The overlap lattice-Dirac operator contains the sign function ϵ(H). Recent practical implementations replace ϵ(H) by a ratio of polynomials, HPn(H2)/Qn(H2), and require storage of 2n+2 large vectors. Here I show that one can use only 4 large vectors at the cost of executing the core conjugate algorithm twice. The slow-down might be less than by a factor of 2, depending on the architecture of the computer one uses.
Cite
@article{arxiv.hep-lat/9811019,
title = {Minimizing storage in implementations of the overlap lattice-Dirac operator},
author = {Herbert Neuberger},
journal= {arXiv preprint arXiv:hep-lat/9811019},
year = {2015}
}