Systolic designs for Aitken's root finding method (Q1195177)

In recent years design and analysis of systolic algorithms has been applied to the provision of high performance architecture with regular structure and rhythmic movement of data. But only little on mapping nonlinear problems onto systolic arrays is known. In the present paper the authors focus their interest on designing a systolic array in order to solve single nonlinear (polynomial or transcendental) equations. Main ingredient is Aitken's table generating method which produces a doubly infinite table whose column entries converge to the solutions of the equation in question. The authors establish a nonlinear scheduling function allowing the table to be partitioned into finite-size blocks. Then using standard synthesis techniques the whole table can be produced iteratively block by block. The paper includes some hints how to improve the described algorithm.