Fast Algorithms for Structural Optimization, Least Squares and Other Computations.
Abstract
Fast algorithms for solving large-scale structural optimization and least squares problems are being investigated. An especially significant aspect of this work is the development and testing of parallel algorithms for alternatives to the often ill-conditioned stiffness equations approach in structural analysis on machines such as the Cray X-MP, the Denelcor HEP and the Intel Hypercube. The principal thrusts in this project on least squares methods have been in developing techniques for the solution of superlarge problems in a stable way, i.e., employing orthogonal factorization techniques, on multiprocessors. These computations involve various levels of parallelism, including domain decomposition as well as pipelining type schemes for orthogonal factorization. Additional keywords: linear algebra; parallel processing. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 19, 1985
- Accession Number
- ADA159136
Entities
People
- R. J. Plemmons
Organizations
- North Carolina State University