Page-mapping techniques for CC-NUMA multiprocessors

Abstract

Careful page mapping has been shown in the past to be effective for reducing cache conflicts on both uniprocessor and Uniform Memory Access (UMA) multiprocessors. This paper extends previous page-mapping schemes to the more recent Cache-Coherent Non-Uniform Memory Access (CC-NUMA) multiprocessors. These extensions maintain the program's data-task affinity, which is important to CC-NUMA, while reducing cache set conflicts by carefully selecting the page frames. Using an execution-driven simulator that simulates a CC-NUMA machine with a 4-MB secondary cache and a 16-KB primary cache on each of the 4-issue super-scalar processors, we find that, when non-coherence cache misses are relatively heavy, it is quite important for page mapping to preserve the compiler-generated memory module ID (MID) which determines data distribution among the processors. We also find that straight application of page-coloring performs worse than bin-hopping by 10-45%, while by hashing the page color with part of the MID, page-coloring can perform closely to bin-hopping.