Algorithms and experiments for structural mechanics on high-performance architectures (Q1087382)

Presented here is a detailed study of the performance of some promising alterative algorithms to the stiffness equations approach on high- performance architectures. Implementations of direct and iterative methods for the solution of structural mechanics problems have been made on the Alliant FX/8 and Cray X-MP systems. The direct methods include the classical Cholesky factorization of the stiffness matrices, the natural factor method of Argyris, and a weighted least-squares method with iterative improvement of Van Loan. The iterative methods include new preconditioned conjugate gradient methods for weighted least-squares problems by Freund and for least squares with equality constraints by Balow, Nichols, and Plemmons. Such alternatives can be, of course, more computation intensive. Thus the need for efficient algorithms for high- speed architectures is apparent. Performance issues discussed here include the speedups obtained for vectorization on a Cray X-MP as well as the speedups for vectorization and concurrency on an Alliant FX/8.