Evaluation of disconnected quark loops for hadron structure using GPUs

TL;DR

This paper compares stochastic methods for calculating fermion loops in hadron structure on GPUs, focusing on efficiency, convergence, and noise reduction techniques for observables like the nucleon axial charge and sigma-terms.

Contribution

It evaluates and compares various stochastic methods for fermion loop calculations on GPUs, highlighting their efficiency and noise reduction strategies in hadron structure studies.

Findings

Certain methods show faster convergence on GPUs.

Noise reduction techniques improve statistical accuracy.

Efficient evaluation of sigma-terms with reduced stochastic noise.

Abstract

A number of stochastic methods developed for the calculation of fermion loops are investigated and compared, in particular with respect to their efficiency when implemented on Graphics Processing Units (GPUs). We assess the performance of the various methods by studying the convergence and statistical accuracy obtained for observables that require a large number of stochastic noise vectors, such as the isoscalar nucleon axial charge. The various methods are also examined for the evaluation of sigma-terms where noise reduction techniques specific to the twisted mass formulation can be utilized thus reducing the required number of stochastic noise vectors.