SYSTEM FOR MULTI THREADED MULTI PROCESSOR SHARING OF ASYNCHRONOUS HARDWARE UNITS

• US 20070261053A1
• Filed: 05/06/2006
• Published: 11/08/2007
• Est. Priority Date: 05/06/2006
• Status: Active Grant

• Current Assignee: NVIDIA Corporation
• Original Assignee: Portal Player Incorporated
• Inventors: Avkarogullari, Gokhan, Petersen, Steven
• Granted Patent: US 8,726,279 B2
• Time in Patent Office: Days
• US Class Current: 718/100
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms

Abstract

Methods and system for sharing a hardware resource in a computer system running at least one software process having multiple threads. A lock_indicator is provided in data structures within the computer system. A request is received to use the hardware resource by one of the threads that is defined to be a requesting tread. Based on the lock_indicator, it is determined whether the hardware resource is available for use by the requesting thread. If this indicates that the hardware resource is available, the lock_indicator is set under control of the hardware resource to instead indicate that the hardware resource is unavailable, and a go_indicator signals to indicate that use of the hardware resource for the request can now proceed.

Claims

1. A method for sharing a hardware resource in a computer system running at least one software process having multiple threads, the method comprising:

(a) providing a lock_indicator in data structures within the computer system;

(b) receiving a request to use the hardware resource by one of the threads defined to be a requesting tread;

(c) determining based on said lock_indicator whether the hardware resource is available for use by said requesting thread; and

(d) if said (c) indicates that the hardware resource is available;

(1) setting said lock_indicator under control of the hardware resource to instead indicate that the hardware resource is unavailable; and

(2) signaling with a go_indicator to indicate that use of the hardware resource for said request can proceed.

2. The method of claim 1, wherein:

said (a) includes further providing a write-waiter_list in said data structures that is able to store at least one thread-identity_value;

and the method further comprising;

(e) if said (c) indicates that the hardware resource is unavailable;

(1) adding a said thread-identity_value identifying said requesting thread to said write-waiter_list; and

(2) waiting said requesting thread pending availability of the hardware resource.

3. The method of claim 2, wherein:

said (e)(1) is performed atomically with respect to the threads, under control of the hardware resource.

4. The method of claim 1, wherein:

said (a) includes further providing a ready_indicator in said data structures;

and the method further comprising;

(e) upon completion of said request;

(1) setting said ready_indicator under control of the hardware resource, thereby signaling the requesting thread that said request is completed; and

(2) resetting said lock_indicator under control of the hardware resource, thereby indicating that the hardware resource is again available.

5. The method of claim 4, wherein:

said (a) includes further providing a read-waiter_list in said data structures able to store at least one thread-identity_value;

and the method further comprising;

(f) determining based on said ready_indicator whether the hardware resource has completed said request; and

(g) if said (f) indicates that the hardware resource has not completed said request;

(1) adding a said thread-identity_value identifying said requesting thread to said read-waiter_list; and

(2) waiting said requesting thread pending completion of said request.

6. The method of claim 5, wherein:

said (g)(1) is performed atomically with respect to the threads, under control of the hardware resource.

7. The method of claim 4, further comprising:

(f) after said (e), determining whether any of the threads are waiting to write or to read the hardware resource and if so;

(1) generating an interrupt in the computer system under control of the hardware resource; and

(2) within said interrupt, signaling at least one of the threads that are waiting to write or to read the hardware resource that it should proceed.

8. The method of claim 7, wherein:

said (e) and said (f) are performed atomically with respect to the threads, under control of the hardware resource.

9. The method of claim 7, wherein:

said (f)(1) is performed atomically with respect to the threads, under control of the hardware resource.

10. A system for sharing a hardware resource in a computer system able to run at least one software process having multiple threads, comprising:

data structures within the computer system that include a lock_indicator;

a logic to receive a request to use the hardware resource by one of the threads defined to be a requesting tread;

a lock logic to determine based on said lock_indicator whether the hardware resource is available for use by said requesting thread; and

a logic responsive to said lock logic to control;

a logic to set said lock_indicator under control of the hardware resource to instead indicate that the hardware resource is unavailable; and

a logic to signal with a go_indicator to indicate that use of the hardware resource for said request can proceed.

