Application-specific memory partitioning for joint energy and lifetime optimization

Power management of caches based on turning idle cache lines into a low-energy state is also beneficial for the aging effects caused by Negative Bias Temperature Instability (NBTI), provided that idleness is correctly exploited; unlike energy, aging, being a measure of delay, is in fact a worst-case metric. In this work we propose an application-specific partitioned cache architecture in which a cache is organized as a set of independently addressable sub-blocks; by properly using the idleness of the various banks to drive how the partition is determined, it is possible to extend the effective lifetime of the cache while saving extra energy. Two are the distinctive features of our approach: First, we allow the cache sub-blocks age at different rates, achieving a sort of graceful degradation of performance while extending lifetime beyond the limits of previously published works. Proper architectural arrangements are also introduced in order to cope with the issue of using a progressively smaller cache. Second, the sub-blocks have non-uniform sizes, so to maximally exploit idleness for joint energy and aging optimization. Simulation results show that it is possible to extend the effective lifetime of the cache by more than 2x with respect to previous methods, while concurrently improving energy consumption by about 50%.