System and method for aggregating core-cache clusters in order to produce multi-core processors
View Patent ↗According to one embodiment of the invention, a processor comprises a memory, a plurality of core-cache clusters and a scalability agent unit that operates as an interface between an on-die interconnect and multiple core-cache clusters. The scalability agent operates in accordance with a protocol to ensure that the plurality of core-cache clusters appear as a single caching agent.
1. A method comprising:
providing a plurality of core-cache clusters, each of the plurality of core-cache clusters including a cache memory and at least one processor core in communication with the cache memory; and
operating in accordance with a protocol to ensure that the plurality of core-cache clusters appear as a single caching agent.
2. The method of claim 1 , wherein the operating in accordance with a protocol comprises controlling, by at least one scalability agent, all coherent transactions issued by the plurality of core-cache clusters.
3. The method of claim 2 , wherein the at least one scalability agent issues snoop requests to core-cache clusters of the plurality of core-cache clusters while concurrently issuing a transaction to a system interconnect.
4. The method of claim 3 , wherein the at least one scalability agent collects snoop responses from the core-cache clusters.
5. The method of claim 4 , wherein the at least one scalability agent issues an aggregate response to a requesting core-cache cluster of the plurality of core-cache clusters.
6. The method of claim 4 , wherein the at least one scalability agent performs address checking operations in order to ensure that only one transaction for a given address can be issued to the system interconnect.
7. The method of claim 2 , wherein the at least one scalability agent issues snoop requests to the plurality of core-cache clusters and collects snoop responses from all of the plurality of core-cache clusters.
8. The method of claim 7 , wherein the at least one scalability agent sends a single snoop response to a requesting socket.
9. The method of claim 1 , wherein the scalability agent includes a plurality of scalability agent units each uniquely corresponding to a core-cache cluster of the plurality of core-cache clusters and responsible for a subset of an address space.
10. A method comprising:
configuring a multi-core processor with a plurality of core-cache clusters, each core-cache cluster including at least one processor core and a scalability agent unit to manage a flow of information with the plurality of core-cache clusters; and
coupling the multi-core processor to a system interconnect, the plurality of core-cache clusters to appear as a single caching agent to devices communicatively coupled to the system interconnect.
11. The method of claim 10 further comprising:
handling local coherence between the plurality of core-cache clusters by each scalability agent unit issuing snoops and collecting responses to check for address conflicts.
12. The method of claim 11 , wherein each of the plurality of scalability agent units is further adapted to send a snoop response message aggregated from the collected snoop responses over the system interconnect so that the plurality of core-cache clusters appear as the single caching agent.
13. The method of claim 11 , wherein each scalability agent unit is responsible for a different portion of a system memory address space.
14. The method of claim 11 , wherein the configuring of the multi-core processor further comprises configuring each scalability agent to handle an equal amount of work.
15. A system comprising:
a memory; and
a processor coupled to the memory, the processor comprises:
an on-die interconnect,
a plurality of core-cache clusters, each of the plurality of core-cache clusters including both a cache memory and one or more processor cores in communication with the cache memory, and
a scalability agent coupled to the on-die interconnect, the scalability agent operating in accordance with a protocol to ensure that the plurality of core-cache clusters appear as a single caching agent.
16. The system of claim 15 , wherein the scalability agent of the processor includes a plurality of scalability agent units each uniquely associated with one of the plurality of core-cache clusters and adapted to issue a snoop message and collect snoop responses to the snoop message.
17. The system of claim 16 , wherein each of the plurality of scalability agent units of the processor is further adapted to send a snoop response message aggregated from the collected snoop responses over a system interconnect so that the plurality of core-cache clusters appear as the single caching agent.
18. The system of claim 16 , wherein the plurality of scalability agent units of the processor correspond in number to the plurality of core-cache clusters and each of the plurality of scalability agent units.
19. The system of claim 15 , wherein the plurality of core-cache clusters of the processor comprises a first core-cache cluster including (i) a first cache memory, (ii) a first plurality of processor cores in communication with the first cache memory, and (iii) a first scalability agent unit coupled to the on-die interconnect and in communication with the first cache memory and the first plurality of processor cores, the first scalability agent being responsible for a first portion of a system memory.
20. The system of claim 19 wherein the plurality of core-cache clusters of the processor further comprises a second core-cache cluster including (i) a second cache memory, (ii) a second plurality of processor cores in communication with the second cache memory, and (iii) a second scalability agent unit coupled to the on-die interconnect and in communication with the second cache memory and the second plurality of processor cores, the second scalability agent being responsible for a second portion of the system memory differing from the first portion of the system memory.