Submission code for course term project : E0243 High performance computer architecture.
Project Name: A METHOD FOR COMBINING PREFETCHING ALGORITHMS USING ORCHESTRATION TO IMPROVE OVERALL ACCURACY, COVERAGE AND ADAPTIVITY ON VARYING ADDRESS STREAMS
Please read the proposal here.
The following results have been obtained on ChampSim simulator using SPEC 2017 benchmark program traces available here. The x-axis represents benchmark id.
We use our proposal to create a hybrid prefetcher for only the L2 cache using two separate prefetchers - Berti prefetcher and PPF prefetcher. The overall accuracy on the prefetches for L2 cache has improved for various benchmarks. 
For some benchmarks, it remain same while for others it gets reduced. Please note that overall cache miss count depends on all the caches - L1D, L1C, L2 and LLC. We've implemented our proposal on only the L2 cache.
