Energy Performance Evaluation of Quasi-Monte Carlo Algorithms on Hybrid HPC

The increasing demands of scientific applications and the increasing capacity of modern computing systems lead to the need of evaluating energy consumption and, consequently, to the development of energy efficient algorithms. In this paper we study the energy performance of a class of quasi-Monte Carlo algorithms on hybrid HPC systems. These algorithms are applied to solve quantum kinetic integral equations using Sobol and Halton sequences. The energy performance results are compared on a CPU-based computer platform and computer platforms with accelerators like GPU cards and Intel Xeon Phi coprocessors with respect to several metrics. Directions for future work are also given.