Algorithmic bombardment for the iterative solution of linear systems: A poly-iterative approach (Q2564251)

The paper describes the combination of three iterative methods for solving nonsymmetric linear systems, namely CGS, BiCGSTAB and QMR, in an algorithmic bombardment approach. Since each iterative solver mentioned above suffers from possible breakdowns in the case of nonsymmetric linear systems of equations, the poly-iterative algorithm applies suitable methods simultaneously on the data set, removes methods from the process that break down and terminates the whole process when one method has converged. The authors show through various experiments that this increases the chance of finding the solution, and in parallel environment this does not increase threefold the time to find the solution. In fact, even when all three algorithms would have found the solution, bombardment may be faster than the slowest of the three.