This thesis presents efficient algorithms for internal and external parallel sorting and remote data update. The sorting algorithms approach the problem by concentrating first on highly efficient but incorrect algorithms followed by a cleanup phase that completes the sort. The remote data update algorithm, rsync, operates by exchanging block signature information followed by a simple hash search algorithm for block matching at arbitrary byte boundaries. The last chapter of the thesis examines a number of related algorithms for text compression, differencing and incremental backup.

www.hpcresearch.nl

A preliminary MPI library has been implemented for the Fujitsu AP1000+ multicomputer running the AP/Linux operating system. Under this environment, parallel programs may be dedicated to a fixed partition, or a number of parallel programs may share a partition. Therefore, the MPI library has been constructed so that messaging operations can be driven by polling and/or interrupt techniques. It has been found that polling works well when a single parallel program is running on a given partition, and that interrupt-driven communication makes far better use of the machine when multiple parallel programs are executing. Gang scheduling of multiple parallel programs which use polling was found to be relatively ineffective. 1 Introduction MPI has previously been implemented on both the Fujitsu AP1000 [3] and AP1000+ [2, 4] multicomputers running the CellOS operating system. Under CellOS, when a parallel program is initiated, the machine is reset and the kernel is loaded before launching the pa...

Abstract not found

Least Squares is a standard problem in linear algebra. It can be solved by adapting the algorithm that carries out the QR decomposition. This approach is adequate due to the good numerical stability of the QR factorization. The algorithm has been developed for Fujitsu's AP1000 computer and it takes into account the sparse nature of the matrices to get memory savings and decrease the computation times. We will describe the parallel algorithm which solves this problem and show an evaluation of its efficiency using the Harwell-Boeing matrix collection.

The research that has gone into this thesis has been thoroughly enjoyable. That enjoyment is largely a result of the interaction that I have had with my supervisors, colleagues and, in the case of rsync, the people who have tested and used the resulting software. I feel very privileged to have worked with my supervisors, Richard Brent, Paul Mackerras and Brendan McKay. To each of them I owe a great debt of gratitude for their patience, inspiration and friendship. Richard, Paul and Brendan have taught me a great deal about the field of Computer Science by sharing with me the joy of discovery and investigation that is the heart of research. I would also like to thank Bruce Millar and Iain Macleod, who supervised my initial research in automatic speech recognition. Through Bruce, Iain and the Computer Sciences Laboratory I made the transition from physicist to computer scientist, turning my hobby into a career.