Lattice QCD based on OpenCL

TL;DR

This paper introduces a platform-independent OpenCL implementation of Lattice QCD simulations, achieving significant speedups on GPUs and CPUs for pure gauge and fermion calculations.

Contribution

It presents a novel OpenCL-based Lattice QCD application capable of running efficiently on various hardware platforms, including AMD and NVIDIA GPUs and CPUs.

Findings

Double precision dslash achieves 60 GFLOPS on AMD Radeon HD 5870.

Hybrid Monte-Carlo runs four times faster than CPU reference code.

Implementation is platform independent and versatile.

Abstract

We present an OpenCL-based Lattice QCD application using a heatbath algorithm for the pure gauge case and Wilson fermions in the twisted mass formulation. The implementation is platform independent and can be used on AMD or NVIDIA GPUs, as well as on classical CPUs. On the AMD Radeon HD 5870 our double precision dslash implementation performs at 60 GFLOPS over a wide range of lattice sizes. The hybrid Monte-Carlo presented reaches a speedup of four over the reference code running on a server CPU.