摘要：We introduce a multi-level prefetching framework with three setups, respectively aimed tominimize cost (Mincost), minimize losses in individual applications (Minloss) or maximizeperformance with moderate cost (Maxperf). Performance is boosted in all cases by a sequentialtagged prefetcher in the L1 cache, with an effective static degree policy. In both cache levels (L1and L2), we also apply prefetch filters. In the L2 cache we use a novel adaptive policy that selectsthe best prefetching degree within a fixed set of values, by tracking the performance gradient.Mincost resorts to sequential tagged prefetching in the L2 cache as well. Minloss relies on anaccurate, home-made, correlating prefetcher (PDFCM, Differencial Finite Context MethodPrefetcher). Maxperf maximizes performance at the expense of slight performance losses in asmall number of benchmarks, by integrating a sequential tagged prefetcher with PDFCM in the L2cache
