This project will evaluate the use of supercomputers to solve large mathematical programming (MP) problems. Specifically, experimental codes will be written to implement and evaluate in practice the vector/parallel techniques identified as promising in the Phase I report. These codes will be written for the Cray X-MP and the IBM 3090 Vector facility and will evaluate techniques usable in either or both of the simplex and interior point (barrier) methods. The research aims for an understanding of the capabilities and requirements of large-scale vector processors, in efficiently carrying out the extensive vector and (especially sparse) matrix calculations associated with real MP models.
