Performance Portability Strategies for Grid C++ Expression Templates

TL;DR

This paper explores strategies for achieving performance portability of Grid C++ expression templates across various architectures, focusing on GPU offloading with CUDA, OpenACC, and JIT, highlighting successes and challenges encountered.

Contribution

It presents new GPU offloading strategies for Grid C++ expression templates, including experimental results and analysis of using CUDA, OpenACC, and JIT compilation.

Findings

Successful GPU offloading with CUDA and OpenACC

Challenges with OpenMP 4.x for GPU offloading

Performance results on GPUs with SU(3)×SU(3) streaming test

Abstract

One of the key requirements for the Lattice QCD Application Development as part of the US Exascale Computing Project is performance portability across multiple architectures. Using the Grid C++ expression template as a starting point, we report on the progress made with regards to the Grid GPU offloading strategies. We present both the successes and issues encountered in using CUDA, OpenACC and Just-In-Time compilation. Experimentation and performance on GPUs with a SU(3)$\times$SU(3) streaming test will be reported. We will also report on the challenges of using current OpenMP 4.x for GPU offloading in the same code.