Acceleration of a QM/MM-QMC simulation using GPU

TL;DR

This paper demonstrates significant acceleration of a quantum Monte Carlo simulation using GPU computing, achieving over 20-fold speedup while maintaining acceptable accuracy, and confirms linear scalability across multiple nodes.

Contribution

It introduces GPU acceleration for a QM/MM-QMC simulation, replacing the bottleneck with CUDA to significantly improve performance and scalability.

Findings

23.6 times faster in single precision on GPU

Energy deviation within 10^{-5} hartree accuracy

Linear scalability across GPU nodes

Abstract

We accelerated an ab-initio molecular QMC calculation by using GPGPU. Only the bottle-neck part of the calculation is replaced by CUDA subroutine and performed on GPU. The performance on a (single core CPU + GPU) is compared with that on a (single core CPU with double precision), getting 23.6 (11.0) times faster calculations in single (double) precision treatments on GPU. The energy deviation caused by the single precision treatment was found to be within the accuracy required in the calculation, \sim 10^{-5} hartree. The accelerated computational nodes mounting GPU are combined to form a hybrid MPI cluster on which we confirmed the performance linearly scales to the number of nodes.