11. The system of claim 10, wherein:

said data structures further include a write-waiter_list able to store at least one thread-identity_value; and

said logic responsive to said lock logic is also to control;

a logic to add a said thread-identity_value identifying said requesting thread to said write-waiter_list; and

a logic to wait said requesting thread pending availability of the hardware resource.

12. The system of claim 11, wherein said logic to add operates atomically with respect to the threads, under control of the hardware resource.

13. The system of claim 10, wherein:

said data structures further include a ready_indicator;

and the system further comprises;

a logic responsive to completion of said request to control;

a logic to set said ready_indicator under control of the hardware resource, thereby signaling the requesting thread that said request is completed; and

a logic to reset said lock_indicator under control of the hardware resource, thereby indicating that the hardware resource is again available.

14. The system of claim 13, wherein:

said data structures further include a read-waiter_list able to store at least one thread-identity_value;

and the system further comprises;

a ready logic to determine based on said ready_indicator whether the hardware resource has completed said request; and

a logic responsive to said ready logic to control;

a logic to add a said thread-identity_value identifying said requesting thread to said read-waiter_list; and

a logic to wait said requesting thread pending completion of said request.

15. The system of claim 14, wherein said logic to add operates atomically with respect to the threads, under control of the hardware resource.

16. The system of claim 13, further comprising:

a waiter logic to determine whether any of the threads are waiting to write or to read the hardware resource; and

a logic responsive to said waiter logic to control;

a logic to generate an interrupt in the computer system under control of the hardware resource; and

a logic to signal, within said interrupt, that at least one of the threads that are waiting to write or to read the hardware resource should proceed.

17. The system of claim 16, wherein said logic responsive to completion of said request to control, said logic to set said ready_indicator, said logic to reset said lock_indicator, and said waiter logic operate atomically with respect to the threads, under control of the hardware resource.

18. The system of claim 16, wherein said logic to generate operates atomically with respect to the threads, under control of the hardware resource.

19. A system for sharing a hardware resource in a computer system running at least one software process having multiple threads, comprising:

data structures within the computer system including a lock_indicator;

means for receiving a request to use the hardware resource by one of the threads defined to be a requesting tread;

lock means for determining based on said lock_indicator whether the hardware resource is available for use by said requesting thread; and

means responsive to said lock means for controlling;

means for setting said lock_indicator under control of the hardware resource to instead indicate that the hardware resource is unavailable; and

means for signaling with a go_indicator that use of the hardware resource for said request can proceed.

20. The system of claim 19, wherein:

said data structures further include a write-waiter_list to store at least one thread-identity_value; and

said means responsive to said lock means is further for controlling;

means for adding a said thread-identity_value identifying said requesting thread to said write-waiter_list that operates atomically with respect to the threads, under control of the hardware resource; and

means for waiting said requesting thread pending availability of the hardware resource.

21. The system of claim 19, wherein:

said data structures further include a ready_indicator;

and further comprising;

means responsive to completion of said request for controlling;

means for setting said ready_indicator under control of the hardware resource; and

means for resetting said lock_indicator under control of the hardware resource.

22. The system of claim 21, wherein:

said data structures further include a read-waiter_list to store at least one thread-identity_value;

and further comprising;

ready means for determining based on said ready_indicator whether the hardware resource has completed said request; and

means responsive to said ready means for controlling;

means for adding a said thread-identity_value identifying said requesting thread to said read-waiter_list that operates atomically with respect to the threads, under control of the hardware resource; and

means for waiting said requesting thread pending completion of said request.

23. The system of claim 21, further comprising:

waiter means for determining whether any of the threads are waiting to write or to read the hardware resource; and

means responsive to said waiter means to control;

means for generating an interrupt in the computer system under control of the hardware resource that operates atomically with respect to the threads, under control of the hardware resource; and

means for signaling, within said interrupt, that at least one of the threads that are waiting to write or to read the hardware resource should proceed.

24. The system of claim 23, wherein said means responsive to completion of said request for controlling, said means for setting said ready_indicator, said means for resetting said lock_indicator, and said waiter means operate atomically with respect to the threads, under control of the hardware resource.