摘要 |
A back-end vector parallel computer system suitable for supercomputing in engineering and science, comprising N vector processor units and a cubic array of N3 memory banks which are shared by N boards and are accessed through row and column common busses wired on all boards in the lattice form from vector processors through anyone of four routing ways in both 2- and 3- dimensional problems, and practicing computation of data accessed through any routing way in parallel for vectors on any section of data array with the selected direction, each vector data being processed by each corresponding vector processor, such parallel computations being usually repeated to the direction successively, and further changing alternatingly the direction of access if necessary for computation to be continued.
|