Scalable spinlocks for non-uniform memory access

• US 10,310,916 B2
• Filed: 09/14/2017
• Issued: 06/04/2019
• Est. Priority Date: 09/14/2017
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Shaw, Steven Mark, Nueckel, Hubert
• Primary Examiner(s): Giardino, Jr., Mark A
• Application Number:
  US15/704,799
• Publication Number:
  US 20190079807A1
• Time in Patent Office: 628 Days
• CPC Class Codes: G06F 12/1466 Key-lock mechanism
• CPC Class Codes: G06F 2212/1052 Security improvement
• CPC Class Codes: G06F 2212/2542 Non-uniform memory access [NUMA] architecture
• CPC Class Codes: G06F 9/5022 Mechanisms to release resources
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms

Abstract

Techniques are disclosed to provide scalable spinlocks for non-uniform memory access (NUMA). In some examples, a global spinlock configured to protect access to a shareable resource is protected by multiple local spinlocks, which are each configured to control access to the global spinlock. In a multi-socket NUMA system, the global spinlock is allocated on one of the sockets, and the local spinlocks are distributed over the multiple sockets. In some embodiments, one local spinlock is allocated on each of the multiple sockets. In other embodiments, the multiple local spinlocks may be equally distributed over the NUMA sockets. When contention for the global spinlock is low, processes can attempt to directly acquire the global spinlock. In contrast, when contention for the global spinlock is high, processes need to first acquire one of the local spinlocks associated with the global spinlock before attempting to acquire the global spinlock.

Claims

1. A computer system comprising:

a plurality of sockets, each socket of the plurality of sockets comprising at least one central processing unit (CPU) and locally attached memory, wherein each CPU in each socket is communicatively coupled to the locally attached memory of the socket;

a global spinlock configured to control access to a shareable resource, the global spinlock comprising a first flag and a second flag; and

a plurality of local spinlocks, each local spinlock of the plurality of local spinlocks configured to control access to the global spinlock;

wherein the first flag indicates whether the global spinlock is directly acquired, and the second flag indicates whether the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.

2. The system of claim 1, wherein the plurality of local spinlocks and the plurality of sockets are same in number, and each local spinlock of the plurality of local spinlocks is allocated on a respective socket of the plurality of sockets.

3. The system of claim 1, wherein the plurality of local spinlocks are equally distributed amongst the plurality of sockets.

4. The system of claim 1, wherein each local spinlock is a test-and-set spinlock.

5. The system of claim 1, wherein each local spinlock is a test and test-and-set spinlock.

6. The system of claim 1, wherein the global spinlock is permitted to be globally acquired if the first flag indicates that the global spinlock is directly acquired.

7. The system of claim 1, wherein the global spinlock is not permitted to be globally acquired if the second flag indicates that the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.

8. The system of claim 1, wherein the global spinlock is acquirable only through a local spinlock of the plurality of local spinlocks if the second flag indicates that the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.

9. A computer-implemented method to control access to a shareable resource, the method comprising:

allocating a global spinlock configured to control access to a shareable resource, the global spinlock comprising a first flag and a second flag; and

allocating a plurality of local spinlocks, each local spinlock of the plurality of local spinlocks configured to control access to the global spinlock;

wherein the first flag indicates whether the global spinlock is directly acquired, and the second flag indicates whether the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.

10. The method of claim 9, wherein the global spinlock is allocated on one socket of a plurality of sockets.

11. The method of claim 10, wherein the plurality of local spinlocks and the plurality of sockets are same in number, and each local spinlock of the plurality of local spinlocks is allocated on a respective socket of the plurality of sockets.

12. The method of claim 10, wherein the plurality of local spinlocks are distributed amongst the plurality of sockets.

13. The method of claim 9, wherein the global spinlock is one of a test and test-and-set spinlock or a test-and-set spinlock.

14. The method of claim 9, further comprising:

in response to globally acquiring the global spinlock, setting the first flag to indicate that the global spinlock is directly acquired.

15. The method of claim 9, further comprising:

in response to determining that the first flag indicates that the global spinlock is directly acquired, attempting to globally acquire the global spinlock.

16. The method of claim 15, further comprising:

in response to spinning a first specific number of times attempting to globally acquire the global spinlock without acquiring the global spinlock, attempting to acquire a first local spinlock of the plurality of local spinlocks, wherein the first local spinlock is randomly selected.

17. The method of claim 16, further comprising:

in response to acquiring the first local spinlock, attempting to locally acquire the global spinlock.

18. The method of claim 17, further comprising:

in response to locally acquiring the global spinlock, setting the second flag to indicate that the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.

19. The method of claim 16, further comprising:

in response to spinning a second specific number of times attempting to acquire the first local spinlock without acquiring the first local spinlock, attempting to acquire a second local spinlock of the plurality of local spinlocks, wherein the second local spinlock is randomly selected.

20. The method of claim 9, further comprising:

in response to determining that the second flag indicates that the global spinlock is acquired through a local spinlock configured to control access to the global spinlock, attempting to acquire a first local spinlock of the plurality of local spinlocks.

21. A computer program product including one or more non-transitory machine readable mediums encoded with instructions that when executed by one or more processors cause a process to be carried out to control access to a shareable resource, the process comprising:

allocating a global spinlock configured to control access to a shareable resource, the global spinlock comprising a first flag and a second flag; and

allocating a plurality of local spinlocks, each local spinlock of the plurality of local spinlocks configured to control access to the global spinlock;

wherein the first flag indicates whether the global spinlock is directly acquired, and the second flag indicates whether the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.

22. The computer program product of claim 21, the process further comprising:

attempting to globally acquire the global spinlock; and

in response to globally acquiring the global spinlock, setting the first flag to indicate that the global spinlock is directly acquired.

23. The computer program product of claim 21, the process further comprising:

spinning a first specific number of times attempting to acquire the global spinlock; and

in response to spinning the first specific number of times attempting to globally acquire the global spinlock without acquiring the global spinlock, attempting to acquire a first local spinlock of the plurality of local spinlocks.

24. The computer program product of claim 23, the process further comprising:

in response to acquiring the first local spinlock, attempting to locally acquire the global spinlock.

25. The computer program product of claim 24, the process further comprising:

in response to locally acquiring the global spinlock, setting the second flag to indicate that the global spinlock is acquired through a local spinlock configured to control access to the global spinlock.