GPU-Accelerated and CPU SIMD Optimized Monte Carlo Simulation of boldsymbolphi4 Model

Piotr Białas

Faculty of Physics, Astronomy and Applied Computer Science
Jagiellonian University, ul. Reymonta 4, 30-059 Krakow, Poland
&
Mark Kac Complex Systems Research Centre
Faculty of Physics, Astronomy and Applied Computer Science
Jagiellonian University, ul. Reymonta 4, 30-059 Krakow, Poland
Jakub Kowal

Faculty of Physics, Astronomy and Applied Computer Science
Jagiellonian University, ul. Reymonta 4, 30-059 Krakow, Poland
Adam Strzelecki

Faculty of Physics, Astronomy and Applied Computer Science
Jagiellonian University, ul. Reymonta 4, 30-059 Krakow, Poland

GPU-Accelerated and CPU SIMD Optimized Monte Carlo Simulation of boldsymbolphi4 Model

keywords: GPU computing, vectorization, SIMD, Monte-Carlo simulations

In this contribution we describe an efficient GPU implementation of the Monte-Carlo simulation of the Ginzburg-Landau model. We achieve the performance close to 50 % of the peak performance of the used GPU. We compare this performance with a parallelized and vectorized CPU code and discuss the observed differences.

mathematics subject classification 2000: 65Y05, 65Y10, 82B99

reference: Vol. 33, 2014, No. 5, pp. 1191–1208

Computing and Informatics

formerly Computers and Artificial Intelligence

GPU-Accelerated and CPU SIMD Optimized Monte Carlo Simulation of boldsymbolphi4 Model