A parallel, iterative method for solving large linear systems

The need for parallel iterative algorithms for solving large linear systems is discussed. A brief background about what from concurrent computer architecture should take is given. The performance of a parallel, iterative program for linear systems which was tested on a variety of linear problems is compared to that of an identical serial linear systems program. For the small-sized systems tested, the parallel program was only slightly superior to the serial program where the system encountered was well conditioned. When an ill-conditioned system was run, however, the parallel processing version of the program was clearly the superior of the two. Some suggestions on how the parallel program could be further improved are given.<
>