This is a project to develop algorithms for nonlinear programming, particularly problems with many variables, with an emphasis on algorithms suitable for parallel computers. The focus is on Newton-like methods. The research is motivated by earlier success in developing algorithms and software for parallel unconstrained optimization. The research is primarily aimed at methods suitable for coarse-grained parallel computers, where each processor is itself moderately powerful. This includes existing machines such as the Intel iPSC hypercube and the Sequent parallel computers. Large-scale nonlinear programming problems arise frequently in industrial, economic, and military applications, however nonlinear models can be difficult to solve. This research aims to produce easy-to-use, general purpose algorithms (and software) that will make available the power of parallel computers to the users of nonlinear models, extending the range of nonlinear programming as a day-to-day modelling tool.//
