Self-healing coarse-grained snoop filter

Abstract

The invention relates to filtering snoops in coherent multiprocessor systems. For example, in response to a request to update a target memory location at a Level-2 (L2) cache shared among multiple local processing units each having a Level-1 (L1) cache, a lookup based on the target memory location may be performed in a snoop filter that tracks entries in the LI caches. If the lookup misses the snoop filter and the snoop filter lacks space to store a new entry, a victim entry to evict from the snoop filter may be selected and a request to invalidate every cache line that maps to the victim entry may be sent to at least one of the processing units with one or more cache lines that map to the victim entry. The victim entry may then be replaced in the snoop filter with the new entry corresponding to the target memory location.

Description

technical field [0001] Aspects and embodiments described herein relate generally to coherent multiprocessor systems, and in particular, to self-healing coarse-grain snoop filters that can filter snoops to one or more processors in a multiprocessor system. Background technique [0002] A typical multiprocessor integrated circuit (ie, chip) utilizes multiple processor cores interconnected using an interconnect bus. In general, each processor core is supported by one or more cache memories, where each cache memory is usually stored in main memory and cache memory in blocks of fixed size, usually called "cache lines." Data files transferred between. In conventional directory-based approaches, each cache memory contains a directory containing all addresses associated with data files cached therein. Data cached at each processor core can be shared among all other processor cores on the interconnect bus. Thus, a multiprocessor system can potentially have many copies of the same ...

AI Technical Summary

Problems solved by technology

[0005] While snooping protocols generally scale better than directory-based protocols, snooping protocols have scaling weaknesses, whereby an increase in the number of active processors results in a corresponding increase in the amount of snoop traffic each active processor receives

Therefore, snoop filtering usually has the primary goal of reducing the number of unnecessary snoops as much as possible without incurring an area or latency cost, or reducing filter snoops due to accumulated false positives over time. Ability

Images