Systems and methods for coherence in clustered cache hierarchies
According to embodiments described herein, the hierarchical complexity for coherence protocols associated with clustered cache architectures can be encapsulated in a simple function, i.e., that of determining when a data block is shared entirely within a cluster (i.e., a sub-tree of the hierarchy) and is private from the outside. This allows embodiments to eliminate complex recursive coherence operations that span the hierarchy and instead employ simple coherence mechanisms such as self-invalidation and write-through but which are restricted to operate where a data block is shared. Thus embodiments recognize that, in the context of clustered cache hierarchies, data can be shared entirely within one cluster but can be private (unshared) to this cluster when viewed from the perspective of other clusters. This characteristic of the data can be determined and then used to locally simplify coherence protocols.
1. A method for cache coherence in a computer system having a clustered cache hierarchy, the method comprising:
storing a common shared level (CSL) value for a data block stored in the clustered cache hierarchy;
when the data block is written, using a coherence mechanism to update a status of the data block for one or more caches within a cache cluster indicated by the CSL value and treating the data block as private for one or more caches outside of the cache cluster indicated by the CSL value,
detecting a translation look-aside buffer (TLB) miss associated with the data block; and
determining a new CSL value for the data block in response to the detecting step.
2. The method of claim 1 , wherein the clustered cache hierarchy includes at least three levels of cache memories wherein a first level, which is connected to multiple processor cores, includes at least three level one cache memories and wherein in each level with at least two cache memories, at least a pair of the at least two cache memories are connected to, and share, one of the cache memories at a higher level that is closer to main memory.
3. The method of claim 1 , wherein the step of using a cache coherence mechanism further comprises:
self-invalidating the data block in the one or more caches within the cache cluster indicated by the CSL value.
4. The method of claim 3 , wherein the one or more caches within the cache cluster indicated by the CSL value which are self-invalidated include each level one cache and, when at least one intermediate cache is present in the cluster, each intermediate cache in the cluster.
5. The method of claim 1 , wherein the CSL value and an identity of a core which first accessed the data block defines the cluster within the clustered cache hierarchy.
6. The method of claim 5 , wherein the cluster includes one or more level two caches and one or more level 3 caches, wherein when the CSL value is three or higher, the one or more level 2 caches within the cluster are intermediate caches.
7. The method of claim 1 , wherein the step of determining a new CSL value further comprises:
retrieving a stored identity of a core which is a first owner of the data block;
determining an identity of another core which attempted to access the data block causing the TLB miss; and
determining to be said new CSL value, a lowest common cache level of the core which is the first owner of the data block and the another core.