Acceleration of Diagrammatic Determinantal Quantum Monte Carlo Calculations using GPUs
Georg-August-Universiẗat G̈ottingen, Institut f̈ur Theoretische Physik; The University of Edinburgh
@article{schmitt2014acceleration,
title={Acceleration of Diagrammatic Determinantal Quantum Monte Carlo Calculations using GPUs},
author={Schmitt, Markus and Bethune, Iain and Haase, Patrick and Pruschke, Thomas},
year={2014}
}
Diagrammatic Determinantal Quantum Monte Carlo (DDQMC) algorithms are used to solve quantum impurity models such as the Anderson model. The calculation of acceptance rates and observables during the Monte Carlo walk involves linear algebra operations whose computational expense increases with decreasing temperature. Thus, the lower boundary of the treatable temperature range is limited by the available compute capacity. In order to make use of GPUs as cheap and powerful accelerators parts of a DDQMC code (CT-INT, [GML+11]) were ported to CUDA.
April 2, 2014 by hgpu
