Systolic ring for parallel computation
Omar Wing · 1985
A computation structure suitable for the iterative solution of linear equations is described. The structure is a ring of processors interspersed with latches. Each processor is connected to a private memory where the elements of a row of the matrix of the equations are stored. We show how the Gauss-Seidel algorithm can be mapped onto such a ring and how an ideal speed-up can be achieved.