Allocation and write policy for a glueless area-efficient directory cache for hotly contested cache lines
View Patent ↗Methods and apparatus relating to allocation and/or write policy for a glueless area-efficient directory cache for hotly contested cache lines are described. In one embodiment, a directory cache stores data corresponding to a caching status of a cache line. The caching status of the cache line is stored for each of a plurality of caching agents in the system. An write-on-allocate policy is used for the directory cache by using a special state (e.g., snoop-all state) that indicates one or more snoops are to be broadcasted to all agents in the system. Other embodiments are also disclosed.
1. An apparatus comprising:
a first agent to receive a request corresponding to a cache line associated with the first agent, from a second agent; and
a directory cache, coupled to the first agent, to store data corresponding to a caching status of the cache line in each of a plurality of caching agents coupled to the first agent,
wherein an entry in the directory cache is to be allocated in response to a miss, corresponding to the request, in the directory cache and a determination that the cache status of the cache line is to be updated to a snoop-all state, based on a snoop response from a different agent than the first agent, and wherein the snoop-all state is to indicate that one or more snoops are to be broadcasted to all of the plurality of caching agents in response to detection of the snoop-all state.
2. The apparatus of claim 1 , wherein the caching status is to be determined based on a presence vector corresponding to a plurality of prior requests or snoop responses from two or more corresponding caching agents of the plurality of caching agents.
3. The apparatus of claim 2 , further comprising a memory to store a copy of the presence vector with a copy of the cache line.
4. The apparatus of claim 1 , wherein the caching status is to be determined based on a presence vector corresponding to a plurality of prior requests or snoop responses from all of the plurality of caching agents.
5. The apparatus of claim 4 , further comprising a memory to store a copy of the presence vector with a copy of the cache line.
6. The apparatus of claim 1 , wherein entries are to be allocated in the directory cache for cache lines having a high probability of encountering a future snoop critical access.
7. The apparatus of claim 1 , wherein the cache line has a state selected from a group consisting of: a modified state and exclusive state.
8. The apparatus of claim 1 , wherein the cache line has a state other than invalid or shared.
9. The apparatus of claim 1 , wherein the caching status is to be determined based on a prior request or snoop response from a corresponding caching agent of the plurality of caching agents.
10. The apparatus of claim 1 , wherein in response to a hit, corresponding to the request, in the directory cache and a determination that the request is a local request, a corresponding entry in the directory cache is to be deallocated.
11. The apparatus of claim 1 , wherein the first agent is to send one or more snoops to one or more of the plurality of caching agents identified by the directory cache to have a copy of the data corresponding to the request.
12. The apparatus of claim 1 , wherein the entry in the directory cache is to be allocated in response to response forwarded from a non-local caching agent.
13. The apparatus of claim 1 , wherein two or more of the first agent, the second agent, and the directory cache are on a same integrated circuit die.
14. A method comprising:
receiving a request corresponding to a cache line associated with a first agent, from a second agent;
storing data corresponding to a caching status of the cache line in each of a plurality of caching agents coupled to the first agent; and
allocating an entry in the directory cache in response to a miss, corresponding to the request, in the directory cache and a determination that the cache status of the cache line is to be updated to a snoop-all state, based on a snoop response from a different agent than the first agent, and wherein the snoop-all state is to indicate that one or more snoops are to be broadcasted to all of the plurality of caching agents in response to detection of the snoop-all state.
15. The method of claim 14 , further comprising determining the caching status based on a presence vector corresponding to a plurality of prior requests or snoop responses from two or more corresponding caching agents of the plurality of caching agents.
16. The method of claim 14 , further comprising determining the caching status based on a presence vector corresponding to a plurality of prior requests or snoop responses from all of the plurality of caching agents.
17. The method of claim 14 , further comprising allocating entries in the directory cache for cache lines having a high probability of encountering a future snoop critical access.
18. The method of claim 14 , further comprising determining the caching status based on a prior request or snoop response from a corresponding caching agent of the plurality of caching agents.
19. The method of claim 14 , further comprising deallocating an entry in the directory cache in response to a hit, corresponding to the request, in the directory cache and a determination that the request is a local request.
20. The method of claim 14 , further comprising the first agent sending one or more snoops to one or more of the plurality of caching agents identified by the directory cache to have a copy of the data corresponding to the request.
21. A computing system comprising:
a memory to store a copy of data stored in a directory cache;
a first processor to receive a request corresponding to a cache line associated with the first processor, from a second processor; and
a directory cache, coupled to the first processor, to store data corresponding to a caching status of the cache line in each of a plurality of caching agents coupled to the first processor,
wherein an entry in the directory cache is to be allocated in response to a miss, corresponding to the request, in the directory cache and a determination that the cache status of the cache line is to be updated to a snoop-all state, based on a snoop response from a different agent than the first processor, and wherein the snoop-all state is to indicate that one or more snoops are to be broadcasted to all of the plurality of caching agents in response to detection of the snoop-all state.
22. The system of claim 21 , wherein the caching status is to be determined based on a presence vector corresponding to a plurality of prior requests or snoop responses from two or more corresponding caching agents of the plurality of caching agents.
23. The system of claim 22 , wherein the memory is to store a copy of the presence vector with a copy of the cache line.
24. The system of claim 21 , wherein the caching status is to be determined based on a presence vector corresponding to a plurality of prior requests or snoop responses from all of the plurality of caching agents.
25. The system of claim 24 , wherein the memory is to store a copy of the presence vector with a copy of the cache line.
26. The system of claim 21 , wherein entries are to be allocated in the directory cache for cache lines having a high probability of encountering a future snoop critical access.
27. The system of claim 21 , wherein the cache line has a state selected from a group consisting of: a modified state and exclusive state.
28. The system of claim 21 , wherein the cache line has a state other than invalid or shared.
29. The system of claim 21 , wherein the caching status is to be determined based on a prior request or snoop response from a corresponding caching agent of the plurality of caching agents.
30. The system of claim 21 , wherein in response to a hit, corresponding to the request, in the directory cache and a determination that the request is a local request, a corresponding entry in the directory cache is to be deallocated.