cacheflush

CacheFlush is a generic term used in computing to describe an operation that clears or synchronizes a cache. In practice, a flush either writes back modified data from the cache to a lower level of storage and then invalidates the cached copy, or simply invalidates entries to prevent stale data from being used. The concept applies to hardware caches in processors and GPUs as well as to software caches used by operating systems, databases, and applications.

In hardware, caches are organized in levels and may be write-back or write-through. A flush ensures coherence and visibility of changes when data is shared with devices or when a processor context changes. Modern CPUs provide cache-flush primitives that force lines to be written back or invalidated; the exact semantics depend on the instruction set and whether the operation affects only certain lines or the entire cache.

In software, flush operations clear in-memory caches or push updates to the underlying data stores. For example, a database cache may be flushed after bulk writes to guarantee that subsequent queries see fresh data; a web or application cache may be flushed after deploying new content. Flushing trades latency for correctness and freshness.

Performance and design considerations: frequent flushing can negate cache benefits; designers balance timing, durability, and coherence with memory barriers and cache-invalidation strategies.

See also: cache coherence, cache invalidation, write-back cache, memory barrier.