摘要: What should the architecture of each node in a general purpose, massively parallel (MPA) be? We frame question concrete terms by describing two fundamental problems that must be solved well any purpose MPA. From this, we systematically develop required logical organization an MPA node, and present some details *T (pronounced Start, designed to these requirements. is direct descendant dynamic dataflow architectures, unifies them with von Neumann architectures. discuss hand-compiled example compilation issues.