The Householder transformation on parallel computers with dynamic ring architecture has been investigated. Innovative systolic-like communication technique is proposed which eliminate the transposing of a matrix. The experiments on CM-5 supercomputer have been done.
The design of optimal array processors for parallel computing of the two-dimensional discrete Fourier transform has been done. We proposed three different schemes for designing array processors using systematic approach. Efficient algorithms have been developed for both compile and run time task scheduling. Experimental results have been obtained, showing that both of our algorithms are superior to existing approaches.