Efficiency of different matrix inversion methods applied to Wilson fermions

Abstract

We compare different conjugate gradient -- like matrix inversion methods (CG, BiCGstab1 and BiCGstab2) employing for this purpose the compact lattice quantum electrodynamics (QED) with Wilson fermions. The main goals of this investigation are the CPU time efficiency of the methods as well as the influence of machine precision on the reliability of (physical) results especially close to the 'critical' line ${\kappa }_c(\text{bt})$.

Comments: 27 pages LaTeX (epsf), all figures included