Projection of the low-lying eigenmodes of the overlap Dirac operator in lattice QCD

TL;DR

This paper presents an implementation of an adaptive Lanczos algorithm for efficiently computing low-lying eigenmodes of the overlap Dirac operator in lattice QCD, comparing its performance with existing methods.

Contribution

The authors develop and evaluate an adaptive thick-restart Lanczos algorithm tailored for lattice QCD eigenmode projection, offering potential performance improvements.

Findings

The $a$-TRLan algorithm shows competitive performance with ARPACK.

Implementation details improve eigenmode projection efficiency.

Comparison results demonstrate advantages in specific lattice QCD scenarios.

Abstract

We outline our implementation of the adaptive thick-restart Lanczos algorithm ($a$-TRLan) for the projection of the low-lying eigenmodes of the overlap Dirac operator in lattice QCD, and compare the performances of our code and the widely used package ARPACK.