Berti-Chip · Energy-Efficient Highly Accurate Data Prefetching

Data prefetchers are ubiquitous in current high-performance computers such as those manufactured by Intel, ARM, AMD, or IBM, as they play a fundamental role in hiding long-latency memory accesses. Prefetchers predict the data that will be needed by a processor in the future and fetch such data ahead of execution. State-of-the-art data prefetchers, push the limits of performance but often without caring about energy efficiency. The energy efficiency of a prefetcher is dictated by its accuracy, which measures the percentage of data moved by the prefetcher that satisfies the processor demands.Berti is a data prefetcher sited at the first-level data cache (L1D) that makes a compelling case for timeliness and accuracy. Berti can boost processors performance by 33% (with respect to those not using prefetching mechanisms) and 8.5% (when compared to mainstream prefetchers) while providing an accuracy above 90%, which translates into a low energy overhead of the memory hierarchy. In addition, Berti is a cost-effective prefetcher that just requires 2.55KB of storage.The objective of this project is to elaborate a hardware design for the Berti data prefetcher, as we strongly believe that the notable boost in processor performance and efficiency along with its design simplicity makes Berti a serious candidate both for the emerging low-power edge market and for high-performance computers.