Kokkos software runs lattice gauge simulations efficiently on diverse supercomputers
Performance Portable $\mathrm{SU}(N)$ Lattice Gauge Theory Simulation with Kokkos
Distributed, Parallel, and Cluster Computing
Summary
Lattice gauge theory simulations, important in physics, often run on many types of computers but need different software for each. The authors created one program using Kokkos that works well on many computer types without rewriting the code. Their program handles complex math for physics simulations and runs fast on GPUs and CPUs, including top supercomputers. This helps researchers run these simulations more easily across different machines.
What this means in practice
- •For high performance computing teams: Run advanced lattice gauge physics simulations efficiently across diverse supercomputer architectures without needing separate implementations for each hardware type.
- •For scientific software developers: Develop and maintain physics simulation software using a single source code that targets multiple CPU and GPU platforms with Kokkos for better portability and performance.
Authors
Wei Sun
Abstract
The increasing diversity of high performance computing systems makes separate, architecture specific implementations of lattice gauge theory algorithms costly to maintain. We present \texttt{kwqft}, a performance portable Kokkos implementation of Wilson pure gauge Monte Carlo simulation for $\mathrm{SU}(N)$ Yang-Mills theory in an arbitrary number of space-time dimensions. The gauge group order $N$ and the dimension $D$ are compile time parameters. A single source targets the Serial, OpenMP, CUDA, HIP, and SYCL execution spaces, with MPI halo exchange overlapped with interior updates. The implementation reproduces the exact two-dimensional plaquette and published three and four dimensional values for gauge groups up to $\mathrm{SU}(17)$. On an NVIDIA A100 the Kokkos CUDA backend is competitive with a native CUDA code, SIMD acceleration improves the OpenMP path on Armv9 processors, and a large scale speedup is demonstrated for an $\mathrm{SU}(4)$ lattice on the LineShine supercomputer, currently ranked first on the TOP500 list.