A Case for Runtime-Assisted Global Cache Management

Rapport, 2016

Dead blocks are handled inefficiently in the multi-level cache hierarchies of many-core architectures because the decision whether a block is dead has to be made locally at each level. This paper introduces runtime-assisted global cache management to quickly deem blocks dead at all levels. The scheme is based on a cooperative hardware/software approach that leverages dynamic information about future region accesses. We show that our proposed runtime-assisted global cache management approach outperforms previously proposed local dead-block management schemes.