Methods and apparatus for speculative probing with early completion and delayed request

• US 20030182514A1
• Filed: 03/22/2002
• Published: 09/25/2003
• Est. Priority Date: 03/22/2002
• Status: Active Grant

• Current Assignee: Sanmina-SCI Corporation
• Original Assignee: Newisys, Inc. (Sanmina-SCI Corporation)
• Inventors: Glasco, David B.
• Granted Patent: US 7,103,725 B2
• Time in Patent Office: Days
• US Class Current: 711/141
• CPC Class Codes: G06F 12/0813 with a network or matrix configuration
• CPC Class Codes: G06F 12/0815 Cache consistency protocols
• CPC Class Codes: G06F 2212/507 using speculative control

Abstract

According to the present invention, methods and apparatus are provided for increasing the efficiency of data access in multiple processor, multiple cluster systems. A cache coherence controller associated with a first cluster of processors can determine whether speculative probing can be performed before forwarding a data access request to a second cluster. The cache coherence controller can send the data access request to the second cluster if the data access request can not be completed locally.

Claims

1. A computer system, comprising:

a first cluster including a first plurality of processors and a first cache coherence controller, the first plurality of processors and the first cache coherence controller interconnected in a point-to-point architecture;

a second cluster including a second plurality of processors and a second cache coherence controller, the second plurality of processors and the second cache coherence controller interconnected in a point-to-point architecture, the first cache coherence controller coupled to the second cache coherence controller;

wherein the first cache coherence controller is configured to receive a cache access request originating from the first plurality of processors and send a probe to the first plurality of processors in the first cluster before the cache access request is received by a serialization point in the second cluster and wherein the first cache coherence controller is further configured to send the cache access request to the second cluster after receiving a probe response from the first node.

2. The computer system of claim 1, wherein the first cache coherence controller sends the cache access request after determining that the cache access request can not be completed locally.

3. The computer system of claim 2, wherein the cache access request can be completed locally if the cache access request is a read and the state of the cache corresponding to the cache access request is valid.

4. The computer system of claim 3, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and exclusive.

5. The computer system of claim 3, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and owned.

6. The computer system of claim 1, wherein the memory access serialization point is a memory controller in the second cluster.

7. The computer system of claim 1, wherein the first cache coherence controller is further configured to receive a probe originating from the second cluster.

8. The computer system of claim 7, wherein the probe is associated with the memory line corresponding to the cache access request.

9. The computer system of claim 8, wherein the first cache coherence controller is further configured to respond to the probe originating from the second cluster using information obtained from the probe of the first plurality of processors.

10. A computer system, comprising:

a first cluster including a first plurality of processors and a first cache coherence controller, the first plurality of processors and the first cache coherence controller interconnected in a point-to-point architecture;

a second cluster including a second plurality of processors and a second cache coherence controller, the second plurality of processors and the second cache coherence controller interconnected in a point-to-point architecture, the first cache coherence controller coupled to the second cache coherence controller and constructed to receive a cache access request originating from the first plurality of processors, send a probe to the first plurality of processors, receive responses to the probe, and send the cache access request to the second cluster after the responses to the probe have been received.

11. A cache coherence controller, the cache coherence controller comprising:

interface circuitry coupled to a plurality of local processors in a local cluster and a non-local cache coherence controller in a non-local cluster, wherein the plurality of local processors are arranged in a point-to-point architecture;

a protocol engine coupled to the interface circuitry, the protocol engine configured to receive a cache access request from a first processor in the local cluster and speculatively probe a local node, wherein the protocol engine is configured to send the cache access request to the non-local cache coherence controller in the non-local cluster after receiving a probe response from the local node associated with the cache.

12. The cache coherence controller of claim 11, wherein the protocol engine sends the cache access request after determining that the cache access request can not be completed locally.

13. The cache coherence controller of claim 12, wherein the cache access request can be completed locally if the cache access request is a read and the state of the cache corresponding to the cache access request is valid.

14. The cache coherence controller of claim 13, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and exclusive.

15. The cache coherence controller of claim 13, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and owned.

16. The cache coherence controller of claim 11, wherein the memory access serialization point is a memory controller in the non-local cluster.

17. The cache coherence controller of claim 11, further comprising determining if speculative probing of the local node can be performed.

18. The cache coherence controller of claim 17, wherein determining if speculative probing can be performed comprises verifying that a memory line associated with the cache access request is not locked.

19. The cache coherence controller of claim 18, wherein the protocol engine is further configured to receive a probe originating from a second processor, wherein the probe is associated with the memory line corresponding to the cache access request.

