GPU-based simulation of the long-range Potts model via parallel tempering

TL;DR

This paper demonstrates a GPU-accelerated parallel tempering method for simulating the long-range one-dimensional Potts model, significantly improving computational efficiency and enabling detailed thermodynamic analysis.

Contribution

It introduces an efficient GPU implementation with multispin coding for long-range Potts model simulations, achieving substantial speedups over traditional methods.

Findings

Speedup factors of up to 37 with GPU implementation

Effective analysis of thermodynamic properties and phase transition regimes

Enhanced computational efficiency for long-range interaction models

Abstract

We discuss the efficiency of parallelization on graphical processing units (GPUs) for the simulation of the one dimensional Potts model with long range interactions via parallel tempering. We investigate the behaviour of some thermodynamic properties, such as equilibrium energy and magnetization, critical temperatures as well as the separation between the first- and second-order regime. By implementing multispin coding techniques and an efficient parallelization of the interaction energy computation among threads, the GPU-accelerated approach reached speedup factors of up to 37.