Method,system and apparatus for main memory access subsystem usage to different partitions in a socket with sub-socket partitioning

• US 20090164751A1
• Filed: 11/07/2008
• Published: 06/25/2009
• Est. Priority Date: 12/20/2007
• Status: Active Grant

• Current Assignee: Intel Corporation
• Original Assignee: Intel Corporation
• Inventors: Thomas, Tessil, Harikumar, Ajay, Simon, Biju Puthur
• Granted Patent: US 8,370,508 B2
• Time in Patent Office: Days
• US Class Current: 712/1
• CPC Class Codes: G06F 12/0813 with a network or matrix configuration
• CPC Class Codes: G06F 9/5016 the resource being the memory
• CPC Class Codes: G06F 9/5061 Partitioning or combining of resources

Abstract

Embodiments enable sub-socket partitioning that facilitates access among a plurality of partitions to a shared resource. A round robin arbitration policy is to allow each partition, within a socket, that may utilize a different operating system, access to the shared resource based at least in part on whether an assigned bandwidth parameter for each partition is consumed. Embodiments may further include support for virtual channels.

Claims

1. A method comprising:

allocating a bandwidth parameter for each of a plurality of partitions in a socket;

decrementing the bandwidth parameter for each partition for each completed request from the corresponding partition; and

rotating arbitration priority among partitions of the plurality of partitions having a non-zero bandwidth parameter, otherwise rotating the arbitration priority among the plurality of partitions if all of the plurality of partitions have a zero bandwidth parameter.

2. The method of claim 1 wherein the bandwidth parameter is a number of allocated cycles consumed for a shared resource for a request from the corresponding partition.

3. The method of claim 1 wherein the bandwidth parameter is utilized during a defined time period.

4. The method of claim 1 wherein the bandwidth parameter is stored in a register in a shared resource for each partition..

5. The method of claim 2 wherein the number of allocated cycles consumed is tracked in a consumed cycles counter in a shared resource for each partition.

6. The method of claim 3 wherein the defined time period is tracked with an epoch counter in a shared resource for each partition.

7. A system comprising:

a processor;

a dynamic random access memory, coupled to the processor, to receive requests from the processor;

the processor to support sub-socket partitioning to utilize at least a first and a second operating system within a first partition and a second partition;

a shared resource, coupled to the processor, with;

a first counter to define a time period;

a register, coupled to the first counter, to store a bandwidth parameter for the first partition and the second partition; and

a second counter, for each partition, coupled to the first counter, to track a number of consumed cycles for each request from the corresponding partition that is used for access to the shared resource.

8. The system of claim 7 wherein the bandwidth parameter is a number of allocated cycles for each partition.

9. The system of claim 7 wherein the bandwidth parameter is utilized during the defined time period stored in the first counter.

10. The system of claim 7 wherein the shared resource is a home agent.

11. The system of claim 7 wherein the home agent is a cache controller.

12. An agent to support sub-socket partitioning for a first partition and a second partition comprising:

a first counter to define a time period for both the first partition and the second partition;

a register, coupled to the first counter, to store a bandwidth parameter for both the first partition and the second partition;

a second counter, coupled to the first counter, to track a number of consumed cycles for the first partition that is used for access to the agent; and

a third counter, coupled to the first counter, to track a number of consumed cycles for the second partition that is used for access to the agent.

13. The agent of claim 12 wherein the bandwidth parameter is a number of allocated cycles for each partition.

14. The agent of claim 12 wherein the bandwidth parameter is utilized during the defined time period stored in the first counter.

15. The agent of claim 12 wherein the agent is a home agent.

16. The agent of claim 15 wherein the home agent is a cache controller.

17. A processor comprising:

a plurality of processor cores that support sub-socket partitioning such that each one of the plurality of processor cores can utilize a different operating system;

an interface, coupled to the processor cores, to generate and transmit a packet to a plurality of agents coupled to the plurality of processor cores via the interface;

a transmit logic, integrated within the interface, with;

a queue for each virtual channel supported by the interface to store all requests for each partition associated with the processor; and

a counter, coupled to the queue, to be incremented when the virtual channel transmits the packet.

18. The processor of claim 17 further comprising a plurality of last level cache banks.

19. The processor of claim 17 wherein the interface is to transmit the packet based at least in part on a round robin arbitration policy that rotates arbitration priority among partitions of the plurality of partitions having a non-zero bandwidth parameter, and otherwise rotates the arbitration priority among the plurality of partitions if all of the plurality of partitions have a zero bandwidth parameter.

20. The processor of claim 19 wherein the bandwidth parameter is a number of allocated cycles consumed for transmission of the packet to a shared resource.