Profile-driven lock handling

• US 7,987,452 B2
• Filed: 12/15/2005
• Issued: 07/26/2011
• Est. Priority Date: 12/15/2005
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Schroeder, Mark Douglas, Olson, Gregory Allan, Moore, Scott Alan, Kronlund, Curtis Duane
• Primary Examiner(s): Dam; Tuan Q
• Assistant Examiner(s): Aguilera; Todd
• Application Number:
  US11/304,143
• Publication Number:
  US 20070169002A1
• Time in Patent Office: 2,049 Days
• Field of Search: 710200-244, 717/148, 717151-161, 707 8- 9, 711151-152, 718/104
• US Class Current: 717/127
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms

Abstract

An apparatus, program product and method utilize metadata collected for a lock in a computer program to determine an optimal lock acquisition strategy for the lock. The metadata is typically collected via profiling of the computer program to determine a wait characteristic for the lock in the particular runtime environment in which the computer program executes, and the metadata is used to select from among different lock access implementations to be used to attempt to acquire the lock, often enabling an optimum lock access implementation to be selected for a lock.

Claims

1. A computer-implemented method of processing program code configured to access a lock, the method comprising:

profiling the program code to collect metadata associated with the lock for a particular runtime environment;

in a first thread attempting to access the lock being held by a second thread, accessing the metadata associated with the lock to determine an expected total wait time for the lock to be held;

tracking a how long the lock has already been held by the second thread while the lock is held by the second thread;

in the first thread, determining from the expected total wait time and how long the lock has already been held by the second thread an expected remaining wait time before the lock is released by the second thread; and

based upon the expected remaining wait time and using at least one hardware-implemented processor, selecting from among a plurality of lock access implementations to access the lock.

2. The method of claim 1, wherein the program code comprises program code that is compatible with a JAVA programming environment, and wherein the lock comprises a lock that is compatible with the JAVA programming environment.

3. The method of claim 1, wherein the metadata comprises profile information associated with at least one prior access of the lock.

4. The method of claim 3, further comprising collecting the profile information during execution of the program code.

5. The method of claim 1, wherein accessing the metadata and selecting from among the plurality of lock access implementations are performed during compilation of the program code.

6. The method of claim 5, wherein accessing the metadata and selecting from among the plurality of lock access implementations are performed during just in time compilation of the program code.

7. The method of claim 6, further comprising collecting the metadata during interpretation of the program code.

8. The method of claim 1, wherein accessing the metadata and selecting from among the plurality of lock access implementations are performed during execution of the program code.

9. The method of claim 1, wherein the metadata includes metadata selected from the group consisting of average wait time, maximum wait time, minimum wait time, average time held by a thread, maximum time held by a thread, minimum time held by a thread, number of acquisitions without contention, number of acquisitions with contention, queue depth, and combinations thereof.

10. The method of claim 1, wherein the plurality of lock access implementations includes first and second lock access implementations.

11. The method of claim 10, wherein the first and second lock access implementations differ from one another based upon access method type.

12. The method of claim 11, wherein each of the first and second lock access implementations is selected from the group consisting of an inline fast path lock access implementation, a spin access implementation, a yield access implementation, a thread suspension access implementation, and combinations thereof.

13. The method of claim 11, wherein the first lock access implementation is configured to proceed to the second lock access implementation if unable to obtain the lock, and wherein selecting from among the plurality of lock access implementations based upon the accessed metadata comprises selecting the second lock access implementation such that the first lock access implementation is skipped.

14. The method of claim 10, wherein the first and second lock access implementations differ from one another based upon a parameter common to each lock access implementation.

15. The method of claim 14, wherein the parameter comprises a wait parameter.

16. The method of claim 10, further comprising performing lock coarsening on the program code.

17. The method of claim 10, wherein the first lock access implementation includes an inline access check, and wherein the second lock access implementation is configured to call an external function.

18. The method of claim 10, wherein the first lock access implementation is configured to discard data from at least one volatile register, and wherein the second lock access implementation is configured to save the data from the volatile register.

19. The method of claim 10, wherein selecting from among the plurality of lock access implementations to access the lock includes selecting from among a plurality of register allocation strategies to be used in connection with accessing the lock.

20. The method of claim 1, wherein selecting from among the plurality of lock access implementations to access the lock is performed in a first thread, and includes determining from the accessed metadata how much longer the lock is likely to be held by a second thread that is currently holding the lock.

21. A computer-implemented method of processing program code configured to access a lock, the method comprising:

profiling the program code to collect metadata associated with the lock for a particular runtime environment;

accessing the metadata associated with the lock to determine a wait characteristic for the lock; and

using a register allocator executing on at least one hardware-implemented processor during compilation of a method that accesses the lock using an inline lock access implementation that, if unsuccessful, calls an external function access implementation to acquire the lock, selecting from among a plurality of register allocation strategies to be used in connection with accessing the lock, wherein selecting from among the plurality of register allocation strategies comprises selecting a first register allocation strategy to use during compilation of the method to minimize register spills upon determining from the metadata that lock contention is common for the lock, and selecting a second register allocation strategy to use during compilation of the method to maximize use of volatile registers upon determining from the metadata that lock contention is not common for the lock.

22. A computer-implemented method of processing program code configured to access a lock, the method comprising:

profiling the program code to collect metadata associated with the lock for a particular runtime environment;

in a first thread, accessing the metadata associated with the lock when the lock is held by a second thread to determine an expected total wait time for the lock to be held by the second thread;

tracking a how long the lock has already been held by the second thread while the lock is held by the second thread;

in the first thread, determining from the expected total wait time and how long the lock has already been held by the second thread an expected remaining wait time before the lock is released by the second thread; and

based upon the expected remaining wait time and using at least one hardware-implemented processor, selecting a lock access implementation for the first thread to access the lock, wherein selecting the lock access implementation includes selecting a wait parameter that controls how long to wait to change from using a first lock access implementation for the first thread to using a second lock access implementation for the first thread that has an access method type that differs from that of the first lock access implementation.

23. An apparatus, comprising:

a processor; and

a program configured to execute on the processor to process program code configured to access a lock by profiling the program code to collect metadata associated with the lock for a particular runtime environment, in a first thread attempting to access the lock being held by a second thread, accessing the metadata associated with the lock to determine an expected total wait time for the lock to be held, tracking a how long the lock has already been held by the second thread while the lock is held by the second thread, in the first thread, determining from the expected total wait time and how long the lock has already been held by the second thread an expected remaining wait time before the lock is released by the second thread, and based upon the expected remaining wait time, selecting from among a plurality of lock access implementations to access the lock.

24. A program product, comprising:

a program configured to process program code configured to access a lock by profiling the program code to collect metadata associated with the lock for a particular runtime environment, in a first thread attempting to access the lock being held by a second thread, accessing the metadata associated with the lock to determine an expected total wait time for the lock to be held, tracking a how long the lock has already been held by the second thread while the lock is held by the second thread, in the first thread, determining from the expected total wait time and how long the lock has already been held by the second thread an expected remaining wait time before the lock is released by the second thread, and based upon the expected remaining wait time, selecting from among a plurality of lock access implementations to access the lock; and

a computer readable recordable type medium bearing the program.