An iterative method for large sparse linear systems on a vector computer (Q915363): Difference between revisions

Theoretical investigation of the ``method of the arithmetic mean'': to solve \(Ax=b,\) \(A=L+D+U,\) solve simultaneously equations of type \((L+D)x_ 1=b-Ux_{old}\) and \((L+U)x_ 2=b-Lx_{old},\) \(x_{new}=(x_ 1+x_ 2)/2.\) For the solution of a 2-D elliptic partial differential equation the number of iterations of the method of the arithmetic mean and of the biconjugate gradient method (BICO) are compared: depending on the coefficients of the partial differential equation one or the other method is better, but I consider this comparison of no value without computation times because BICO can be vectorized excellently, whereas the solution of triangular systems is strongly recursive. Another example is presented for the solution of an inverse eigenproblem.