A Self-Tuning Algorithm for Managing Reconfigurable Cache

Reconfigurable cache with a set of adjustable configurations can be configured dynamically to adapt itself to the change in program characteristics and has tremendous benefits for performance and energy. However, how to dynamically manage reconfigurable cache is still a cumbersome task left for designers. We introduce a self-tuning algorithm (ETCA), which can dynamically manage reconfigurable cache on a per-phase basis. In contrast with previous works, ETCA seeks not only to lower cache´s energy consumption effectively, but also reduce the performance loss due to unnecessary reconfigurations. By simulating numerous MiBench benchmark, the results show that ETCA, when applied to reconfigurable cache, saves on average 38% of total memory access energy compared with a conventional cache and the associated performance loss is close to 1.8%