further comprising:

configuring an identifier for identifying a probe point corresponding to the most recently executed probe function.

9. A computer readable storage medium storing instructions that, when executed by a computer, cause the computer to perform a method of monitoring performance of a program executing in a computer, the method comprising:

installing a uniquely identifiable probe at each of N probe points (N≧

2) by;

copying an original instruction at the probe point to a corresponding probe function;

modifying the original instruction path to branch to said probe function prior to the original instruction; and

configuring said probe function to return to the original instruction path; and

during execution of a first probe function corresponding to a first probe point;

updating a first timestamp associated with the first probe point with a current timestamp; and

for each of N-1 other probe points;

retrieving another timestamp associated with another probe point that preceded the first probe point during execution of the program; and

updating a probe path matrix on a computer readable medium to reflect the time difference between said current timestamp and said retrieved timestamp.

10. A method of monitoring performance of a program executing in a computer system, comprising:

installing a uniquely identifiable probe at each of N probe points (N≧

2) by;

copying an original instruction at the probe point to a corresponding probe function;

modifying the original instruction path to branch to said probe function prior to the original instruction; and

configuring said probe function to return to the original instruction path; and

during execution of a first probe function corresponding to a first probe point;

obtaining a measurement of a predetermined performance metric of the computer system;

updating a first performance metric measurement associated with the first probe point; and

for each of the N-1 other probe points;

retrieving another performance metric measurement associated with another probe point executed before the first probe point; and

updating a matrix on a computer readable medium to reflect a difference between said first performance metric and said other performance metric.

11. The method of claim 10, further comprising:

maintaining a resource array comprising an element for each of said N probe points;

wherein each element of said resource array is configured to store a performance metric measurement obtained during the most recent execution of the probe function corresponding to the associated probe point.

12. The method of claim 10, further comprising:

initializing said probe path matrix for N×

N elements, wherein each element corresponds to a path from one probe point to another probe point, and is configured to store;

a total number of times said path is traversed during said execution of the program; and

the accumulated difference between measurements of the performance metric at the one probe point and the other probe point during said total number of times said path is traversed.

13. The method of claim 12, wherein said updating a matrix comprises:

identifying an element of said matrix corresponding to the path from said other probe point to said first probe point; and

within said identified element;

incrementing said total number of times said path from said other probe point to said first probe point has been traversed; and

increasing said accumulated difference by said difference between said first performance metric and said other performance metric.

14. The computer system of claim 10, wherein said performance metric comprises a number of instructions executed.

15. The computer system of claim 10, wherein said performance metric comprises a number of cache misses.

16. The computer system of claim 10, wherein said performance metric comprises a number of memory addresses accessed.

17. The computer system of claim 10, wherein said performance metric comprises a number of Translation Lookaside Buffer invalidations.

18. The computer system of claim 10, wherein said performance metric comprises time.

19. A computer readable storage medium storing instructions that, when executed by a computer, cause the computer to perform a method of monitoring performance of a program executing in a computer system, the method comprising:

installing a uniquely identifiable probe at each of N probe points (N≧

2) by;

copying an original instruction at the probe point to a corresponding probe function;

modifying the original instruction path to branch to said probe function prior to the original instruction; and

configuring said probe function to return to the original instruction path; and

during execution of a first probe function corresponding to a first probe point;

obtaining a measurement of a predetermined performance metric of the computer system;

updating a first performance metric measurement associated with the first probe point; and

for each of the N-1 other probe points;

retrieving another performance metric measurement associated with another probe point executed before the first probe point; and

updating a matrix on a computer readable medium to reflect a difference between said first performance metric and said other performance metric.

20. A computer readable storage medium containing a data structure configured for facilitating the measurement of an amount of time spent traversing a path of execution of a computer program, the data structure comprising:

an array configured to store N timestamps, wherein each said timestamp reflects a most recent time of execution of a corresponding one of N probes inserted into the computer program;

an identifier configured to identify which of said N probes was executed most recently; and

an N×

N array of path elements, wherein each said path element corresponds to a path from a first of said N probes to a second of said N probes and comprises;

a number of times said path has been traversed; and

the total amount of time spent traversing said path during said number of times;

wherein said N×

N array of path elements is updated each time one of said N probes is executed.

