作者: Robert Schreiber , Steven W. Hammond
DOI:
关键词:
摘要: We present a highly parallel graph mapping technique that enables one to solve unstructured grid problems on massively computers. Many implicit and explicit methods for solving discretizated partial differential equations require each point in the discretization exchange data with its neighboring points every time step or iteration. The spent communicating can limit high performance promised by computing. To eliminate this bottleneck, we map of irregular problem representing interconnection topology computer such sum distances messages travel is minimized. show that, comparison naive assignment processors, our heuristic algorithm significantly reduces communication Connection Machine, CM-2.