Multigrid Solver on Fugaku
Ken-Ichi Ishikawa, Issaku Kanamori, Hideo Matsufuru

TL;DR
This paper presents a highly optimized multigrid solver for Clover fermions on the Fugaku supercomputer, demonstrating good scalability and performance using the Bridge++ framework and QWS library.
Contribution
It introduces a novel implementation of a multigrid solver utilizing the QWS library's domain decomposed preconditioner as a smoother on Fugaku.
Findings
The solver scales well on Fugaku's architecture.
It achieves improved performance over previous implementations.
The code demonstrates effective use of A64FX CPU features.
Abstract
We report an implementation of a multigrid solver for the Clover fermion on supercomputer Fugaku, which uses A64FX CPU with Arm architecture. On Fugaku, a highly optimized implementation of BiCGStab solver with domain decomposed preconditioner for the Clover fermion, called QCD Wide SIMD library (QWS), is available. We use the preconditioner in QWS as a smoother of the multigrid solver. Our implementation shows reasonably good scaling and performance. The code is developed by using Bridge++ code framework and its extension.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsParticle physics theoretical and experimental studies · Quantum Chromodynamics and Particle Interactions · Parallel Computing and Optimization Techniques