20. The cache coherence controller of claim 19, wherein the protocol engine is further configured to respond to the probe originating from the second processor using information obtained from the speculative probe of the local node.

21. The cache coherence controller of claim 20, further comprising a pending buffer coupled to the protocol engine.

22. The cache coherence controller of claim 21, wherein the protocol engine responds to the probe originating from the second processor by using probe information in the pending buffer.

23. The cache coherence controller of claim 22, wherein the cache coherence controller is configurable as an aggregate remote cache, a probing agent pair, a remote memory and/or as a remote processor.

24. The cache coherence controller of claim 22, wherein the protocol engine, local processors, and non-local processors support a coherence protocol.

25. The cache coherence controller of claim 24, wherein the local cache coherence controller is coupled to the non-local cache coherence controller through a point-to-point architecture.

26. The cache coherence controller of claim 24, wherein the local cache coherence controller is coupled to the non-local cache coherence controller through a switch architecture.

27. A method for a cache coherence controller to manage data access in a multiprocessor system, the method comprising:

receiving a cache access request from a local processor associated with a local cluster of processors connected through a point-to-point architecture;

determining if speculative probing of a local node associated with a cache can be performed before forwarding the cache access request to a non-local cache coherence controller, the non-local cache coherence controller associated with a remote cluster of processors connected through a point-to-point architecture, wherein the remote cluster of processors shares an address space with the local cluster of processors

sending the cache access request to the non-local cache coherence controller after receiving a probe response from the local node associated with the cache.

28. The method of claim 27, wherein the cache access request is sent after determining that the cache access request can not be completed locally.

29. The method of claim 28, wherein the cache access request can be completed locally if the cache access request is a read and the state of the cache corresponding to the cache access request is valid.

30. The method of claim 29, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and exclusive.

31. The method of claim 29, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and owned.

32. The method of claim 27, wherein speculatively probing the local node comprises sending a probe to the local node before a memory line associated the probe is locked.

33. The method of claim 27, wherein the memory access serialization point is a memory controller in the non-local cluster.

34. The method of claim 33, wherein determining if speculative probing can be performed comprises verifying that a memory line associated with the cache access request is not locked.

35. The method of claim 34, further comprising receiving a probe originating from a second processor, wherein the probe is associated with the memory line corresponding to the cache access request.

36. The method of claim 35, further comprising responding to the probe originating from the second processor using information obtained from the speculative probe of the local node.

37. An apparatus for managing data access in a multiprocessor system, the apparatus comprising:

means for receiving a cache access request from a local processor associated with a local cluster of processors connected through a point-to-point architecture;

means for determining if speculative probing of a local node associated with a cache can be performed before forwarding the cache access request to a non-local cache coherence controller, the non-local cache coherence controller associated with a remote cluster of processors connected through a point-to-point architecture, wherein the remote cluster of processors shares an address space with the local cluster of processors

means for sending the cache access request to the non-local cache coherence controller after receiving a probe response from the local node associated with the cache.

38. The apparatus of claim 37, wherein the cache access request is sent after determining that the cache access request can not be completed locally.

39. The apparatus of claim 38, wherein the cache access request can be completed locally if the cache access request is a read and the state of the cache corresponding to the cache access request is valid.

40. The apparatus of claim 39, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and exclusive.

41. The apparatus of claim 39, wherein the cache access request can be completed locally if the cache access request is a write and the state of the cache corresponding to the cache access request is valid and owned.

42. The apparatus of claim 37, wherein speculatively probing the local node comprises sending a probe to the local node before a memory line associated the probe is locked.

43. The apparatus of claim 37, wherein the memory access serialization point is a memory controller in the non-local cluster.

44. The apparatus of claim 43, wherein determining if speculative probing can be performed comprises verifying that a memory line associated with the cache access request is not locked.

45. The apparatus of claim 44, further comprising means for receiving a probe originating from a second processor, wherein the probe is associated with the memory line corresponding to the cache access request.

46. The apparatus of claim 45, further comprising means for responding to the probe originating from the second processor using information obtained from the speculative probe of the local node.

47. A method for a cache coherence controller to manage data access in a multiprocessor system, the method comprising:

receiving a cache access request originating from a first cluster of processors;

sending a probe to nodes associated with the first cluster of processors;

receiving responses to the probe; and

sending the cache access request to a second cluster of processors after receiving responses to the probe.

48. The method of claim 47, wherein the cache coherence controller is constructed to act as an aggregate remote cache.

49. The method of claim 47, wherein the cache coherence controller is constructed to act as a probing agent pair.

50. The method of claim 47, wherein the cache coherence controller is constructed to act as a remote memory.

51. The method of claim 47, wherein the cache coherence controller is constructed to act as a remote processor.