发明名称 System for parallel computation with three phase processing in processor tiers in which new instructions trigger execution and forwarding
摘要 Multicomputer systems frequently consist of a network of node computers connected to one another in a particular topology, and at least one host computer which controls the network. To execute a task, the host computer loads a program into the node computer, and this is then followed by data, either from the host computer or possibly from a peripheral device also connected to the network, which are processed in accordance with the loaded program. In order to then avoid during the execution of plurality of similar tasks in succession, as is the case during the processing of successive images for instance, that the network has to be loaded again in each case, the network program is designed in such a way that all node computers execute a cycle composed of three parts, wherein in the first part an instruction for starting a task is received from the host computer, in the second part execution is effected with exchange of data packets between node computers, and in the third part a ready message is collected from all node computers, following which all node computers await a new instruction from the host computer in order to execute the same or a similar task with new data. In addition, various methods of information transfer in the network of node computers are described.
申请公布号 US5317735(A) 申请公布日期 1994.05.31
申请号 US19910715193 申请日期 1991.06.14
申请人 U.S. PHILIPS CORPORATION 发明人 SCHOMBERG, HERMANN
分类号 G06F15/173;G06F15/80;(IPC1-7):G06F15/16 主分类号 G06F15/173
代理机构 代理人
主权项
地址