Solving lattice QCD systems of equations using mixed precision solvers on GPUs

DOI10.1016/J.CPC.2010.05.002MaRDI QIDQ536950zbMATH OpenOpenAlexFDO

Authors R. Babich, Michael A. Clark, Kipton Barros, R. C. Brower, C. Rebbi

Publication date 31 May 2011

Published in Computer Physics Communications (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/0911.3191

zbMATH Keywords

CUDAGPGPUlattice QCDGPUmixed precision

Mathematics Subject Classification ID

Strong interaction, including quantum chromodynamics (81V05) Quantum field theory on lattices (81T25)

Abstract: Modern graphics hardware is designed for highly parallel numerical tasks and promises significant cost and performance benefits for many scientific applications. One such application is lattice quantum chromodyamics (lattice QCD), where the main computational challenge is to efficiently solve the discretized Dirac equation in the presence of an SU(3) gauge field. Using NVIDIA's CUDA platform we have implemented a Wilson-Dirac sparse matrix-vector product that performs at up to 40 Gflops, 135 Gflops and 212 Gflops for double, single and half precision respectively on NVIDIA's GeForce GTX 280 GPU. We have developed a new mixed precision approach for Krylov solvers using reliable updates which allows for full double precision accuracy while using only single or half precision arithmetic for the bulk of the computation. The resulting BiCGstab and CG solvers run in excess of 100 Gflops and, in terms of iterations until convergence, perform better than the usual defect-correction approach for mixed precision.

Recommendations

• Multi-mass solvers for lattice QCD on GPUs
• SU(2) lattice gauge theory simulations on Fermi GPUs
• Coulomb, Landau and maximally abelian gauge fixing in lattice QCD with multi-gpus
• Lattice QCD based on OpenCL
• A factored sparse approximate inverse preconditioned conjugate gradient solver on graphics processing units

Cites work

• Iterative refinement of the solution of a positive definite system of equations
• Optimizing a lattice QCD simulation program
• Reliable updated residuals in hybrid Bi-CG methods

Cited in

(16)

• Mixed precision algorithms in numerical linear algebra
• The Möbius domain wall fermion algorithm
• Nuclear matrix elements from lattice QCD for electroweak and beyond-standard-model processes
• A new class of AMG interpolation methods based on matrix-matrix multiplications
• Algebraic Error Analysis for Mixed-Precision Multigrid Solvers
• Coulomb, Landau and maximally abelian gauge fixing in lattice QCD with multi-gpus
• Lattice QCD based on OpenCL
• Multi-mass solvers for lattice QCD on GPUs
• SU(2) lattice gauge theory simulations on Fermi GPUs
• Coupled-channel \(D\pi\), \(D\eta\) and \(D_s\overline K\) scattering from lattice QCD
• Pseudo-random number generators for Monte Carlo simulations on ATI graphics processing units
• Mixed-precision conjugate gradient algorithm using the groupwise update strategy
• High-precision numerical simulations on a CUDA GPU: Kerr black hole tails
• Speeding up fermionic lattice calculations with photonic accelerated inverters
• Algebraic Error Analysis for Mixed-Precision Multigrid Solvers
• Towards lattice quantum chromodynamics on FPGA devices

Describes a project that uses

Uses Software

• QUDA
• Chroma
• OpenCL
• BiCGstab
• CUDA

This page was built for publication: Solving lattice QCD systems of equations using mixed precision solvers on GPUs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q536950)