Multi-mass solvers for lattice QCD on GPUs

Authors:
A. Alexandru;C. Pelissier;B. Gamari;F. X. Lee
Affiliations:
Department of Physics, The George Washington University, Washington, DC 20052, United States;Department of Physics, The George Washington University, Washington, DC 20052, United States;Department of Physics, The George Washington University, Washington, DC 20052, United States;Department of Physics, The George Washington University, Washington, DC 20052, United States
Venue:
Journal of Computational Physics
Year:
2012

Citing 3
Cited 0

BI-CGSTAB: a fast and smoothly converging variant of BI-CG for the solution of nonsymmetric linear systems

SIAM Journal on Scientific and Statistical Computing
Iterative Methods for Sparse Linear Systems

Iterative Methods for Sparse Linear Systems
Efficient Implementation of the Overlap Operator on Multi-GPUs

SAAHPC '11 Proceedings of the 2011 Symposium on Application Accelerators in High-Performance Computing

Quantified Score

Hi-index	31.45

Visualization

Abstract

Graphical Processing Units (GPUs) are more and more frequently used for lattice QCD calculations. Lattice studies often require computing the quark propagators for several masses. These systems can be solved using multi-shift solvers but these algorithms are memory intensive which limits the size of the problem that can be solved using GPUs. In this paper, we show how to efficiently use a memory-lean single-mass solver to solve multi-mass problems. We focus on the BiCGstab algorithm for Wilson fermions and show that the single-mass solver not only requires less memory but also outperforms the multi-shift variant by a factor of two.