Large Scale Structural Analysis on Massively Parallel Computers

We discuss the use of a massively parallel computer in large scale structural analysis computations. In particular, we consider the necessary modifications to a standard industrial code and a strategy where the code may be allowed to evolve gradually from a pure F77 version to a form more suitable for a range of parallel computers. We present preliminary computational results from a DECmpp computer, showing that we can achieve good price performance when compared to traditional vector supercomputers.