Sidecache

Sidecache is a secondary cache layer intended to complement a primary cache in a computer system. It stores data likely to be reused soon but that would be costly to fetch from main memory if not cached. Sidecache aims to reduce latency and widen effective memory bandwidth for selected workloads.

In architecture, a sidecache may be on-die with the processor, on the same package, or external but tightly integrated via a high-speed interconnect. It is managed by a cache controller or software, and kept coherent with other caches through standard coherence protocols. It can use common cache organizations and replacement policies.

Typical use cases include accelerating workloads with repetitive but not perfectly predictable access patterns, caching data near storage or I/O devices, and improving performance for accelerators that access host memory. In storage stacks, a sidecache can be implemented as RAM or as an adapter cache near SSDs, holding hot blocks to reduce latency.

Challenges include added coherence traffic, design complexity, and the risk of marginal gains if data locality is weak. Costs, power consumption, and integration with existing memory hierarchies must be balanced against expected performance improvements.

The term is not universally standardized and appears in varied vendor and research contexts. Related concepts include memory hierarchy, near-memory caches, sidecar caches in storage, and cache architectures for heterogeneous systems.