21. A computer readable storage medium containing a data structure configured to facilitate the measurement of the performance of a portion of a computer program, the data structure comprising:

an array configured to store N measurements of a predetermined performance metric, wherein each said measurement reflects a measure of the performance metric at the time of execution of a corresponding one of N probes inserted into the computer program;

an identifier configured to identify which of said N probes was executed most recently; and

an N×

N array of path elements, wherein each said path element corresponds to a path from a first of said N probes to a second of said N probes and comprises;

a number of times said path has been traversed; and

the accumulated differences between said performance metric measurement at the time of execution of the first probe and said performance metric measurement at the time of execution of the second probe;

wherein said N×

N array of path elements is updated each time one of said N probes is executed.

22. A computer system for measuring the time spent traversing a portion of a computer program, comprising:

a computer program comprising a series of processor executable instructions;

an instrumentation module configured to insert N probes (N>

2) into the computer program;

a processor configured to execute the computer program and said N probes;

for each of said N probes, a corresponding probe function configured to generate a timestamp at which said probe is executed; and

a data module tangibly embodied on a computer readable medium and configured to store;

said timestamps of said probe executions;

an identifier of the last probe executed; and

for each pair of probes within said N probes, the number of times a program path between said pair of probes has been traversed and the accumulated amount of time spent traversing the program path;

wherein said data module is updated each time one of said N probes is executed.

23. The computer system of claim 22, wherein said data module includes a timestamp array configured to store said timestamps of said probe executions.

24. The computer system of claim 22, wherein said timestamp reflects the most recent execution of the corresponding probe function.

25. The computer system of claim 22, wherein said data module includes a probe path matrix configured to store said number of times the program path has been traversed and said accumulated amount of time spent traversing the program path.

26. The computer system of claim 25, wherein said probe function is further configured to update the probe path matrix by:

incrementing said number of times the program path has been traversed; and

increasing said accumulated amount of time spent traversing the program path.

27. The computer system of claim 26, wherein said probe function increases said accumulated amount of time by:

retrieving a timestamp of the last probe executed;

calculating the difference between the generated timestamp and the retrieved timestamp; and

adding said difference to said accumulated amount of time spent traversing the program path.

28. The computer system of claim 22, further comprising:

one or more processors in addition to the processor; and

a corresponding data module for each of said one or more processors;

wherein each of said one or more processors is configured to execute the computer program and said two or more probes.

29. The computer system of claim 28, wherein said corresponding data module further comprises:

an identifier of the processor on which a probe is executed.

30. The computer system of claim 22, wherein said instrumentation module is further configured to dynamically enable one or more of said N probes.

31. The computer system of claim 22, wherein said instrumentation module is further configured to dynamically disable one or more of said N probes.

32. The computer system of claim 22, wherein said instrumentation module is further configured to dynamically remove one or more of said N probes.

33. A computer system for measuring resource usage of a computer program, comprising:

a computer program comprising a series of processor executable instructions;

an instrumentation module configured to insert N probes (N>

2) into the computer program;

a processor configured to execute the computer program and said N probes;

for each of said N probes, a corresponding probe function configured to obtain a measurement of a performance metric of the computer system at the time said probe is executed; and

a data module tangibly embodied on a computer readable medium and configured to store;

said measurements of said performance metrics;

an identifier of the last probe executed; and

for each pair of probes within said N probes;

the number of times a program path between said pair of probes has been traversed; and

the accumulated differences between said performance metric measurements of said pair of probes during the number of times a program path between said pair of probes was traversed;

wherein said data module is updated each time one of said N probes is executed.

34. The computer system of claim 33, wherein said data module includes a resource array configured to store said performance metric measurements.

35. The computer system of claim 34, wherein said probe function is further configured to update the data module by:

incrementing said number of times the program path has been traversed;

determining a difference between a measurement of the performance metric when a first probe in said pair of probes is executed and a measurement of the performance metric when the second probe in said pair of probes is executed; and

increasing said accumulated differences by said difference.

36. The computer system of claim 33, wherein said performance metric comprises a number of instructions executed.

37. The computer system of claim 33, wherein said performance metric comprises a number of cache misses.

38. The computer system of claim 33, wherein said performance metric comprises a number of memory addresses executed.

39. The computer system of claim 33, wherein said performance metric comprises a number of Translation Lookaside Buffer invalidations.

40. The computer system of claim 33, wherein said performance metric comprises time.