Experimental Study on Parallel Methods for Solving Systems of Equations

The paper presents experimental results on parallel variants of some classical methods for solving systems of equations. The following four methods are studied: Newton method, Chebyshev method, Gradient method with Fridman control sequence, a method of conjugate gradient type. The main conclusion is that the precondition of the sparse systems (both linear and nonlinear) improves in a great extent the performance of the parallel algorithms.