TAP: Taxonomy for Adaptive Prefetching

Data prefetching is an important technique for overcoming the increasing gap between the processor and memory speeds. Traditonal metrics for prefetching have been shown to be ineffective and misleading. Prefetch taxonomies provide interesting insights into the performance of prefetch algorithms. In this report, we present an analysis of two prefetch algorithms using a prefetch classification taxonomy PTMT. We also come up with the drawbacks of PTMT and propose an alternative taxonomy TAP. An adaptive tagged prefetch algorithm (A-TAG) that uses TAP is also proposed and analyzed.