next up previous contents index
Next: Computational Aspects of Up: Chebyshev Iteration Previous: Convergence



In Chebyshev Iteration the iteration parameters are known as soon as one knows the ellipse containing the eigenvalues (or rather, the field of values) of the operator. Therefore the computation of inner products, as is necessary in methods like GMRES or CG, is avoided . This avoids the synchronization points required of CG-type methods, so machines with hierarchical or distributed memory may achieve higher performance (it also suggests strong parallelization properties ; for a discussion of this see Saad [185], and Dongarra, et al. [71]). Specifically, as soon as some segment of is computed, we may begin computing, in sequence, corresponding segments of , , and .

Figure: The Preconditioned Chebyshev Method

The pseudocode for the Preconditioned Chebyshev Method with preconditioner is given in Figure gif. It handles the case of a symmetric positive definite coefficient matrix . The eigenvalues of are assumed to be all real and in the interval , which does not include zero.


Jack Dongarra
Mon Nov 20 08:52:54 EST 1995