摘要: Abstract We investigate the use of a hypercube packet switching network as shared memory server for vector multiprocessors. Using generalization high performance switch node introduced in an earlier paper, we develop switched capable providing bandwidth access to memory. The exhibits adaptive behavior, absorbing conflicts operation proceeds, and delivers full all processors simultaneously. In addition its performance, has another feature that makes it attractive server. words are not equidistant from processors. A hierarchy distances occurs. By taking advantage this one can provide segments fast within global environment. This very promising general purpose parallel computer.