Instruction Based Management of Faulty Data Caches

We propose a new approach to mitigate the impact of faulty bits in data caches. Our technique assumes that faulty caches are enhanced with the ability of disabling their defective parts at cache subblock granularity. Our experiments reveal that while the occurrence of hard-errors in faulty caches may have a significant impact in performance, a lot of room for improvement exists, if someone is able to take into account the spatial reuse patterns of the to-be-referenced blocks (not all the data fetched into the cache is accessed). We propose frugal PC-indexed spatial predictors to orchestrate the (re)placement decisions among the fully and partially faulty blocks. Using cycleaccurate simulations, we show that our approach is able to offer significant benefits.