关键词：算法；通信；拓扑；消息传递接口
摘 要：Communication plays a central role in parallel computing algorithms. For collective communication, significant gains in performance can be achieved by implementing topology- and performance-aware collectives. In this chapter, we offer a comprehensive overview of the existing research in this area. We observe both message passing interface (MPI) collectives as well as alternatives from the distributed computing domain. The existing challenges in analytical and experimental solutions for heterogeneous platforms are presented and discussed.