Implementation of preconditioned s-step Conjugate Gradient methods on a multiprocessor system with memory hierarchy
Anthony Theodore Chronopoulos, C. William Gear · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1987
Multiprocessor architectures combining vector and parallel processing capabilities on a two-level shared memory structure have been implemented. This memory hierarchy structure requires that numerical algorithms possess good data locality in order to achieve high performance rates. The s-step Conjugate Gradient method is a generalization of the classical method with improved data locality and parallel properties. Numerical experiments on a multiprocessor with memory hierarchy show the superiority in performance of the s-step methods to their one-step counterparts.