TWQCD's dynamical DWF project

TL;DR

This paper describes the TWQCD project's GPU-accelerated unquenched lattice QCD simulations using optimal domain-wall quarks, detailing the computational setup, ongoing production runs, and preliminary results on pseudoscalar properties.

Contribution

It introduces a high-performance GPU cluster for unquenched lattice QCD with optimal domain-wall quarks and reports on the simulation setup and initial findings.

Findings

Simulation of two-flavor QCD on large lattices is underway.

Preliminary results on pseudoscalar mass and decay constant are presented.

Achieved sustained computing power of 15.36 Teraflops.

Abstract

We present an overview of our project of simulation of unquenched lattice QCD with optimal domain-wall quarks, using a GPU cluster currently constituting of 16 units of Nvidia Tesla S1070 plus 64 graphic cards with Nvidia GTX285 (total 128 GPUs with 128 Teraflops peak), attaining sustained computing power of 15.36 Teraflops. The first production run in two-flavor QCD is on-going, using the Iwasaki gauge action on a set of lattices with sizes $ 16^3 \times (32,10,8,6,4) \times (16,32) $ at the lattice spacing $ a \sim 0.1$ fm, with eight sea quark masses down to $ m_\pi \simeq 200 $ MeV. We outline our simulation algorithm, and describe the present status of the production run. Preliminary results of pseudoscalar mass and decay constant are also presented.