Grid: OneCode and FourAPIs

TL;DR

This paper presents an updated Grid library for Lattice QCD that achieves cross-platform performance portability across multiple GPU architectures and CPUs, supporting CUDA, HIP, and SYCL environments.

Contribution

The work introduces an internal abstraction interface that enables single-source code to run efficiently on diverse GPU and CPU architectures, including future Exascale systems.

Findings

Supports Nvidia CUDA, AMD HIP, and SYCL environments

Enables performance portability across multiple architectures

Retains CPU vectorisation and SIMD execution within OpenMP

Abstract

We discuss a substantial update to the Grid software library for Lattice QCD, enabling it to port to multiple GPU architectures while retaining CPU vectorisation and SIMD execution within OpenMP threads. The GPU environments supported include vendor specific Nvidia CUDA and AMD HIP environments and a (mostly) standards based SYCL implementation. This is performed by an internal abstraction interface giving single source cross-platform performance portability across all number of planned Exascale architectures, and all those planned by the US Department of Energy.